ti.sil/open-webui
1
0
Fork 0
mirror of https://github.com/open-webui/open-webui.git synced 2025-12-12 04:15:25 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 4

Merge branch 'main' into feat/google-oauth-groups-dev

Browse source 
# Conflicts:
#	backend/open_webui/utils/oauth.py
#	uv.lock
This commit is contained in:
Luke Garceau 2025-11-27 16:53:12 -05:00
commit 89a5dbda45
641 changed files with 95650 additions and 34849 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

19
.github/ISSUE_TEMPLATE/bug_report.yaml vendored
View file

@ -11,7 +11,9 @@ body:
## Important Notes
- **Before submitting a bug report**: Please check the [Issues](https://github.com/open-webui/open-webui/issues) or [Discussions](https://github.com/open-webui/open-webui/discussions) sections to see if a similar issue has already been reported. If unsure, start a discussion first, as this helps us efficiently focus on improving the project.
- **Before submitting a bug report**: Please check the [Issues](https://github.com/open-webui/open-webui/issues) and [Discussions](https://github.com/open-webui/open-webui/discussions) sections to see if a similar issue has already been reported. If unsure, start a discussion first, as this helps us efficiently focus on improving the project. Duplicates may be closed without notice. **Please search for existing issues AND discussions. No matter open or closed.**
- Check for opened, **but also for (recently) CLOSED issues** as the issue you are trying to report **might already have been fixed on the dev branch!**
- **Respectful collaboration**: Open WebUI is a volunteer-driven project with a single maintainer and contributors who also have full-time jobs. Please be constructive and respectful in your communication.
@ -19,13 +21,19 @@ body:
- **Bug Reproducibility**: If a bug cannot be reproduced using a `:main` or `:dev` Docker setup or with `pip install` on Python 3.11, community assistance may be required. In such cases, we will move it to the "[Issues](https://github.com/open-webui/open-webui/discussions/categories/issues)" Discussions section. Your help is appreciated!
- **Scope**: If you want to report a SECURITY VULNERABILITY, then do so through our [GitHub security page](https://github.com/open-webui/open-webui/security).
- type: checkboxes
id: issue-check
attributes:
label: Check Existing Issues
description: Confirm that you’ve checked for existing reports before submitting a new one.
options:
- label: I have searched the existing issues and discussions.
- label: I have searched for any existing and/or related issues.
required: true
- label: I have searched for any existing and/or related discussions.
required: true
- label: I have also searched in the CLOSED issues AND CLOSED discussions and found no related items (your issue might already be addressed on the development branch!).
required: true
- label: I am using the latest version of Open WebUI.
required: true
@ -47,7 +55,7 @@ body:
id: open-webui-version
attributes:
label: Open WebUI Version
description: Specify the version (e.g., v0.3.11)
description: Specify the version (e.g., v0.6.26)
validations:
required: true
@ -63,7 +71,7 @@ body:
id: operating-system
attributes:
label: Operating System
description: Specify the OS (e.g., Windows 10, macOS Sonoma, Ubuntu 22.04)
description: Specify the OS (e.g., Windows 10, macOS Sonoma, Ubuntu 22.04, Debian 12)
validations:
required: true
@ -126,6 +134,7 @@ body:
description: |
Please provide a **very detailed, step-by-step guide** to reproduce the issue. Your instructions should be so clear and precise that anyone can follow them without guesswork. Include every relevant detail—settings, configuration options, exact commands used, values entered, and any prerequisites or environment variables.
**If full reproduction steps and all relevant settings are not provided, your issue may not be addressed.**
**If your steps to reproduction are incomplete, lacking detail or not reproducible, your issue can not be addressed.**
placeholder: |
Example (include every detail):
@ -163,5 +172,5 @@ body:
attributes:
value: |
## Note
If the bug report is incomplete or does not follow instructions, it may not be addressed. Ensure that you've followed all the **README.md** and **troubleshooting.md** guidelines, and provide all necessary information for us to reproduce the issue.
**If the bug report is incomplete, does not follow instructions or is lacking details it may not be addressed.** Ensure that you've followed all the **README.md** and **troubleshooting.md** guidelines, and provide all necessary information for us to reproduce the issue.
Thank you for contributing to Open WebUI!

26
.github/ISSUE_TEMPLATE/feature_request.yaml vendored
View file

@ -8,8 +8,19 @@ body:
value: |
## Important Notes
### Before submitting
Please check the [Issues](https://github.com/open-webui/open-webui/issues) or [Discussions](https://github.com/open-webui/open-webui/discussions) to see if a similar request has been posted.
Please check the **open AND closed** [Issues](https://github.com/open-webui/open-webui/issues) AND [Discussions](https://github.com/open-webui/open-webui/discussions) to see if a similar request has been posted.
It's likely we're already tracking it! If you’re unsure, start a discussion post first.
#### Scope
If your feature request is likely to take more than a quick coding session to implement, test and verify, then open it in the **Ideas** section of the [Discussions](https://github.com/open-webui/open-webui/discussions) instead.
**We will close and force move your feature request to the Ideas section, if we believe your feature request is not trivial/quick to implement.**
This is to ensure the issues tab is used only for issues, quickly addressable feature requests and tracking tickets by the maintainers.
Other feature requests belong in the **Ideas** section of the [Discussions](https://github.com/open-webui/open-webui/discussions).
If your feature request might impact others in the community, definitely open a discussion instead and evaluate whether and how to implement it.
This will help us efficiently focus on improving the project.
### Collaborate respectfully
@ -22,7 +33,6 @@ body:
We appreciate your time and ask that you **respect ours**.
### Contributing
If you encounter an issue, we highly encourage you to submit a pull request or fork the project. We actively work to prevent contributor burnout to maintain the quality and continuity of Open WebUI.
@ -35,14 +45,22 @@ body:
label: Check Existing Issues
description: Please confirm that you've checked for existing similar requests
options:
- label: I have searched the existing issues and discussions.
- label: I have searched for all existing **open AND closed** issues and discussions for similar requests. I have found none that is comparable to my request.
required: true
- type: checkboxes
id: feature-scope
attributes:
label: Verify Feature Scope
description: Please confirm the feature's scope is within the described scope
options:
- label: I have read through and understood the scope definition for feature requests in the Issues section. I believe my feature request meets the definition and belongs in the Issues section instead of the Discussions.
required: true
- type: textarea
id: problem-description
attributes:
label: Problem Description
description: Is your feature request related to a problem? Please provide a clear and concise description of what the problem is.
placeholder: "Ex. I'm always frustrated when..."
placeholder: "Ex. I'm always frustrated when... / Not related to a problem"
validations:
required: true
- type: textarea

6
.github/dependabot.yml vendored
View file

@ -12,12 +12,6 @@ updates:
interval: monthly
target-branch: 'dev'
- package-ecosystem: npm
directory: '/'
schedule:
interval: monthly
target-branch: 'dev'
- package-ecosystem: 'github-actions'
directory: '/'
schedule:

20
.github/pull_request_template.md vendored
View file

@ -1,17 +1,20 @@
# Pull Request Checklist
### Note to first-time contributors: Please open a discussion post in [Discussions](https://github.com/open-webui/open-webui/discussions) and describe your changes before submitting a pull request.
### Note to first-time contributors: Please open a discussion post in [Discussions](https://github.com/open-webui/open-webui/discussions) to discuss your idea/fix with the community before creating a pull request, and describe your changes before submitting a pull request.
This is to ensure large feature PRs are discussed with the community first, before starting work on it. If the community does not want this feature or it is not relevant for Open WebUI as a project, it can be identified in the discussion before working on the feature and submitting the PR.
**Before submitting, make sure you've checked the following:**
- [ ] **Target branch:** Please verify that the pull request targets the `dev` branch.
- [ ] **Description:** Provide a concise description of the changes made in this pull request.
- [ ] **Target branch:** Verify that the pull request targets the `dev` branch. **Not targeting the `dev` branch will lead to immediate closure of the PR.**
- [ ] **Description:** Provide a concise description of the changes made in this pull request down below.
- [ ] **Changelog:** Ensure a changelog entry following the format of [Keep a Changelog](https://keepachangelog.com/) is added at the bottom of the PR description.
- [ ] **Documentation:** Have you updated relevant documentation [Open WebUI Docs](https://github.com/open-webui/docs), or other documentation sources?
- [ ] **Documentation:** If necessary, update relevant documentation [Open WebUI Docs](https://github.com/open-webui/docs) like environment variables, the tutorials, or other documentation sources.
- [ ] **Dependencies:** Are there any new dependencies? Have you updated the dependency versions in the documentation?
- [ ] **Testing:** Have you written and run sufficient tests to validate the changes?
- [ ] **Testing:** Perform manual tests to **verify the implemented fix/feature works as intended AND does not break any other functionality**. Take this as an opportunity to **make screenshots of the feature/fix and include it in the PR description**.
- [ ] **Agentic AI Code:** Confirm this Pull Request is **not written by any AI Agent** or has at least **gone through additional human review AND manual testing**. If any AI Agent is the co-author of this PR, it may lead to immediate closure of the PR.
- [ ] **Code review:** Have you performed a self-review of your code, addressing any coding standard issues and ensuring adherence to the project's coding standards?
- [ ] **Prefix:** To clearly categorize this pull request, prefix the pull request title using one of the following:
- [ ] **Title Prefix:** To clearly categorize this pull request, prefix the pull request title using one of the following:
- **BREAKING CHANGE**: Significant changes that may affect compatibility
- **build**: Changes that affect the build system or external dependencies
- **ci**: Changes to our continuous integration processes or workflows
@ -73,4 +76,7 @@
### Contributor License Agreement
By submitting this pull request, I confirm that I have read and fully agree to the [Contributor License Agreement (CLA)](/CONTRIBUTOR_LICENSE_AGREEMENT), and I am providing my contributions under its terms.
By submitting this pull request, I confirm that I have read and fully agree to the [Contributor License Agreement (CLA)](https://github.com/open-webui/open-webui/blob/main/CONTRIBUTOR_LICENSE_AGREEMENT), and I am providing my contributions under its terms.
> [!NOTE]
> Deleting the CLA section will lead to immediate closure of your PR and it will not be merged in.

6
.github/workflows/build-release.yml vendored
View file

@ -11,7 +11,7 @@ jobs:
steps:
- name: Checkout repository
uses: actions/checkout@v4
uses: actions/checkout@v5
- name: Check for changes in package.json
run: |
@ -36,7 +36,7 @@ jobs:
echo "::set-output name=content::$CHANGELOG_ESCAPED"
- name: Create GitHub release
uses: actions/github-script@v7
uses: actions/github-script@v8
with:
github-token: ${{ secrets.GITHUB_TOKEN }}
script: |
@ -61,7 +61,7 @@ jobs:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
- name: Trigger Docker build workflow
uses: actions/github-script@v7
uses: actions/github-script@v8
with:
script: |
github.rest.actions.createWorkflowDispatch({

2
.github/workflows/deploy-to-hf-spaces.yml vendored
View file

@ -27,7 +27,7 @@ jobs:
HF_TOKEN: ${{ secrets.HF_TOKEN }}
steps:
- name: Checkout repository
uses: actions/checkout@v4
uses: actions/checkout@v5
with:
lfs: true

180
.github/workflows/docker-build.yaml vendored
View file

@ -43,7 +43,7 @@ jobs:
echo "PLATFORM_PAIR=${platform//\//-}" >> $GITHUB_ENV
- name: Checkout repository
uses: actions/checkout@v4
uses: actions/checkout@v5
- name: Set up QEMU
uses: docker/setup-qemu-action@v3
@ -141,8 +141,11 @@ jobs:
platform=${{ matrix.platform }}
echo "PLATFORM_PAIR=${platform//\//-}" >> $GITHUB_ENV
- name: Delete huge unnecessary tools folder
run: rm -rf /opt/hostedtoolcache
- name: Checkout repository
uses: actions/checkout@v4
uses: actions/checkout@v5
- name: Set up QEMU
uses: docker/setup-qemu-action@v3
@ -243,8 +246,11 @@ jobs:
platform=${{ matrix.platform }}
echo "PLATFORM_PAIR=${platform//\//-}" >> $GITHUB_ENV
- name: Delete huge unnecessary tools folder
run: rm -rf /opt/hostedtoolcache
- name: Checkout repository
uses: actions/checkout@v4
uses: actions/checkout@v5
- name: Set up QEMU
uses: docker/setup-qemu-action@v3
@ -347,7 +353,7 @@ jobs:
echo "PLATFORM_PAIR=${platform//\//-}" >> $GITHUB_ENV
- name: Checkout repository
uses: actions/checkout@v4
uses: actions/checkout@v5
- name: Set up QEMU
uses: docker/setup-qemu-action@v3
@ -419,6 +425,108 @@ jobs:
if-no-files-found: error
retention-days: 1
build-slim-image:
runs-on: ${{ matrix.runner }}
permissions:
contents: read
packages: write
strategy:
fail-fast: false
matrix:
include:
- platform: linux/amd64
runner: ubuntu-latest
- platform: linux/arm64
runner: ubuntu-24.04-arm
steps:
# GitHub Packages requires the entire repository name to be in lowercase
# although the repository owner has a lowercase username, this prevents some people from running actions after forking
- name: Set repository and image name to lowercase
run: |
echo "IMAGE_NAME=${IMAGE_NAME,,}" >>${GITHUB_ENV}
echo "FULL_IMAGE_NAME=ghcr.io/${IMAGE_NAME,,}" >>${GITHUB_ENV}
env:
IMAGE_NAME: '${{ github.repository }}'
- name: Prepare
run: |
platform=${{ matrix.platform }}
echo "PLATFORM_PAIR=${platform//\//-}" >> $GITHUB_ENV
- name: Checkout repository
uses: actions/checkout@v5
- name: Set up QEMU
uses: docker/setup-qemu-action@v3
- name: Set up Docker Buildx
uses: docker/setup-buildx-action@v3
- name: Log in to the Container registry
uses: docker/login-action@v3
with:
registry: ${{ env.REGISTRY }}
username: ${{ github.actor }}
password: ${{ secrets.GITHUB_TOKEN }}
- name: Extract metadata for Docker images (slim tag)
id: meta
uses: docker/metadata-action@v5
with:
images: ${{ env.FULL_IMAGE_NAME }}
tags: |
type=ref,event=branch
type=ref,event=tag
type=sha,prefix=git-
type=semver,pattern={{version}}
type=semver,pattern={{major}}.{{minor}}
type=raw,enable=${{ github.ref == 'refs/heads/main' }},prefix=,suffix=,value=slim
flavor: |
latest=${{ github.ref == 'refs/heads/main' }}
suffix=-slim,onlatest=true
- name: Extract metadata for Docker cache
id: cache-meta
uses: docker/metadata-action@v5
with:
images: ${{ env.FULL_IMAGE_NAME }}
tags: |
type=ref,event=branch
${{ github.ref_type == 'tag' && 'type=raw,value=main' || '' }}
flavor: |
prefix=cache-slim-${{ matrix.platform }}-
latest=false
- name: Build Docker image (slim)
uses: docker/build-push-action@v5
id: build
with:
context: .
push: true
platforms: ${{ matrix.platform }}
labels: ${{ steps.meta.outputs.labels }}
outputs: type=image,name=${{ env.FULL_IMAGE_NAME }},push-by-digest=true,name-canonical=true,push=true
cache-from: type=registry,ref=${{ steps.cache-meta.outputs.tags }}
cache-to: type=registry,ref=${{ steps.cache-meta.outputs.tags }},mode=max
build-args: |
BUILD_HASH=${{ github.sha }}
USE_SLIM=true
- name: Export digest
run: |
mkdir -p /tmp/digests
digest="${{ steps.build.outputs.digest }}"
touch "/tmp/digests/${digest#sha256:}"
- name: Upload digest
uses: actions/upload-artifact@v4
with:
name: digests-slim-${{ env.PLATFORM_PAIR }}
path: /tmp/digests/*
if-no-files-found: error
retention-days: 1
merge-main-images:
runs-on: ubuntu-latest
needs: [build-main-image]
@ -433,7 +541,7 @@ jobs:
IMAGE_NAME: '${{ github.repository }}'
- name: Download digests
uses: actions/download-artifact@v4
uses: actions/download-artifact@v5
with:
pattern: digests-main-*
path: /tmp/digests
@ -487,7 +595,7 @@ jobs:
IMAGE_NAME: '${{ github.repository }}'
- name: Download digests
uses: actions/download-artifact@v4
uses: actions/download-artifact@v5
with:
pattern: digests-cuda-*
path: /tmp/digests
@ -543,7 +651,7 @@ jobs:
IMAGE_NAME: '${{ github.repository }}'
- name: Download digests
uses: actions/download-artifact@v4
uses: actions/download-artifact@v5
with:
pattern: digests-cuda126-*
path: /tmp/digests
@ -599,7 +707,7 @@ jobs:
IMAGE_NAME: '${{ github.repository }}'
- name: Download digests
uses: actions/download-artifact@v4
uses: actions/download-artifact@v5
with:
pattern: digests-ollama-*
path: /tmp/digests
@ -640,3 +748,59 @@ jobs:
- name: Inspect image
run: |
docker buildx imagetools inspect ${{ env.FULL_IMAGE_NAME }}:${{ steps.meta.outputs.version }}
merge-slim-images:
runs-on: ubuntu-latest
needs: [build-slim-image]
steps:
# GitHub Packages requires the entire repository name to be in lowercase
# although the repository owner has a lowercase username, this prevents some people from running actions after forking
- name: Set repository and image name to lowercase
run: |
echo "IMAGE_NAME=${IMAGE_NAME,,}" >>${GITHUB_ENV}
echo "FULL_IMAGE_NAME=ghcr.io/${IMAGE_NAME,,}" >>${GITHUB_ENV}
env:
IMAGE_NAME: '${{ github.repository }}'
- name: Download digests
uses: actions/download-artifact@v5
with:
pattern: digests-slim-*
path: /tmp/digests
merge-multiple: true
- name: Set up Docker Buildx
uses: docker/setup-buildx-action@v3
- name: Log in to the Container registry
uses: docker/login-action@v3
with:
registry: ${{ env.REGISTRY }}
username: ${{ github.actor }}
password: ${{ secrets.GITHUB_TOKEN }}
- name: Extract metadata for Docker images (default slim tag)
id: meta
uses: docker/metadata-action@v5
with:
images: ${{ env.FULL_IMAGE_NAME }}
tags: |
type=ref,event=branch
type=ref,event=tag
type=sha,prefix=git-
type=semver,pattern={{version}}
type=semver,pattern={{major}}.{{minor}}
type=raw,enable=${{ github.ref == 'refs/heads/main' }},prefix=,suffix=,value=slim
flavor: |
latest=${{ github.ref == 'refs/heads/main' }}
suffix=-slim,onlatest=true
- name: Create manifest list and push
working-directory: /tmp/digests
run: |
docker buildx imagetools create $(jq -cr '.tags | map("-t " + .) | join(" ")' <<< "$DOCKER_METADATA_OUTPUT_JSON") \
$(printf '${{ env.FULL_IMAGE_NAME }}@sha256:%s ' *)
- name: Inspect image
run: |
docker buildx imagetools inspect ${{ env.FULL_IMAGE_NAME }}:${{ steps.meta.outputs.version }}

4
.github/workflows/format-backend.yaml vendored
View file

@ -30,10 +30,10 @@ jobs:
- 3.12.x
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v5
- name: Set up Python
uses: actions/setup-python@v5
uses: actions/setup-python@v6
with:
python-version: '${{ matrix.python-version }}'

12
.github/workflows/format-build-frontend.yaml vendored
View file

@ -24,15 +24,15 @@ jobs:
runs-on: ubuntu-latest
steps:
- name: Checkout Repository
uses: actions/checkout@v4
uses: actions/checkout@v5
- name: Setup Node.js
uses: actions/setup-node@v4
uses: actions/setup-node@v5
with:
node-version: '22'
- name: Install Dependencies
run: npm install
run: npm install --force
- name: Format Frontend
run: npm run format
@ -51,15 +51,15 @@ jobs:
runs-on: ubuntu-latest
steps:
- name: Checkout Repository
uses: actions/checkout@v4
uses: actions/checkout@v5
- name: Setup Node.js
uses: actions/setup-node@v4
uses: actions/setup-node@v5
with:
node-version: '22'
- name: Install Dependencies
run: npm ci
run: npm ci --force
- name: Run vitest
run: npm run test:frontend

6
.github/workflows/release-pypi.yml vendored
View file

@ -16,15 +16,15 @@ jobs:
id-token: write
steps:
- name: Checkout repository
uses: actions/checkout@v4
uses: actions/checkout@v5
with:
fetch-depth: 0
- name: Install Git
run: sudo apt-get update && sudo apt-get install -y git
- uses: actions/setup-node@v4
- uses: actions/setup-node@v5
with:
node-version: 22
- uses: actions/setup-python@v5
- uses: actions/setup-python@v6
with:
python-version: 3.11
- name: Build

5
.gitignore vendored
View file

@ -1,3 +1,5 @@
x.py
yarn.lock
.DS_Store
node_modules
/build
@ -12,7 +14,8 @@ vite.config.ts.timestamp-*
__pycache__/
*.py[cod]
*$py.class
.nvmrc
CLAUDE.md
# C extensions
*.so

954
CHANGELOG.md
View file

@ -5,6 +5,960 @@ All notable changes to this project will be documented in this file.
The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
## [0.6.40] - 2025-11-25
### Fixed
- 🗄️ A critical PostgreSQL user listing performance issue was resolved by removing a redundant count operation that caused severe database slowdowns and potential timeouts when viewing user lists in admin panels.
## [0.6.39] - 2025-11-25
### Added
- 💬 A user list modal was added to channels, displaying all users with access and featuring search, sorting, and pagination capabilities. [Commit](https://github.com/open-webui/open-webui/commit/c0e120353824be00a2ef63cbde8be5d625bd6fd0)
- 💬 Channel navigation now displays the total number of users with access to the channel. [Commit](https://github.com/open-webui/open-webui/commit/3b5710d0cd445cf86423187f5ee7c40472a0df0b)
- 🔌 Tool servers and MCP connections now support function name filtering, allowing administrators to selectively enable or block specific functions using allow/block lists. [Commit](https://github.com/open-webui/open-webui/commit/743199f2d097ae1458381bce450d9025a0ab3f3d)
- ⚡ A toggle to disable parallel embedding processing was added via "ENABLE_ASYNC_EMBEDDING", allowing sequential processing for rate-limited or resource-constrained local embedding setups. [#19444](https://github.com/open-webui/open-webui/pull/19444)
- 🔄 Various improvements were implemented across the frontend and backend to enhance performance, stability, and security.
- 🌐 Localization improvements were made for German (de-DE) and Portuguese (Brazil) translations.
### Fixed
- 📝 Inline citations now render correctly within markdown lists and nested elements instead of displaying as "undefined" values. [#19452](https://github.com/open-webui/open-webui/issues/19452)
- 👥 Group member selection now works correctly without randomly selecting other users or causing the user list to jump around. [#19426](https://github.com/open-webui/open-webui/issues/19426)
- 👥 Admin panel user list now displays the correct total user count and properly paginates 30 items per page after fixing database query issues with group member joins. [#19429](https://github.com/open-webui/open-webui/issues/19429)
- 🔍 Knowledge base reindexing now works correctly after resolving async execution chain issues by implementing threadpool workers for embedding operations. [#19434](https://github.com/open-webui/open-webui/pull/19434)
- 🖼️ OpenAI image generation now works correctly after fixing a connection adapter error caused by incorrect URL formatting. [#19435](https://github.com/open-webui/open-webui/pull/19435)
### Changed
- 🔧 BREAKING: Docling configuration has been consolidated from individual environment variables into a single "DOCLING_PARAMS" JSON configuration and now supports API key authentication via "DOCLING_API_KEY", requiring users to migrate existing Docling settings to the new format. [#16841](https://github.com/open-webui/open-webui/issues/16841), [#19427](https://github.com/open-webui/open-webui/pull/19427)
- 🔧 The environment variable "REPLACE_IMAGE_URLS_IN_CHAT_RESPONSE" has been renamed to "ENABLE_CHAT_RESPONSE_BASE64_IMAGE_URL_CONVERSION" for naming consistency.
## [0.6.38] - 2025-11-24
### Fixed
- 🔍 Hybrid search now works reliably after recent changes.
- 🛠️ Tool server saving now handles errors gracefully, preventing failed saves from impacting the UI.
- 🔐 SSO/OIDC code fixed to improve login reliability and better handle edge cases.
## [0.6.37] - 2025-11-24
### Added
- 🔐 Granular sharing permissions are now available with two-tiered control separating group sharing from public sharing, allowing administrators to independently configure whether users can share workspace items with groups or make them publicly accessible, with separate permission toggles for models, knowledge bases, prompts, tools, and notes, configurable via "USER_PERMISSIONS_WORKSPACE_MODELS_ALLOW_SHARING", "USER_PERMISSIONS_WORKSPACE_MODELS_ALLOW_PUBLIC_SHARING", and corresponding environment variables for other workspace item types, while groups can now be configured to opt-out of sharing via the "Allow Group Sharing" setting. [Commit](https://github.com/open-webui/open-webui/commit/7be750bcbb40da91912a0a66b7ab791effdcc3b6), [Commit](https://github.com/open-webui/open-webui/commit/f69e37a8507d6d57382d6670641b367f3127f90a)
- 🔐 Password policy enforcement is now available with configurable validation rules, allowing administrators to require specific password complexity requirements via "ENABLE_PASSWORD_VALIDATION" and "PASSWORD_VALIDATION_REGEX_PATTERN" environment variables, with default pattern requiring minimum 8 characters including uppercase, lowercase, digit, and special character. [#17794](https://github.com/open-webui/open-webui/pull/17794)
- 🔐 Granular import and export permissions are now available for workspace items, introducing six separate permission toggles for models, prompts, and tools that are disabled by default for enhanced security. [#19242](https://github.com/open-webui/open-webui/pull/19242)
- 👥 Default group assignment is now available for new users, allowing administrators to automatically assign newly registered users to a specified group for streamlined access control to models, prompts, and tools, particularly useful for organizations with group-based model access policies. [#19325](https://github.com/open-webui/open-webui/pull/19325), [#17842](https://github.com/open-webui/open-webui/issues/17842)
- 🔒 Password-based authentication can now be fully disabled via "ENABLE_PASSWORD_AUTH" environment variable, enforcing SSO-only authentication and preventing password login fallback when SSO is configured. [#19113](https://github.com/open-webui/open-webui/pull/19113)
- 🖼️ Large stream chunk handling was implemented to support models that generate images directly in their output responses, with configurable buffer size via "CHAT_STREAM_RESPONSE_CHUNK_MAX_BUFFER_SIZE" environment variable, resolving compatibility issues with models like Gemini 2.5 Flash Image. [#18884](https://github.com/open-webui/open-webui/pull/18884), [#17626](https://github.com/open-webui/open-webui/issues/17626)
- 🖼️ Streaming response middleware now handles images in delta updates with automatic base64 conversion, enabling proper display of images from models using the "choices[0].delta.images.image_url" format such as Gemini 2.5 Flash Image Preview on OpenRouter. [#19073](https://github.com/open-webui/open-webui/pull/19073), [#19019](https://github.com/open-webui/open-webui/issues/19019)
- 📈 Model list API performance was optimized by pre-fetching user group memberships and removing profile image URLs from response payloads, significantly reducing both database queries and payload size for instances with large model lists, with profile images now served dynamically via dedicated endpoints. [#19097](https://github.com/open-webui/open-webui/pull/19097), [#18950](https://github.com/open-webui/open-webui/issues/18950)
- ⏩ Batch file processing performance was improved by reducing database queries by 67% while ensuring data consistency between vector and relational databases. [#18953](https://github.com/open-webui/open-webui/pull/18953)
- 🚀 Chat import performance was dramatically improved by replacing individual per-chat API requests with a bulk import endpoint, reducing import time by up to 95% for large chat collections and providing user feedback via toast notifications displaying the number of successfully imported chats. [#17861](https://github.com/open-webui/open-webui/pull/17861)
- ⚡ Socket event broadcasting performance was optimized by implementing user-specific rooms, significantly reducing server overhead particularly for users with multiple concurrent sessions. [#18996](https://github.com/open-webui/open-webui/pull/18996)
- 🗄️ Weaviate is now supported as a vector database option, providing an additional choice for RAG document storage alongside existing ChromaDB, Milvus, Qdrant, and OpenSearch integrations. [#14747](https://github.com/open-webui/open-webui/pull/14747)
- 🗄️ PostgreSQL pgvector now supports HNSW index types and large dimensional embeddings exceeding 2000 dimensions through automatic halfvec type selection, with configurable index methods via "PGVECTOR_INDEX_METHOD", "PGVECTOR_HNSW_M", "PGVECTOR_HNSW_EF_CONSTRUCTION", and "PGVECTOR_IVFFLAT_LISTS" environment variables. [#19158](https://github.com/open-webui/open-webui/pull/19158), [#16890](https://github.com/open-webui/open-webui/issues/16890)
- 🔍 Azure AI Search is now supported as a web search provider, enabling integration with Azure's cognitive search services via "AZURE_AI_SEARCH_API_KEY", "AZURE_AI_SEARCH_ENDPOINT", and "AZURE_AI_SEARCH_INDEX_NAME" configuration. [#19104](https://github.com/open-webui/open-webui/pull/19104)
- ⚡ External embedding generation now processes API requests in parallel instead of sequential batches, reducing document processing time by 10-50x when using OpenAI, Azure OpenAI, or Ollama embedding providers, with large PDFs now processing in seconds instead of minutes. [#19296](https://github.com/open-webui/open-webui/pull/19296)
- 💨 Base64 image conversion is now available for markdown content in chat responses, automatically uploading embedded images exceeding 1KB and replacing them with file URLs to reduce payload size and resource consumption, configurable via "REPLACE_IMAGE_URLS_IN_CHAT_RESPONSE" environment variable. [#19076](https://github.com/open-webui/open-webui/pull/19076)
- 🎨 OpenAI image generation now supports additional API parameters including quality settings for GPT Image 1, configurable via "IMAGES_OPENAI_API_PARAMS" environment variable or through the admin interface, enabling cost-effective image generation with low, medium, or high quality options. [#19228](https://github.com/open-webui/open-webui/issues/19228)
- 🖼️ Image editing can now be independently enabled or disabled via admin settings, allowing administrators to control whether sequential image prompts trigger image editing or new image generation, configurable via "ENABLE_IMAGE_EDIT" environment variable. [#19284](https://github.com/open-webui/open-webui/issues/19284)
- 🔐 SSRF protection was implemented with a configurable URL blocklist that prevents access to cloud metadata endpoints and private networks, with default protections for AWS, Google Cloud, Azure, and Alibaba Cloud metadata services, customizable via "WEB_FETCH_FILTER_LIST" environment variable. [#19201](https://github.com/open-webui/open-webui/pull/19201)
- ⚡ Workspace models page now supports server-side pagination dramatically improving load times and usability for instances with large numbers of workspace models.
- 🔍 Hybrid search now indexes file metadata including filenames, titles, headings, sources, and snippets alongside document content, enabling keyword queries to surface documents where search terms appear only in metadata, configurable via "ENABLE_RAG_HYBRID_SEARCH_ENRICHED_TEXTS" environment variable. [#19095](https://github.com/open-webui/open-webui/pull/19095)
- 📂 Knowledge base upload page now supports folder drag-and-drop with recursive directory handling, enabling batch uploads of entire directory structures instead of requiring individual file selection. [#19320](https://github.com/open-webui/open-webui/pull/19320)
- 🤖 Model cloning is now available in admin settings, allowing administrators to quickly create workspace models based on existing base models through a "Clone" option in the model dropdown menu. [#17937](https://github.com/open-webui/open-webui/pull/17937)
- 🎨 UI scale adjustment is now available in interface settings, allowing users to increase the size of the entire interface from 1.0x to 1.5x for improved accessibility and readability, particularly beneficial for users with visual impairments. [#19186](https://github.com/open-webui/open-webui/pull/19186)
- 📌 Default pinned models can now be configured by administrators for all new users, mirroring the behavior of default models where admin-configured defaults apply only to users who haven't customized their pinned models, configurable via "DEFAULT_PINNED_MODELS" environment variable. [#19273](https://github.com/open-webui/open-webui/pull/19273)
- 🎙️ Text-to-Speech and Speech-to-Text services now receive user information headers when "ENABLE_FORWARD_USER_INFO_HEADERS" is enabled, allowing external TTS and STT providers to implement user-specific personalization, rate limiting, and usage tracking. [#19323](https://github.com/open-webui/open-webui/pull/19323), [#19312](https://github.com/open-webui/open-webui/issues/19312)
- 🎙️ Voice mode now supports custom system prompts via "VOICE_MODE_PROMPT_TEMPLATE" configuration, allowing administrators to control response style and behavior for voice interactions. [#18607](https://github.com/open-webui/open-webui/pull/18607)
- 🔧 WebSocket and Redis configuration options are now available including debug logging controls, custom ping timeout and interval settings, and arbitrary Redis connection options via "WEBSOCKET_SERVER_LOGGING", "WEBSOCKET_SERVER_ENGINEIO_LOGGING", "WEBSOCKET_SERVER_PING_TIMEOUT", "WEBSOCKET_SERVER_PING_INTERVAL", and "WEBSOCKET_REDIS_OPTIONS" environment variables. [#19091](https://github.com/open-webui/open-webui/pull/19091)
- 🔧 MCP OAuth dynamic client registration now automatically detects and uses the appropriate token endpoint authentication method from server-supported options, enabling compatibility with OAuth servers that only support "client_secret_basic" instead of "client_secret_post". [#19193](https://github.com/open-webui/open-webui/issues/19193)
- 🔧 Custom headers can now be configured for remote MCP and OpenAPI tool server connections, enabling integration with services that require additional authentication headers. [#18918](https://github.com/open-webui/open-webui/issues/18918)
- 🔍 Perplexity Search now supports custom API endpoints via "PERPLEXITY_SEARCH_API_URL" configuration and automatically forwards user information headers to enable personalized search experiences. [#19147](https://github.com/open-webui/open-webui/pull/19147)
- 🔍 User information headers can now be optionally forwarded to external web search engines when "ENABLE_FORWARD_USER_INFO_HEADERS" is enabled. [#19043](https://github.com/open-webui/open-webui/pull/19043)
- 📊 Daily active user metric is now available for monitoring, tracking unique users active since midnight UTC via the "webui.users.active.today" Prometheus gauge. [#19236](https://github.com/open-webui/open-webui/pull/19236), [#19234](https://github.com/open-webui/open-webui/issues/19234)
- 📊 Audit log file path is now configurable via "AUDIT_LOGS_FILE_PATH" environment variable, enabling storage in separate volumes or custom locations. [#19173](https://github.com/open-webui/open-webui/pull/19173)
- 🎨 Sidebar collapse states for model lists and group information are now persistent across page refreshes, remembering user preferences through browser-based storage. [#19159](https://github.com/open-webui/open-webui/issues/19159)
- 🎨 Background image display was enhanced with semi-transparent overlays for navbar and sidebar, creating a seamless and visually cohesive design across the entire interface. [#19157](https://github.com/open-webui/open-webui/issues/19157)
- 📋 Tables in chat messages now include a copy button that appears on hover, enabling quick copying of table content alongside the existing CSV export functionality. [#19162](https://github.com/open-webui/open-webui/issues/19162)
- 📝 Notes can now be created directly via the "/notes/new" URL endpoint with optional title and content query parameters, enabling faster note creation through bookmarks and shortcuts. [#19195](https://github.com/open-webui/open-webui/issues/19195)
- 🏷️ Tag suggestions are now context-aware, displaying only relevant tags when creating or editing models versus chat conversations, preventing confusion between model and chat tags. [#19135](https://github.com/open-webui/open-webui/issues/19135)
- ✍️ Prompt autocompletion is now available independently of the rich text input setting, improving accessibility to the feature. [#19150](https://github.com/open-webui/open-webui/issues/19150)
- 🔄 Various improvements were implemented across the frontend and backend to enhance performance, stability, and security.
- 🌐 Translations for Simplified Chinese, Traditional Chinese, Portuguese (Brazil), Catalan, Spanish (Spain), Finnish, Irish, Farsi, Swedish, Danish, German, Korean, and Thai were improved and expanded.
### Fixed
- 🤖 Model update functionality now works correctly, resolving a database parameter binding error that prevented saving changes to model configurations via the Save & Update button. [#19335](https://github.com/open-webui/open-webui/issues/19335)
- 🖼️ Multiple input images for image editing and generation are now correctly passed as an array using the "image[]" parameter syntax, enabling proper multi-image reference functionality with models like GPT Image 1. [#19339](https://github.com/open-webui/open-webui/issues/19339)
- 📱 PWA installations on iOS now properly refresh after server container restarts, resolving freezing issues by automatically unregistering service workers when version or deployment changes are detected. [#19316](https://github.com/open-webui/open-webui/pull/19316)
- 🗄️ S3 Vectors collection detection now correctly handles buckets with more than 2000 indexes by using direct index lookup instead of paginated list scanning, improving performance by approximately 8x and enabling RAG queries to work reliably at scale. [#19238](https://github.com/open-webui/open-webui/pull/19238), [#19233](https://github.com/open-webui/open-webui/issues/19233)
- 📈 Feedback retrieval performance was optimized by eliminating N+1 query patterns through database joins, adding server-side pagination and sorting, significantly reducing database load for instances with large feedback datasets. [#17976](https://github.com/open-webui/open-webui/pull/17976)
- 🔍 Chat search now works correctly with PostgreSQL when chat data contains null bytes, with comprehensive sanitization preventing null bytes during data writes, cleaning existing data on read, and stripping null bytes during search queries to ensure reliable search functionality. [#15616](https://github.com/open-webui/open-webui/issues/15616)
- 🔍 Hybrid search with reranking now correctly handles attribute validation, preventing errors when collection results lack expected structure. [#19025](https://github.com/open-webui/open-webui/pull/19025), [#17046](https://github.com/open-webui/open-webui/issues/17046)
- 🔎 Reranking functionality now works correctly after recent refactoring, resolving crashes caused by incorrect function argument handling. [#19270](https://github.com/open-webui/open-webui/pull/19270)
- 🤖 Azure OpenAI models now support the "reasoning_effort" parameter, enabling proper configuration of reasoning capabilities for models like GPT-5.1 which default to no reasoning without this setting. [#19290](https://github.com/open-webui/open-webui/issues/19290)
- 🤖 Models with very long IDs can now be deleted correctly, resolving URL length limitations that previously prevented management operations on such models. [#18230](https://github.com/open-webui/open-webui/pull/18230)
- 🤖 Model-level streaming settings now correctly apply to API requests, ensuring "Stream Chat Response" toggle properly controls the streaming parameter. [#19154](https://github.com/open-webui/open-webui/issues/19154)
- 🖼️ Image editing configuration now correctly preserves independent OpenAI API endpoints and keys, preventing them from being overwritten by image generation settings. [#19003](https://github.com/open-webui/open-webui/issues/19003)
- 🎨 Gemini image edit settings now display correctly in the admin panel, fixing an incorrect configuration key reference that prevented proper rendering of edit options. [#19200](https://github.com/open-webui/open-webui/pull/19200)
- 🖌️ Image generation settings menu now loads correctly, resolving validation errors with AUTOMATIC1111 API authentication parameters. [#19187](https://github.com/open-webui/open-webui/issues/19187), [#19246](https://github.com/open-webui/open-webui/issues/19246)
- 📅 Date formatting in chat search and admin user chat search now correctly respects the "DEFAULT_LOCALE" environment variable, displaying dates according to the configured locale instead of always using MM/DD/YYYY format. [#19305](https://github.com/open-webui/open-webui/pull/19305), [#19020](https://github.com/open-webui/open-webui/issues/19020)
- 📝 RAG template query placeholder escaping logic was corrected to prevent unintended replacements of context values when query placeholders appear in retrieved content. [#19102](https://github.com/open-webui/open-webui/pull/19102), [#19101](https://github.com/open-webui/open-webui/issues/19101)
- 📄 RAG template prompt duplication was eliminated by removing redundant user query section from the default template. [#19099](https://github.com/open-webui/open-webui/pull/19099), [#19098](https://github.com/open-webui/open-webui/issues/19098)
- 📋 MinerU local mode configuration no longer incorrectly requires an API key, allowing proper use of local content extraction without external API credentials. [#19258](https://github.com/open-webui/open-webui/issues/19258)
- 📊 Excel file uploads now work correctly with the addition of the missing msoffcrypto-tool dependency, resolving import errors introduced by the unstructured package upgrade. [#19153](https://github.com/open-webui/open-webui/issues/19153)
- 📑 Docling parameters now properly handle JSON serialization, preventing exceptions and ensuring configuration changes are saved correctly. [#19072](https://github.com/open-webui/open-webui/pull/19072)
- 🛠️ UserValves configuration now correctly isolates settings per tool, preventing configuration contamination when multiple tools with UserValves are used simultaneously. [#19185](https://github.com/open-webui/open-webui/pull/19185), [#15569](https://github.com/open-webui/open-webui/issues/15569)
- 🔧 Tool selection prompt now correctly handles user messages without duplication, removing redundant query prefixes and improving prompt clarity. [#19122](https://github.com/open-webui/open-webui/pull/19122), [#19121](https://github.com/open-webui/open-webui/issues/19121)
- 📝 Notes chat feature now correctly submits messages to the completions endpoint, resolving errors that prevented AI model interactions. [#19079](https://github.com/open-webui/open-webui/pull/19079)
- 📝 Note PDF downloads now sanitize HTML content using DOMPurify before rendering, preventing potential DOM-based XSS attacks from malicious content in notes. [Commit](https://github.com/open-webui/open-webui/commit/03cc6ce8eb5c055115406e2304fbf7e3338b8dce)
- 📁 Archived chats now have their folder associations automatically removed to prevent unintended deletion when their previous folder is deleted. [#14578](https://github.com/open-webui/open-webui/issues/14578)
- 🔐 ElevenLabs API key is now properly obfuscated in the admin settings page, preventing plain text exposure of sensitive credentials. [#19262](https://github.com/open-webui/open-webui/pull/19262), [#19260](https://github.com/open-webui/open-webui/issues/19260)
- 🔧 MCP OAuth server metadata discovery now follows the correct specification order, ensuring proper authentication flow compliance. [#19244](https://github.com/open-webui/open-webui/pull/19244)
- 🔒 API key endpoint restrictions now properly enforce access controls for all endpoints including SCIM, preventing unintended access when "API_KEY_ALLOWED_ENDPOINTS" is configured. [#19168](https://github.com/open-webui/open-webui/issues/19168)
- 🔓 OAuth role claim parsing now supports both flat and nested claim structures, enabling compatibility with OAuth providers that deliver claims as direct properties on the user object rather than nested structures. [#19286](https://github.com/open-webui/open-webui/pull/19286)
- 🔑 OAuth MCP server verification now correctly extracts the access token value for authorization headers instead of sending the entire token dictionary. [#19149](https://github.com/open-webui/open-webui/pull/19149), [#19148](https://github.com/open-webui/open-webui/issues/19148)
- ⚙️ OAuth dynamic client registration now correctly converts empty strings to None for optional fields, preventing validation failures in MCP package integration. [#19144](https://github.com/open-webui/open-webui/pull/19144), [#19129](https://github.com/open-webui/open-webui/issues/19129)
- 🔐 OIDC authentication now correctly passes client credentials in access token requests, ensuring compatibility with providers that require these parameters per RFC 6749. [#19132](https://github.com/open-webui/open-webui/pull/19132), [#19131](https://github.com/open-webui/open-webui/issues/19131)
- 🔗 OAuth client creation now respects configured token endpoint authentication methods instead of defaulting to basic authentication, preventing failures with servers that don't support basic auth. [#19165](https://github.com/open-webui/open-webui/pull/19165)
- 📋 Text copied from chat responses in Chrome now pastes without background formatting, improving readability when pasting into word processors. [#19083](https://github.com/open-webui/open-webui/issues/19083)
### Changed
- 🗄️ Group membership data storage was refactored from JSON arrays to a dedicated relational database table, significantly improving query performance and scalability for instances with large numbers of users and groups, while API responses now return member counts instead of full user ID arrays. [#19239](https://github.com/open-webui/open-webui/pull/19239)
- 📄 MinerU parameter handling was refactored to pass parameters directly to the API, improving flexibility and fixing VLM backend configuration. [#19105](https://github.com/open-webui/open-webui/pull/19105), [#18446](https://github.com/open-webui/open-webui/discussions/18446)
- 🔐 API key creation is now controlled by granular user and group permissions, with the "ENABLE_API_KEY" environment variable renamed to "ENABLE_API_KEYS" and disabled by default, requiring explicit configuration at both the global and user permission levels, while related environment variables "ENABLE_API_KEY_ENDPOINT_RESTRICTIONS" and "API_KEY_ALLOWED_ENDPOINTS" were renamed to "ENABLE_API_KEYS_ENDPOINT_RESTRICTIONS" and "API_KEYS_ALLOWED_ENDPOINTS" respectively. [#18336](https://github.com/open-webui/open-webui/pull/18336)
## [0.6.36] - 2025-11-07
### Added
- 🔐 OAuth group parsing now supports configurable separators via the "OAUTH_GROUPS_SEPARATOR" environment variable, enabling proper handling of semicolon-separated group claims from providers like CILogon. [#18987](https://github.com/open-webui/open-webui/pull/18987), [#18979](https://github.com/open-webui/open-webui/issues/18979)
### Fixed
- 🛠️ Tool calling functionality is restored by correcting asynchronous function handling in tool parameter updates. [#18981](https://github.com/open-webui/open-webui/issues/18981)
- 🖼️ The ComfyUI image edit workflow editor modal now opens correctly when clicking the Edit button. [#18978](https://github.com/open-webui/open-webui/issues/18978)
- 🔥 Firecrawl import errors are resolved by implementing lazy loading and using the correct class name. [#18973](https://github.com/open-webui/open-webui/issues/18973)
- 🔌 Socket.IO CORS warning is resolved by properly configuring CORS origins for Socket.IO connections. [Commit](https://github.com/open-webui/open-webui/commit/639d26252e528c9c37a5f553b11eb94376d8792d)
## [0.6.35] - 2025-11-06
### Added
- 🖼️ Image generation system received a comprehensive overhaul with major new capabilities including full image editing support allowing users to modify existing images using text prompts with OpenAI, Gemini, or ComfyUI engines, adding Gemini 2.5 Flash Image (Nano Banana) support, Qwen Image Edit integration, resolution of base64-encoded image display issues, streamlined AUTOMATIC1111 configuration by consolidating parameters into a flexible JSON parameters field, and enhanced UI with a code editor modal for ComfyUI workflow management. [#17434](https://github.com/open-webui/open-webui/pull/17434), [#16976](https://github.com/open-webui/open-webui/issues/16976), [Commit](https://github.com/open-webui/open-webui/commit/8e5690aab4f632a57027e2acf880b8f89a8717c0), [Commit](https://github.com/open-webui/open-webui/commit/72f8539fd2e679fec0762945f22f4b8a6920afa0), [Commit](https://github.com/open-webui/open-webui/commit/8d34fcb586eeee1fac6da2f991518b8a68b00b72), [Commit](https://github.com/open-webui/open-webui/commit/72900cd686de1fa6be84b5a8a2fc857cff7b91b8)
- 🔒 CORS origin validation was added to WebSocket connections as a defense-in-depth security measure against cross-site WebSocket hijacking attacks. [#18411](https://github.com/open-webui/open-webui/pull/18411), [#18410](https://github.com/open-webui/open-webui/issues/18410)
- 🔄 Automatic page refresh now occurs when a version update is detected via WebSocket connection, ensuring users always run the latest version without cache issues. [Commit](https://github.com/open-webui/open-webui/commit/989f192c92d2fe55daa31336e7971e21798b96ae)
- 🐍 Experimental initial preparations for Python 3.13 compatibility by updating dependencies with security enhancements and cryptographic improvements. [#18430](https://github.com/open-webui/open-webui/pull/18430), [#18424](https://github.com/open-webui/open-webui/pull/18424)
- ⚡ Image compression now preserves the original image format instead of converting to PNG, significantly reducing file sizes and improving chat loading performance. [#18506](https://github.com/open-webui/open-webui/pull/18506)
- 🎤 Mistral Voxtral model support was added for text-to-speech, including voxtral-small and voxtral-mini models with both transcription and chat completion API support. [#18934](https://github.com/open-webui/open-webui/pull/18934)
- 🔊 Text-to-speech now uses a global audio queue system to prevent overlapping playback, ensuring only one TTS instance plays at a time with proper stop/start controls and automatic cleanup when switching between messages. [#16152](https://github.com/open-webui/open-webui/pull/16152), [#18744](https://github.com/open-webui/open-webui/pull/18744), [#16150](https://github.com/open-webui/open-webui/issues/16150)
- 🔊 ELEVENLABS_API_BASE_URL environment variable now allows configuration of custom ElevenLabs API endpoints, enabling support for EU residency API requirements. [#18402](https://github.com/open-webui/open-webui/issues/18402)
- 🔐 OAUTH_ROLES_SEPARATOR environment variable now allows custom role separators for OAuth roles that contain commas, useful for roles specified in LDAP syntax. [#18572](https://github.com/open-webui/open-webui/pull/18572)
- 📄 External document loaders can now optionally forward user information headers when ENABLE_FORWARD_USER_INFO_HEADERS is enabled, enabling cost tracking, audit logs, and usage analytics for external services. [#18731](https://github.com/open-webui/open-webui/pull/18731)
- 📄 MISTRAL_OCR_API_BASE_URL environment variable now allows configuration of custom Mistral OCR API endpoints for flexible deployment options. [Commit](https://github.com/open-webui/open-webui/commit/415b93c7c35c2e2db4425e6da1b88b3750f496b0)
- ⌨️ Keyboard shortcut hints are now displayed on sidebar buttons with a refactored shortcuts modal that accurately reflects all available hotkeys across different keyboard layouts. [#18473](https://github.com/open-webui/open-webui/pull/18473)
- 🛠️ Tooltips now display tool descriptions when hovering over tool names on the model edit page, improving usability and providing immediate context. [#18707](https://github.com/open-webui/open-webui/pull/18707)
- 📝 "Create a new note" from the search modal now immediately creates a new private note and opens it in the editor instead of navigating to the generic notes page. [#18255](https://github.com/open-webui/open-webui/pull/18255)
- 🖨️ Code block output now preserves whitespace formatting with monospace font to accurately reflect terminal behavior. [#18352](https://github.com/open-webui/open-webui/pull/18352)
- ✏️ Edit button is now available in the three-dot menu of models in the workspace section for quick access to model editing, with the menu reorganized for better user experience and Edit, Clone, Copy Link, and Share options logically grouped. [#18574](https://github.com/open-webui/open-webui/pull/18574)
- 📌 Sidebar models section is now collapsible, allowing users to expand and collapse the pinned models list for better sidebar organization. [Commit](https://github.com/open-webui/open-webui/commit/82c08a3b5d189f81c96b6548cc872198771015b0)
- 🌙 Dark mode styles for select elements were added using Tailwind CSS classes, improving consistency across the interface. [#18636](https://github.com/open-webui/open-webui/pull/18636)
- 🔄 Various improvements were implemented across the frontend and backend to enhance performance, stability, and security.
- 🌐 Translations for Portuguese (Brazil), Greek, German, Traditional Chinese, Simplified Chinese, Spanish, Georgian, Danish, and Estonian were enhanced and expanded.
### Fixed
- 🔒 Server-Sent Event (SSE) code injection vulnerability in Direct Connections is resolved by blocking event emission from untrusted external model servers; event emitters from direct connected model servers are no longer supported, preventing arbitrary JavaScript execution in user browsers. [Commit](https://github.com/open-webui/open-webui/commit/8af6a4cf21b756a66cd58378a01c60f74c39b7ca)
- 🛡️ DOM XSS vulnerability in "Insert Prompt as Rich Text" is resolved by sanitizing HTML content with DOMPurify before rendering. [Commit](https://github.com/open-webui/open-webui/commit/eb9c4c0e358c274aea35f21c2856c0a20051e5f1)
- ⚙️ MCP server cancellation scope corruption is prevented by reversing disconnection order to follow LIFO and properly handling exceptions, resolving 100% CPU usage when resuming chats with expired tokens or using multiple streamable MCP servers. [#18537](https://github.com/open-webui/open-webui/pull/18537)
- 🔧 UI freeze when querying models with knowledge bases containing inconsistent distance metrics is resolved by properly initializing the distances array in citations. [#18585](https://github.com/open-webui/open-webui/pull/18585)
- 🤖 Duplicate model IDs from multiple OpenAI endpoints are now automatically deduplicated server-side, preventing frontend crashes for users with unified gateway proxies that aggregate multiple providers. [Commit](https://github.com/open-webui/open-webui/commit/fdf7ca11d4f3cc8fe63e81c98dc0d1e48e52ba36)
- 🔐 Login failures with passwords longer than 72 bytes are resolved by safely truncating oversized passwords for bcrypt compatibility. [#18157](https://github.com/open-webui/open-webui/issues/18157)
- 🔐 OAuth 2.1 MCP tool connections now automatically re-register clients when stored client IDs become stale, preventing unauthorized_client errors after editing tool endpoints and providing detailed error messages for callback failures. [#18415](https://github.com/open-webui/open-webui/pull/18415), [#18309](https://github.com/open-webui/open-webui/issues/18309)
- 🔓 OAuth 2.1 discovery, metadata fetching, and dynamic client registration now correctly use HTTP proxy environment variables when trust_env is enabled. [Commit](https://github.com/open-webui/open-webui/commit/bafeb76c411483bd6b135f0edbcdce048120f264)
- 🔌 MCP server connection failures now display clear error messages in the chat interface instead of silently failing. [#18892](https://github.com/open-webui/open-webui/pull/18892), [#18889](https://github.com/open-webui/open-webui/issues/18889)
- 💬 Chat titles are now properly generated even when title auto-generation is disabled in interface settings, fixing an issue where chats would remain labeled as "New chat". [#18761](https://github.com/open-webui/open-webui/pull/18761), [#18717](https://github.com/open-webui/open-webui/issues/18717), [#6478](https://github.com/open-webui/open-webui/issues/6478)
- 🔍 Chat query errors are prevented by properly validating and handling the "order_by" parameter to ensure requested columns exist. [#18400](https://github.com/open-webui/open-webui/pull/18400), [#18452](https://github.com/open-webui/open-webui/pull/18452)
- 🔧 Root-level max_tokens parameter is no longer dropped when proxying to Ollama, properly converting to num_predict to limit output token length as intended. [#18618](https://github.com/open-webui/open-webui/issues/18618)
- 🔑 Self-hosted Marker instances can now be used without requiring an API key, while keeping it optional for datalab Marker service users. [#18617](https://github.com/open-webui/open-webui/issues/18617)
- 🔧 OpenAPI specification endpoint conflict between "/api/v1/models" and "/api/v1/models/" is resolved by changing the models router endpoint to "/list", preventing duplicate operationId errors when generating TypeScript API clients. [#18758](https://github.com/open-webui/open-webui/issues/18758)
- 🏷️ Model tags are now de-duplicated case-insensitively in both the model selector and workspace models page, preventing duplicate entries with different capitalization from appearing in filter dropdowns. [#18716](https://github.com/open-webui/open-webui/pull/18716), [#18711](https://github.com/open-webui/open-webui/issues/18711)
- 📄 Docling RAG parameter configuration is now correctly saved in the admin UI by fixing the typo in the "DOCLING_PARAMS" parameter name. [#18390](https://github.com/open-webui/open-webui/pull/18390)
- 📃 Tika document processing now automatically detects content types instead of relying on potentially incorrect browser-provided mime-types, improving file handling accuracy for formats like RTF. [#18765](https://github.com/open-webui/open-webui/pull/18765), [#18683](https://github.com/open-webui/open-webui/issues/18683)
- 🖼️ Image and video uploads to knowledge bases now display proper error messages instead of showing an infinite spinner when the content extraction engine does not support these file types. [#18514](https://github.com/open-webui/open-webui/issues/18514)
- 📝 Notes PDF export now properly detects and applies dark mode styling consistently across both the notes list and individual note pages, with a shared utility function to eliminate code duplication. [#18526](https://github.com/open-webui/open-webui/issues/18526)
- 💭 Details tags for reasoning content are now correctly identified and rendered even when the same tag is present in user messages. [#18840](https://github.com/open-webui/open-webui/pull/18840), [#18294](https://github.com/open-webui/open-webui/issues/18294)
- 📊 Mermaid and Vega rendering errors now display inline with the code instead of showing repetitive toast notifications, improving user experience when models generate invalid diagram syntax. [Commit](https://github.com/open-webui/open-webui/commit/fdc0f04a8b7dd0bc9f9dc0e7e30854f7a0eea3e9)
- 📈 Mermaid diagram rendering errors no longer cause UI unavailability or display error messages below the input box. [#18493](https://github.com/open-webui/open-webui/pull/18493), [#18340](https://github.com/open-webui/open-webui/issues/18340)
- 🔗 Web search SSL verification is now asynchronous, preventing the website from hanging during web search operations. [#18714](https://github.com/open-webui/open-webui/pull/18714), [#18699](https://github.com/open-webui/open-webui/issues/18699)
- 🌍 Web search results now correctly use HTTP proxy environment variables when WEB_SEARCH_TRUST_ENV is enabled. [#18667](https://github.com/open-webui/open-webui/pull/18667), [#7008](https://github.com/open-webui/open-webui/discussions/7008)
- 🔍 Google Programmable Search Engine now properly includes referer headers, enabling API keys with HTTP referrer restrictions configured in Google Cloud Console. [#18871](https://github.com/open-webui/open-webui/pull/18871), [#18870](https://github.com/open-webui/open-webui/issues/18870)
- ⚡ YouTube video transcript fetching now works correctly when using a proxy connection. [#18419](https://github.com/open-webui/open-webui/pull/18419)
- 🎙️ Speech-to-text transcription no longer deletes or replaces existing text in the prompt input field, properly preserving any previously entered content. [#18540](https://github.com/open-webui/open-webui/issues/18540)
- 🎙️ The "Instant Auto-Send After Voice Transcription" setting now functions correctly and automatically sends transcribed text when enabled. [#18466](https://github.com/open-webui/open-webui/issues/18466)
- ⚙️ Chat settings now load properly when reopening a tab or starting a new session by initializing defaults when sessionStorage is empty. [#18438](https://github.com/open-webui/open-webui/pull/18438)
- 🔎 Folder tag search in the sidebar now correctly handles folder names with multiple spaces by replacing all spaces with underscores. [Commit](https://github.com/open-webui/open-webui/commit/a8fe979af68e47e4e4bb3eb76e48d93d60cd2a45)
- 🛠️ Functions page now updates immediately after deleting a function, removing the need for a manual page reload. [#18912](https://github.com/open-webui/open-webui/pull/18912), [#18908](https://github.com/open-webui/open-webui/issues/18908)
- 🛠️ Native tool calling now properly supports sequential tool calls with shared context, allowing tools to access images and data from previous tool executions in the same conversation. [#18664](https://github.com/open-webui/open-webui/pull/18664)
- 🎯 Globally enabled actions in the model editor now correctly apply as global instead of being treated as disabled. [#18577](https://github.com/open-webui/open-webui/pull/18577)
- 📋 Clipboard images pasted via the "{{CLIPBOARD}}" prompt variable are now correctly converted to base64 format before being sent to the backend, resolving base64 encoding errors. [#18432](https://github.com/open-webui/open-webui/pull/18432), [#18425](https://github.com/open-webui/open-webui/issues/18425)
- 📋 File list is now cleared when switching to models that do not support file uploads, preventing files from being sent to incompatible models. [#18496](https://github.com/open-webui/open-webui/pull/18496)
- 📂 Move menu no longer displays when folders are empty. [#18484](https://github.com/open-webui/open-webui/pull/18484)
- 📁 Folder and channel creation now validates that names are not empty, preventing creation of folders or channels with no name and showing an error toast if attempted. [#18564](https://github.com/open-webui/open-webui/pull/18564)
- 🖊️ Rich text input no longer removes text between equals signs when pasting code with comparison operators. [#18551](https://github.com/open-webui/open-webui/issues/18551)
- ⌨️ Keyboard shortcuts now display the correct keys for international and non-QWERTY keyboard layouts by detecting the user's layout using the Keyboard API. [#18533](https://github.com/open-webui/open-webui/pull/18533)
- 🌐 "Attach Webpage" button now displays with correct disabled styling when a model does not support file uploads. [#18483](https://github.com/open-webui/open-webui/pull/18483)
- 🎚️ Divider no longer displays in the integrations menu when no integrations are enabled. [#18487](https://github.com/open-webui/open-webui/pull/18487)
- 📱 Chat controls button is now properly hidden on mobile for users without admin or explicit chat control permissions. [#18641](https://github.com/open-webui/open-webui/pull/18641)
- 📍 User menu, download submenu, and move submenu are now repositioned to prevent overlap with the Chat Controls sidebar when it is open. [Commit](https://github.com/open-webui/open-webui/commit/414ab51cb6df1ab0d6c85ac6c1f2c5c9a5f8e2aa)
- 🎯 Artifacts button no longer appears in the chat menu when there are no artifacts to display. [Commit](https://github.com/open-webui/open-webui/commit/ed6449d35f84f68dc75ee5c6b3f4748a3fda0096)
- 🎨 Artifacts view now automatically displays when opening an existing conversation containing artifacts, improving user experience. [#18215](https://github.com/open-webui/open-webui/pull/18215)
- 🖌️ Formatting toolbar is no longer hidden under images or code blocks in chat and now displays correctly above all message content.
- 🎨 Layout shift near system instructions is prevented by properly rendering the chat component when system prompts are empty. [#18594](https://github.com/open-webui/open-webui/pull/18594)
- 📐 Modal layout shift caused by scrollbar appearance is prevented by adding a stable scrollbar gutter. [#18591](https://github.com/open-webui/open-webui/pull/18591)
- ✨ Spacing between icon and label in the user menu dropdown items is now consistent. [#18595](https://github.com/open-webui/open-webui/pull/18595)
- 💬 Duplicate prompt suggestions no longer cause the webpage to freeze or throw JavaScript errors by implementing proper key management with composite keys. [#18841](https://github.com/open-webui/open-webui/pull/18841), [#18566](https://github.com/open-webui/open-webui/issues/18566)
- 🔍 Chat preview loading in the search modal now works correctly for all search results by fixing an index boundary check that previously caused out-of-bounds errors. [#18911](https://github.com/open-webui/open-webui/pull/18911)
- ♿ Screen reader support was enhanced by wrapping messages in semantic elements with descriptive aria-labels, adding "Assistant is typing" and "Response complete" announcements for improved accessibility. [#18735](https://github.com/open-webui/open-webui/pull/18735)
- 🔒 Incorrect await call in the OAuth 2.1 flow is removed, eliminating a logged exception during authentication. [#18236](https://github.com/open-webui/open-webui/pull/18236)
- 🛡️ Duplicate crossorigin attribute in the manifest file was removed. [#18413](https://github.com/open-webui/open-webui/pull/18413)
### Changed
- 🔄 Firecrawl integration was refactored to use the official Firecrawl SDK instead of direct HTTP requests and langchain_community FireCrawlLoader, improving reliability and performance with batch scraping support and enhanced error handling. [#18635](https://github.com/open-webui/open-webui/pull/18635)
- 📄 MinerU content extraction engine now only supports PDF files following the upstream removal of LibreOffice document conversion in version 2.0.0; users needing to process office documents should convert them to PDF format first. [#18448](https://github.com/open-webui/open-webui/issues/18448)
## [0.6.34] - 2025-10-16
### Added
- 📄 MinerU is now supported as a document parser backend, with support for both local and managed API deployments. [#18306](https://github.com/open-webui/open-webui/pull/18306)
- 🔒 JWT token expiration default is now set to 4 weeks instead of never expiring, with security warnings displayed in backend logs and admin UI when set to unlimited. [#18261](https://github.com/open-webui/open-webui/pull/18261), [#18262](https://github.com/open-webui/open-webui/pull/18262)
- ⚡ Page loading performance is improved by preventing unnecessary API requests when sidebar folders are not expanded. [#18179](https://github.com/open-webui/open-webui/pull/18179), [#17476](https://github.com/open-webui/open-webui/issues/17476)
- 📁 File hash values are now included in the knowledge endpoint response, enabling efficient file synchronization through hash comparison. [#18284](https://github.com/open-webui/open-webui/pull/18284), [#18283](https://github.com/open-webui/open-webui/issues/18283)
- 🎨 Chat dialog scrollbar visibility is improved by increasing its width, making it easier to use for navigation. [#18369](https://github.com/open-webui/open-webui/pull/18369), [#11782](https://github.com/open-webui/open-webui/issues/11782)
- 🔄 Various improvements were implemented across the frontend and backend to enhance performance, stability, and security.
- 🌐 Translations for Catalan, Chinese, Czech, Finnish, German, Kabyle, Korean, Portuguese (Brazil), Spanish, Thai, and Turkish were enhanced and expanded.
### Fixed
- 📚 Focused retrieval mode now works correctly, preventing the system from forcing full context mode and loading all documents in a knowledge base regardless of settings. [#18133](https://github.com/open-webui/open-webui/issues/18133)
- 🔧 Filter inlet functions now correctly execute on tool call continuations, ensuring parameter persistence throughout tool interactions. [#18222](https://github.com/open-webui/open-webui/issues/18222)
- 🛠️ External tool servers now properly support DELETE requests with body data. [#18289](https://github.com/open-webui/open-webui/pull/18289), [#18287](https://github.com/open-webui/open-webui/issues/18287)
- 🗄️ Oracle23ai vector database client now correctly handles variable initialization, resolving UnboundLocalError when retrieving items from collections. [#18356](https://github.com/open-webui/open-webui/issues/18356)
- 🔧 Model auto-pull functionality now works correctly even when user settings remain unmodified. [#18324](https://github.com/open-webui/open-webui/pull/18324)
- 🎨 Duplicate HTML content in artifacts is now prevented by improving code block detection logic. [#18195](https://github.com/open-webui/open-webui/pull/18195), [#6154](https://github.com/open-webui/open-webui/issues/6154)
- 💬 Pinned chats now appear in the Reference Chats list and can be referenced in conversations. [#18288](https://github.com/open-webui/open-webui/issues/18288)
- 📝 Misleading knowledge base warning text in documents settings is clarified to correctly instruct users about reindexing vectors. [#18263](https://github.com/open-webui/open-webui/pull/18263)
- 🔔 Toast notifications can now be dismissed even when a modal is open. [#18260](https://github.com/open-webui/open-webui/pull/18260)
- 🔘 The "Chats" button in the sidebar now correctly toggles chat list visibility without navigating away from the current page. [#18232](https://github.com/open-webui/open-webui/pull/18232)
- 🎯 The Integrations menu no longer closes prematurely when clicking outside the Valves modal. [#18310](https://github.com/open-webui/open-webui/pull/18310)
- 🛠️ Tool ID display issues where "undefined" was incorrectly shown in the interface are now resolved. [#18178](https://github.com/open-webui/open-webui/pull/18178)
- 🛠️ Model management issues caused by excessively long model IDs are now prevented through validation that limits model IDs to 256 characters. [#18125](https://github.com/open-webui/open-webui/issues/18125)
## [0.6.33] - 2025-10-08
### Added
- 🎨 Workspace interface received a comprehensive redesign across Models, Knowledge, Prompts, and Tools sections, featuring reorganized controls, view filters for created vs shared items, tag selectors, improved visual hierarchy, and streamlined import/export functionality. [Commit](https://github.com/open-webui/open-webui/commit/2c59a288603d8c5f004f223ee00fef37cc763a8e), [Commit](https://github.com/open-webui/open-webui/commit/6050c86ab6ef6b8c96dd3f99c62a6867011b67a4), [Commit](https://github.com/open-webui/open-webui/commit/96ecb47bc71c072aa34ef2be10781b042bef4e8c), [Commit](https://github.com/open-webui/open-webui/commit/2250d102b28075a9611696e911536547abb8b38a), [Commit](https://github.com/open-webui/open-webui/commit/23c8f6d507bfee75ab0015a3e2972d5c26f7e9bf), [Commit](https://github.com/open-webui/open-webui/commit/a743b16728c6ae24b8befbc2d7f24eb9e20c4ad5)
- 🛠️ Functions admin interface received a comprehensive redesign with creator attribution display, ownership filters for created vs shared items, improved organization, and refined styling. [Commit](https://github.com/open-webui/open-webui/commit/f5e1a42f51acc0b9d5b63a33c1ca2e42470239c1)
- ⚡ Page initialization performance is significantly improved through parallel data loading and optimized folder API calls, reducing initial page load time. [#17559](https://github.com/open-webui/open-webui/pull/17559), [#17889](https://github.com/open-webui/open-webui/pull/17889)
- ⚡ Chat overview component is now dynamically loaded on demand, reducing initial page bundle size by approximately 470KB and improving first-screen loading speed. [#17595](https://github.com/open-webui/open-webui/pull/17595)
- 📁 Folders can now be attached to chats using the "#" command, automatically expanding to include all files within the folder for streamlined knowledge base integration. [Commit](https://github.com/open-webui/open-webui/commit/d2cb78179d66dc85188172a08622d4c97a2ea1ee)
- 📱 Progressive Web App now supports Android share target functionality, allowing users to share web pages, YouTube videos, and text directly to Open WebUI from the system share menu. [#17633](https://github.com/open-webui/open-webui/pull/17633), [#17125](https://github.com/open-webui/open-webui/issues/17125)
- 🗄️ Redis session storage is now available as an experimental option for OAuth authentication flows via the ENABLE_STAR_SESSIONS_MIDDLEWARE environment variable, providing shared session state across multi-replica deployments to address CSRF errors, though currently only basic Redis setups are supported. [#17223](https://github.com/open-webui/open-webui/pull/17223), [#15373](https://github.com/open-webui/open-webui/issues/15373), [Docs:Commit](https://github.com/open-webui/docs/commit/14052347f165d1b597615370373d7289ce44c7f9)
- 📊 Vega and Vega-Lite chart visualization renderers are now supported in code blocks, enabling inline rendering of data visualizations with automatic compilation of Vega-Lite specifications. [#18033](https://github.com/open-webui/open-webui/pull/18033), [#18040](https://github.com/open-webui/open-webui/pull/18040), [#18022](https://github.com/open-webui/open-webui/issues/18022)
- 🔗 OpenAI connections now support custom HTTP headers, enabling users to configure authentication and routing headers for specific deployment requirements. [#18021](https://github.com/open-webui/open-webui/pull/18021), [#9732](https://github.com/open-webui/open-webui/discussions/9732)
- 🔐 OpenID Connect authentication now supports OIDC providers without email scope via the ENABLE_OAUTH_WITHOUT_EMAIL environment variable, enabling compatibility with identity providers that don't expose email addresses. [#18047](https://github.com/open-webui/open-webui/pull/18047), [#18045](https://github.com/open-webui/open-webui/issues/18045)
- 🤖 Ollama model management modal now features individual model update cancellation, comprehensive tooltips for all buttons, and streamlined notification behavior to reduce toast spam. [#16863](https://github.com/open-webui/open-webui/pull/16863)
- ☁️ OneDrive file picker now includes search functionality and "My Organization" pivot for business accounts, enabling easier file discovery across organizational content. [#17930](https://github.com/open-webui/open-webui/pull/17930), [#17929](https://github.com/open-webui/open-webui/issues/17929)
- 📊 Chat overview flow diagram now supports toggling between vertical and horizontal layout orientations for improved visualization flexibility. [#17941](https://github.com/open-webui/open-webui/pull/17941)
- 🔊 OpenAI Text-to-Speech engine now supports additional parameters, allowing users to customize TTS behavior with provider-specific options via JSON configuration. [#17985](https://github.com/open-webui/open-webui/issues/17985), [#17188](https://github.com/open-webui/open-webui/pull/17188)
- 🛠️ Tool server list now displays server name, URL, and type (OpenAPI or MCP) for easier identification and management. [#18062](https://github.com/open-webui/open-webui/issues/18062)
- 📁 Folders now remember the last selected model, automatically applying it when starting new chats within that folder. [#17836](https://github.com/open-webui/open-webui/issues/17836)
- 🔢 Ollama embedding endpoint now supports the optional dimensions parameter for controlling embedding output size, compatible with Ollama v0.11.11 and later. [#17942](https://github.com/open-webui/open-webui/pull/17942)
- ⚡ Workspace knowledge page load time is improved by removing redundant API calls, enhancing overall responsiveness. [#18057](https://github.com/open-webui/open-webui/pull/18057)
- ⚡ File metadata query performance is enhanced by selecting only relevant columns instead of retrieving entire records, reducing database overhead. [#18013](https://github.com/open-webui/open-webui/pull/18013)
- 📄 Note PDF exports now include titles and properly render in dark mode with appropriate background colors. [Commit](https://github.com/open-webui/open-webui/commit/216fb5c3db1a223ffe6e72d97aa9551fe0e2d028)
- 📄 Docling document extraction now supports additional parameters for VLM pipeline configuration, enabling customized vision model settings. [#17363](https://github.com/open-webui/open-webui/pull/17363)
- ⚙️ Server startup script now supports passing arbitrary arguments to uvicorn, enabling custom server configuration options. [#17919](https://github.com/open-webui/open-webui/pull/17919), [#17918](https://github.com/open-webui/open-webui/issues/17918)
- 🔄 Various improvements were implemented across the frontend and backend to enhance performance, stability, and security.
- 🌐 Translations for German, Danish, Spanish, Korean, Portuguese (Brazil), Simplified Chinese, and Traditional Chinese were enhanced and expanded.
### Fixed
- 💬 System prompts are no longer duplicated in chat requests, eliminating confusion and excessive token usage caused by repeated instructions being sent to models. [#17198](https://github.com/open-webui/open-webui/issues/17198), [#16855](https://github.com/open-webui/open-webui/issues/16855)
- 🔐 MCP OAuth 2.1 authentication now complies with the standard by implementing PKCE with S256 code challenge method and explicitly passing client credentials during token authorization, resolving "code_challenge: Field required" and "client_id: Field required" errors when connecting to OAuth-secured MCP servers. [Commit](https://github.com/open-webui/open-webui/commit/911a114ad459f5deebd97543c13c2b90196efb54), [#18010](https://github.com/open-webui/open-webui/issues/18010), [#18087](https://github.com/open-webui/open-webui/pull/18087)
- 🔐 OAuth signup flow now handles password hashing correctly by migrating from passlib to native bcrypt, preventing failures when passwords exceed 72 bytes. [#17917](https://github.com/open-webui/open-webui/issues/17917)
- 🔐 OAuth token refresh errors are resolved by properly registering and storing OAuth clients, fixing "Constructor parameter should be str" exceptions for Google, Microsoft, and OIDC providers. [#17829](https://github.com/open-webui/open-webui/issues/17829)
- 🔐 OAuth server metadata URL is now correctly accessed via the proper attribute, fixing automatic token refresh and logout functionality for Microsoft OAuth provider when OPENID_PROVIDER_URL is not set. [#18065](https://github.com/open-webui/open-webui/pull/18065)
- 🔐 OAuth credential decryption failures now allow the application to start gracefully with clear error messages instead of crashing, preventing complete service outages when WEBUI_SECRET_KEY mismatches occur during database migrations or environment changes. [#18094](https://github.com/open-webui/open-webui/pull/18094), [#18092](https://github.com/open-webui/open-webui/issues/18092)
- 🔐 OAuth 2.1 server discovery now correctly attempts all configured discovery URLs in sequence instead of only trying the first URL. [#17906](https://github.com/open-webui/open-webui/pull/17906), [#17904](https://github.com/open-webui/open-webui/issues/17904), [#18026](https://github.com/open-webui/open-webui/pull/18026)
- 🔐 Login redirect now correctly honors the redirect query parameter after authentication, ensuring users are returned to their intended destination with query parameters intact instead of defaulting to the homepage. [#18071](https://github.com/open-webui/open-webui/issues/18071)
- ☁️ OneDrive Business integration authentication regression is resolved, ensuring the popup now properly triggers when connecting to OneDrive accounts. [#17902](https://github.com/open-webui/open-webui/pull/17902), [#17825](https://github.com/open-webui/open-webui/discussions/17825), [#17816](https://github.com/open-webui/open-webui/issues/17816)
- 👥 Default group settings now persist correctly after page navigation, ensuring configuration changes are properly saved and retained. [#17899](https://github.com/open-webui/open-webui/issues/17899), [#18003](https://github.com/open-webui/open-webui/issues/18003)
- 📁 Folder data integrity is now verified on retrieval, automatically fixing orphaned folders with invalid parent references and ensuring proper cascading deletion of nested folder structures. [Commit](https://github.com/open-webui/open-webui/commit/5448618dd5ea181b9635b77040cef60926a902ff)
- 🗄️ Redis Sentinel and Redis Cluster configurations with the experimental ENABLE_STAR_SESSIONS_MIDDLEWARE feature are now properly isolated by making the feature opt-in only, preventing ReadOnlyError failures when connecting to read replicas in multi-node Redis deployments. [#18073](https://github.com/open-webui/open-webui/issues/18073)
- 📊 Mermaid and Vega diagram rendering now displays error toast notifications when syntax errors are detected, helping users identify and fix diagram issues instead of silently failing. [#18068](https://github.com/open-webui/open-webui/pull/18068)
- 🤖 Reasoning models that return reasoning_content instead of content no longer cause NoneType errors during chat title generation, follow-up suggestions, and tag generation. [#18080](https://github.com/open-webui/open-webui/pull/18080)
- 📚 Citation rendering now correctly handles multiple source references in a single bracket, parsing formats like [1,2] and [1, 2] into separate clickable citation links. [#18120](https://github.com/open-webui/open-webui/pull/18120)
- 🔍 Web search now handles individual source failures gracefully, continuing to process remaining sources instead of failing entirely when a single URL is unreachable or returns an error. [Commit](https://github.com/open-webui/open-webui/commit/e000494e488090c5f66989a2b3f89d3eaeb7946b), [Commit](https://github.com/open-webui/open-webui/commit/53e98620bff38ab9280aee5165af0a704bdd99b9)
- 🔍 Hybrid search with reranking now handles empty result sets gracefully instead of crashing with ValueError when all results are filtered out due to relevance thresholds. [#18096](https://github.com/open-webui/open-webui/issues/18096)
- 🔍 Reranking models without defined padding tokens now work correctly by automatically falling back to eos_token_id as pad_token_id, fixing "Cannot handle batch sizes > 1" errors for models like Qwen3-Reranker. [#18108](https://github.com/open-webui/open-webui/pull/18108), [#16027](https://github.com/open-webui/open-webui/discussions/16027)
- 🔍 Model selector search now correctly returns results for non-admin users by dynamically updating the search index when the model list changes, fixing a race condition that caused empty search results. [#17996](https://github.com/open-webui/open-webui/pull/17996), [#17960](https://github.com/open-webui/open-webui/pull/17960)
- ⚡ Task model function calling performance is improved by excluding base64 image data from payloads, significantly reducing token count and memory usage when images are present in conversations. [#17897](https://github.com/open-webui/open-webui/pull/17897)
- 🤖 Text selection "Ask" action now correctly recognizes and uses local models configured via direct connections instead of only showing external provider models. [#17896](https://github.com/open-webui/open-webui/issues/17896)
- 🛑 Task cancellation API now returns accurate response status, correctly reporting successful cancellations instead of incorrectly indicating failures. [#17920](https://github.com/open-webui/open-webui/issues/17920)
- 💬 Follow-up query suggestions are now generated and displayed in temporary chats, matching the behavior of saved chats. [#14987](https://github.com/open-webui/open-webui/issues/14987)
- 🔊 Azure Text-to-Speech now properly escapes special characters like ampersands in SSML, preventing HTTP 400 errors and ensuring audio generation succeeds for all text content. [#17962](https://github.com/open-webui/open-webui/issues/17962)
- 🛠️ OpenAPI tool server calls with optional parameters now execute successfully even when no arguments are provided, removing the incorrect requirement for a request body. [#18036](https://github.com/open-webui/open-webui/issues/18036)
- 🛠️ MCP mode tool server connections no longer incorrectly validate the OpenAPI path field, allowing seamless switching between OpenAPI and MCP connection types. [#17989](https://github.com/open-webui/open-webui/pull/17989), [#17988](https://github.com/open-webui/open-webui/issues/17988)
- 🛠️ Third-party tool responses containing non-UTF8 or invalid byte sequences are now handled gracefully without causing request failures. [#17882](https://github.com/open-webui/open-webui/pull/17882)
- 🎨 Workspace filter dropdown now correctly renders model tags as strings instead of displaying individual characters, fixing broken filtering interface when models have multiple tags. [#18034](https://github.com/open-webui/open-webui/issues/18034)
- ⌨️ Ctrl+Enter keyboard shortcut now correctly sends messages in mobile and narrow browser views on Chrome instead of inserting newlines. [#17975](https://github.com/open-webui/open-webui/issues/17975)
- ⌨️ Tab characters are now preserved when pasting code or formatted text into the chat input box in plain text mode. [#17958](https://github.com/open-webui/open-webui/issues/17958)
- 📋 Text selection copying from the chat input box now correctly copies only the selected text instead of the entire textbox content. [#17911](https://github.com/open-webui/open-webui/issues/17911)
- 🔍 Web search query logging now uses debug level instead of info level, preventing user search queries from appearing in production logs. [#17888](https://github.com/open-webui/open-webui/pull/17888)
- 📝 Debug print statements in middleware were removed to prevent excessive log pollution and respect configured logging levels. [#17943](https://github.com/open-webui/open-webui/issues/17943)
### Changed
- 🗄️ Milvus vector database dependency is updated from pymilvus 2.5.0 to 2.6.2, ensuring compatibility with newer Milvus versions but requiring users on older Milvus instances to either upgrade their database or manually downgrade the pymilvus package. [#18066](https://github.com/open-webui/open-webui/pull/18066)
## [0.6.32] - 2025-09-29
### Added
- ⚡ JSON model import moved to backend processing for significant performance improvements when importing large model files. [#17871](https://github.com/open-webui/open-webui/pull/17871)
- ⚠️ Visual warnings for group permissions that display when a permission is disabled in a group but remains enabled in the default user role, clarifying inheritance behavior for administrators. [#17848](https://github.com/open-webui/open-webui/pull/17848)
- 🗄️ Milvus multi-tenancy mode using shared collections with resource ID filtering for improved scalability, mirroring the existing Qdrant implementation and configurable via ENABLE_MILVUS_MULTITENANCY_MODE environment variable. [#17837](https://github.com/open-webui/open-webui/pull/17837)
- 🛠️ Enhanced tool result processing with improved error handling, better MCP tool result handling, and performance improvements for embedded UI components. [Commit](https://github.com/open-webui/open-webui/commit/4f06f29348b2c9d71c87d1bbe5b748a368f5101f)
- 👥 New user groups now automatically inherit default group permissions, streamlining the admin setup process by eliminating manual permission configuration. [#17843](https://github.com/open-webui/open-webui/pull/17843)
- 🗂️ Bulk unarchive functionality for all chats, providing a single backend endpoint to efficiently restore all archived chats at once. [#17857](https://github.com/open-webui/open-webui/pull/17857)
- 🏷️ Browser tab title toggle setting allows users to control whether chat titles appear in the browser tab or display only "Open WebUI". [#17851](https://github.com/open-webui/open-webui/pull/17851)
- 💬 Reply-to-message functionality in channels, allowing users to reply directly to specific messages with visual threading and context display. [Commit](https://github.com/open-webui/open-webui/commit/1a18928c94903ad1f1f0391b8ade042c3e60205b)
- 🔧 Tool server import and export functionality, allowing direct upload of openapi.json and openapi.yaml files as an alternative to URL-based configuration. [#14446](https://github.com/open-webui/open-webui/issues/14446)
- 🔧 User valve configuration for Functions is now available in the integration menu, providing consistent management alongside Tools. [#17784](https://github.com/open-webui/open-webui/issues/17784)
- 🔐 Admin permission toggle for controlling public sharing of notes, configurable via USER_PERMISSIONS_NOTES_ALLOW_PUBLIC_SHARING environment variable. [#17801](https://github.com/open-webui/open-webui/pull/17801), [Docs:#715](https://github.com/open-webui/docs/pull/715)
- 🗄️ DISKANN index type support for Milvus vector database with configurable maximum degree and search list size parameters. [#17770](https://github.com/open-webui/open-webui/pull/17770), [Docs:Commit](https://github.com/open-webui/docs/commit/cec50ab4d4b659558ca1ccd4b5e6fc024f05fb83)
- 🔄 Various improvements were implemented across the frontend and backend to enhance performance, stability, and security.
- 🌐 Translations for Chinese (Simplified & Traditional) and Bosnian (Latin) were enhanced and expanded.
### Fixed
- 🛠️ MCP tool calls are now correctly routed to the appropriate server when multiple streamable-http MCP servers are enabled, preventing "Tool not found" errors. [#17817](https://github.com/open-webui/open-webui/issues/17817)
- 🛠️ External tool servers (OpenAPI/MCP) now properly process and return tool results to the model, restoring functionality that was broken in v0.6.31. [#17764](https://github.com/open-webui/open-webui/issues/17764)
- 🔧 User valve detection now correctly identifies valves in imported tool code, ensuring gear icons appear in the integrations menu for all tools with user valves. [#17765](https://github.com/open-webui/open-webui/issues/17765)
- 🔐 MCP OAuth discovery now correctly handles multi-tenant configurations by including subpaths in metadata URL discovery. [#17768](https://github.com/open-webui/open-webui/issues/17768)
- 🗄️ Milvus query operations now correctly use -1 instead of None for unlimited queries, preventing TypeError exceptions. [#17769](https://github.com/open-webui/open-webui/pull/17769), [#17088](https://github.com/open-webui/open-webui/issues/17088)
- 📁 File upload error messages are now displayed when files are modified during upload, preventing user confusion on Android and Windows devices. [#17777](https://github.com/open-webui/open-webui/pull/17777)
- 🎨 MessageInput Integrations button hover effect now displays correctly with proper visual feedback. [#17767](https://github.com/open-webui/open-webui/pull/17767)
- 🎯 "Set as default" label positioning is fixed to ensure it remains clickable in all scenarios, including multi-model configurations. [#17779](https://github.com/open-webui/open-webui/pull/17779)
- 🎛️ Floating buttons now correctly retrieve message context by using the proper messageId parameter in createMessagesList calls. [#17823](https://github.com/open-webui/open-webui/pull/17823)
- 📌 Pinned chats are now properly cleared from the sidebar after archiving all chats, ensuring UI consistency without requiring a page refresh. [#17832](https://github.com/open-webui/open-webui/pull/17832)
- 🗑️ Delete confirmation modals now properly truncate long names for Notes, Prompts, Tools, and Functions to prevent modal overflow. [#17812](https://github.com/open-webui/open-webui/pull/17812)
- 🌐 Internationalization function calls now use proper Svelte store subscription syntax, preventing "i18n.t is not a function" errors on the model creation page. [#17819](https://github.com/open-webui/open-webui/pull/17819)
- 🎨 Playground chat interface button layout is corrected to prevent vertical text rendering for Assistant/User role buttons. [#17819](https://github.com/open-webui/open-webui/pull/17819)
- 🏷️ UI text truncation is improved across multiple components including usernames in admin panels, arena model names, model tags, and filter tags to prevent layout overflow issues. [#17805](https://github.com/open-webui/open-webui/pull/17805), [#17803](https://github.com/open-webui/open-webui/pull/17803), [#17791](https://github.com/open-webui/open-webui/pull/17791), [#17796](https://github.com/open-webui/open-webui/pull/17796)
## [0.6.31] - 2025-09-25
### Added
- 🔌 MCP (streamable HTTP) server support was added alongside existing OpenAPI server integration, allowing users to connect both server types through an improved server configuration interface. [#15932](https://github.com/open-webui/open-webui/issues/15932) [#16651](https://github.com/open-webui/open-webui/pull/16651), [Commit](https://github.com/open-webui/open-webui/commit/fd7385c3921eb59af76a26f4c475aedb38ce2406), [Commit](https://github.com/open-webui/open-webui/commit/777e81f7a8aca957a359d51df8388e5af4721a68), [Commit](https://github.com/open-webui/open-webui/commit/de7f7b3d855641450f8e5aac34fbae0665e0b80e), [Commit](https://github.com/open-webui/open-webui/commit/f1bbf3a91e4713039364b790e886e59b401572d0), [Commit](https://github.com/open-webui/open-webui/commit/c55afc42559c32a6f0c8beb0f1bb18e9360ab8af), [Commit](https://github.com/open-webui/open-webui/commit/61f20acf61f4fe30c0e5b0180949f6e1a8cf6524)
- 🔐 To enable MCP server authentication, OAuth 2.1 dynamic client registration was implemented with secure automatic client registration, encrypted session management, and seamless authentication flows. [Commit](https://github.com/open-webui/open-webui/commit/972be4eda5a394c111e849075f94099c9c0dd9aa), [Commit](https://github.com/open-webui/open-webui/commit/77e971dd9fbeee806e2864e686df5ec75e82104b), [Commit](https://github.com/open-webui/open-webui/commit/879abd7feea3692a2f157da4a458d30f27217508), [Commit](https://github.com/open-webui/open-webui/commit/422d38fd114b1ebd8a7dbb114d64e14791e67d7a), [Docs:#709](https://github.com/open-webui/docs/pull/709)
- 🛠️ External & Built-In Tools can now support rich UI element embedding ([Docs](https://docs.openwebui.com/features/plugin/tools/development)), allowing tools to return HTML content and interactive iframes that display directly within chat conversations with configurable security settings. [Commit](https://github.com/open-webui/open-webui/commit/07c5b25bc8b63173f406feb3ba183d375fedee6a), [Commit](https://github.com/open-webui/open-webui/commit/a5d8882bba7933a2c2c31c0a1405aba507c370bb), [Commit](https://github.com/open-webui/open-webui/commit/7be5b7f50f498de97359003609fc5993a172f084), [Commit](https://github.com/open-webui/open-webui/commit/a89ffccd7e96705a4a40e845289f4fcf9c4ae596)
- 📝 Note editor now supports drag-and-drop reordering of list items with visual drag handles, making list organization more intuitive and efficient. [Commit](https://github.com/open-webui/open-webui/commit/e4e97e727e9b4971f1c363b1280ca3a101599d88), [Commit](https://github.com/open-webui/open-webui/commit/aeb5288a3c7a6e9e0a47b807cc52f870c1b7dbe6)
- 🔍 Search modal was enhanced with quick action buttons for starting new conversations and creating notes, with intelligent content pre-population from search queries. [Commit](https://github.com/open-webui/open-webui/commit/aa6f63a335e172fec1dc94b2056541f52c1167a6), [Commit](https://github.com/open-webui/open-webui/commit/612a52d7bb7dbe9fa0bbbc8ac0a552d2b9801146), [Commit](https://github.com/open-webui/open-webui/commit/b03529b006f3148e895b1094584e1ab129ecac5b)
- 🛠️ Tool user valve configuration interface was added to the integrations menu, displaying clickable gear icon buttons with tooltips for tools that support user-specific settings, making personal tool configurations easily accessible. [Commit](https://github.com/open-webui/open-webui/commit/27d61307cdce97ed11a05ec13fc300249d6022cd)
- 👥 Channel access control was enhanced to require write permissions for posting, editing, and deleting messages, while read-only users can view content but cannot contribute. [#17543](https://github.com/open-webui/open-webui/pull/17543)
- 💬 Channel models now support image processing, allowing AI assistants to view and analyze images shared in conversation threads. [Commit](https://github.com/open-webui/open-webui/commit/9f0010e234a6f40782a66021435d3c02b9c23639)
- 🌐 Attach Webpage button was added to the message input menu, providing a user-friendly modal interface for attaching web content and YouTube videos as an alternative to the existing URL syntax. [#17534](https://github.com/open-webui/open-webui/pull/17534)
- 🔐 Redis session storage support was added for OAuth redirects, providing better state handling in multi-pod Kubernetes deployments and resolving CSRF mismatch errors. [#17223](https://github.com/open-webui/open-webui/pull/17223), [#15373](https://github.com/open-webui/open-webui/issues/15373)
- 🔍 Ollama Cloud web search integration was added as a new search engine option, providing access to web search functionality through Ollama's cloud infrastructure. [Commit](https://github.com/open-webui/open-webui/commit/e06489d92baca095b8f376fbef223298c7772579), [Commit](https://github.com/open-webui/open-webui/commit/4b6d34438bcfc45463dc7a9cb984794b32c1f0a1), [Commit](https://github.com/open-webui/open-webui/commit/05c46008da85357dc6890b846789dfaa59f4a520), [Commit](https://github.com/open-webui/open-webui/commit/fe65fe0b97ec5a8fff71592ff04a25c8e123d108), [Docs:#708](https://github.com/open-webui/docs/pull/708)
- 🔍 Perplexity Websearch API integration was added as a new search engine option, providing access to the new websearch functionality provided by Perplexity. [#17756](https://github.com/open-webui/open-webui/issues/17756), [Commit](https://github.com/open-webui/open-webui/pull/17747/commits/7f411dd5cc1c29733216f79e99eeeed0406a2afe)
- ☁️ OneDrive integration was improved to support separate client IDs for personal and business authentication, enabling both integrations to work simultaneously. [#17619](https://github.com/open-webui/open-webui/pull/17619), [Docs](https://docs.openwebui.com/tutorials/integrations/onedrive-sharepoint), [Docs](https://docs.openwebui.com/getting-started/env-configuration/#onedrive)
- 📝 Pending user overlay content now supports markdown formatting, enabling rich text display for custom messages similar to banner functionality. [#17681](https://github.com/open-webui/open-webui/pull/17681)
- 🎨 Image generation model selection was centralized to enable dynamic model override in function calls, allowing pipes and tools to specify different models than the global default while maintaining backward compatibility. [#17689](https://github.com/open-webui/open-webui/pull/17689)
- 🎨 Interface design was modernized with updated visual styling, improved spacing, and refined component layouts across modals, sidebar, settings, and navigation elements. [Commit](https://github.com/open-webui/open-webui/commit/27a91cc80a24bda0a3a188bc3120a8ab57b00881), [Commit](https://github.com/open-webui/open-webui/commit/4ad743098615f9c58daa9df392f31109aeceeb16), [Commit](https://github.com/open-webui/open-webui/commit/fd7385c3921eb59af76a26f4c475aedb38ce2406)
- 📊 Notes query performance was optimized through database-level filtering and separated access control logic, reducing memory usage and eliminating N+1 query problems for better scalability. [#17607](https://github.com/open-webui/open-webui/pull/17607) [Commit](https://github.com/open-webui/open-webui/pull/17747/commits/da661756fa7eec754270e6dd8c67cbf74a28a17f)
- ⚡ Page loading performance was optimized by deferring API requests until components are actually opened, including ChangelogModal, ModelSelector, RecursiveFolder, ArchivedChatsModal, and SearchModal. [#17542](https://github.com/open-webui/open-webui/pull/17542), [#17555](https://github.com/open-webui/open-webui/pull/17555), [#17557](https://github.com/open-webui/open-webui/pull/17557), [#17541](https://github.com/open-webui/open-webui/pull/17541), [#17640](https://github.com/open-webui/open-webui/pull/17640)
- ⚡ Bundle size was reduced by 1.58MB through optimized highlight.js language support, improving page loading speed and reducing bandwidth usage. [#17645](https://github.com/open-webui/open-webui/pull/17645)
- ⚡ Editor collaboration functionality was refactored to reduce package size by 390KB and minimize compilation errors, improving build performance and reliability. [#17593](https://github.com/open-webui/open-webui/pull/17593)
- ♿ Enhanced user interface accessibility through the addition of unique element IDs, improving targeting for testing, styling, and assistive technologies while providing better semantic markup for screen readers and accessibility tools. [#17746](https://github.com/open-webui/open-webui/pull/17746)
- 🔄 Various improvements were implemented across the frontend and backend to enhance performance, stability, and security.
- 🌐 Translations for Portuguese (Brazil), Chinese (Simplified and Traditional), Korean, Irish, Spanish, Finnish, French, Kabyle, Russian, and Catalan were enhanced and improved.
### Fixed
- 🛡️ SVG content security was enhanced by implementing DOMPurify sanitization to prevent XSS attacks through malicious SVG elements, ensuring safe rendering of user-generated SVG content. [Commit](https://github.com/open-webui/open-webui/pull/17747/commits/750a659a9fee7687e667d9d755e17b8a0c77d557)
- ☁️ OneDrive attachment menu rendering issues were resolved by restructuring the submenu interface from dropdown to tabbed navigation, preventing menu items from being hidden or clipped due to overflow constraints. [#17554](https://github.com/open-webui/open-webui/issues/17554), [Commit](https://github.com/open-webui/open-webui/pull/17747/commits/90e4b49b881b644465831cc3028bb44f0f7a2196)
- 💬 Attached conversation references now persist throughout the entire chat session, ensuring models can continue querying referenced conversations after multiple conversation turns. [#17750](https://github.com/open-webui/open-webui/issues/17750)
- 🔍 Search modal text box focus issues after pinning or unpinning chats were resolved, allowing users to properly exit the search interface by clicking outside the text box. [#17743](https://github.com/open-webui/open-webui/issues/17743)
- 🔍 Search function chat list is now properly updated in real-time when chats are created or deleted, eliminating stale search results and preview loading failures. [#17741](https://github.com/open-webui/open-webui/issues/17741)
- 💬 Chat jitter and delayed code block expansion in multi-model sessions were resolved by reverting dynamic CodeEditor loading, restoring stable rendering behavior. [#17715](https://github.com/open-webui/open-webui/pull/17715), [#17684](https://github.com/open-webui/open-webui/issues/17684)
- 📎 File upload handling was improved to properly recognize uploaded files even when no accompanying text message is provided, resolving issues where attachments were ignored in custom prompts. [#17492](https://github.com/open-webui/open-webui/issues/17492)
- 💬 Chat conversation referencing within projects was restored by including foldered chats in the reference menu, allowing users to properly quote conversations from within their project scope. [#17530](https://github.com/open-webui/open-webui/issues/17530)
- 🔍 RAG query generation is now skipped when all attached files are set to full context mode, preventing unnecessary retrieval operations and improving system efficiency. [#17744](https://github.com/open-webui/open-webui/pull/17744)
- 💾 Memory leaks in file handling and HTTP connections are prevented through proper resource cleanup, ensuring stable memory usage during large file downloads and processing operations. [#17608](https://github.com/open-webui/open-webui/pull/17608)
- 🔐 OAuth access token refresh errors are resolved by properly implementing async/await patterns, preventing "coroutine object has no attribute get" failures during token expiry. [#17585](https://github.com/open-webui/open-webui/issues/17585), [#17678](https://github.com/open-webui/open-webui/issues/17678)
- ⚙️ Valve behavior was improved to properly handle default values and array types, ensuring only explicitly set values are persisted while maintaining consistent distinction between custom and default valve states. [#17664](https://github.com/open-webui/open-webui/pull/17664)
- 🔍 Hybrid search functionality was enhanced to handle inconsistent parameter types and prevent failures when collection results are None, empty, or in unexpected formats. [#17617](https://github.com/open-webui/open-webui/pull/17617)
- 📁 Empty folder deletion is now allowed regardless of chat deletion permission restrictions, resolving cases where users couldn't remove folders after deleting all contained chats. [#17683](https://github.com/open-webui/open-webui/pull/17683)
- 📝 Rich text editor console errors were resolved by adding proper error handling when the TipTap editor view is not available or not yet mounted. [#17697](https://github.com/open-webui/open-webui/issues/17697)
- 🗒️ Hidden models are now properly excluded from the notes section dropdown and default model selection, preventing users from accessing models they shouldn't see. [#17722](https://github.com/open-webui/open-webui/pull/17722)
- 🖼️ AI-generated image download filenames now use a clean, translatable "Generated Image" format instead of potentially problematic response text, improving file management and compatibility. [#17721](https://github.com/open-webui/open-webui/pull/17721)
- 🎨 Toggle switch display issues in the Integrations interface are fixed, preventing background highlighting and obscuring on hover. [#17564](https://github.com/open-webui/open-webui/issues/17564)
### Changed
- 👥 Channel permissions now require write access for message posting, editing, and deletion, with existing user groups defaulting to read-only access requiring manual admin migration to write permissions for full participation.
- ☁️ OneDrive environment variable configuration was updated to use separate ONEDRIVE_CLIENT_ID_PERSONAL and ONEDRIVE_CLIENT_ID_BUSINESS variables for better client ID separation, while maintaining backward compatibility with the legacy ONEDRIVE_CLIENT_ID variable. [Docs](https://docs.openwebui.com/tutorials/integrations/onedrive-sharepoint), [Docs](https://docs.openwebui.com/getting-started/env-configuration/#onedrive)
## [0.6.30] - 2025-09-17
### Added
- 🔑 Microsoft Entra ID authentication type support was added for Azure OpenAI connections, enabling enhanced security and streamlined authentication workflows.
### Fixed
- ☁️ OneDrive integration was fixed after recent breakage, restoring reliable account connectivity and file access.
## [0.6.29] - 2025-09-17
### Added
- 🎨 The chat input menu has been completely overhauled with a revolutionary new design, consolidating attachments under a unified '+' button, organizing integrations into a streamlined options menu, and introducing powerful, interactive selectors for attaching chats, notes, and knowledge base items. [Commit](https://github.com/open-webui/open-webui/commit/a68342d5a887e36695e21f8c2aec593b159654ff), [Commit](https://github.com/open-webui/open-webui/commit/96b8aaf83ff341fef432649366bc5155bac6cf20), [Commit](https://github.com/open-webui/open-webui/commit/4977e6d50f7b931372c96dd5979ca635d58aeb78), [Commit](https://github.com/open-webui/open-webui/commit/d973db829f7ec98b8f8fe7d3b2822d588e79f94e), [Commit](https://github.com/open-webui/open-webui/commit/d4c628de09654df76653ad9bce9cb3263e2f27c8), [Commit](https://github.com/open-webui/open-webui/commit/cd740f436db4ea308dbede14ef7ff56e8126f51b), [Commit](https://github.com/open-webui/open-webui/commit/5c2db102d06b5c18beb248d795682ff422e9b6d1), [Commit](https://github.com/open-webui/open-webui/commit/031cf38655a1a2973194d2eaa0fbbd17aca8ee92), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/3ed0a6d11fea1a054e0bc8aa8dfbe417c7c53e51), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/eadec9e86e01bc8f9fb90dfe7a7ae4fc3bfa6420), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/c03ca7270e64e3a002d321237160c0ddaf2bb129), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/b53ddfbd19aa94e9cbf7210acb31c3cfafafa5fe), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/c923461882fcde30ae297a95e91176c95b9b72e1)
- 🤖 AI models can now be mentioned in channels to automatically generate responses, enabling multi-model conversations where mentioned models participate directly in threaded discussions with full context awareness. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/4fe97d8794ee18e087790caab9e5d82886006145)
- 💬 The Channels feature now utilizes the modern rich text editor, including support for '/', '@', and '#' command suggestions. [Commit](https://github.com/open-webui/open-webui/commit/06c1426e14ac0dfaf723485dbbc9723a4d89aba9), [Commit](https://github.com/open-webui/open-webui/commit/02f7c3258b62970ce79716f75d15467a96565054)
- 📎 Channel message input now supports direct paste functionality for images and files from the clipboard, streamlining content sharing workflows. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/6549fc839f86c40c26c2ef4dedcaf763a9304418)
- ⚙️ Models can now be configured with default features (Web Search, Image Generation) and filters that automatically activate when a user selects the model. [Commit](https://github.com/open-webui/open-webui/commit/9a555478273355a5177bfc7f7211c64778e4c8de), [Commit](https://github.com/open-webui/open-webui/commit/384a53b339820068e92f7eaea0d9f3e0536c19c2), [Commit](https://github.com/open-webui/open-webui/commit/d7f43bfc1a30c065def8c50d77c2579c1a3c5c67), [Commit](https://github.com/open-webui/open-webui/commit/6a67a2217cc5946ad771e479e3a37ac213210748)
- 💬 The ability to reference other chats as context within a conversation was added via the attachment menu. [Commit](https://github.com/open-webui/open-webui/commit/e097bbdf11ae4975c622e086df00d054291cdeb3), [Commit](https://github.com/open-webui/open-webui/commit/f3cd2ffb18e7dedbe88430f9ae7caa6b3cfd79d0), [Commit](https://github.com/open-webui/open-webui/commit/74263c872c5d574a9bb0944d7984f748dc772dba), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/aa8ab349ed2fcb46d1cf994b9c0de2ec2ea35d0d), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/025eef754f0d46789981defd473d001e3b1d0ca2)
- 🎨 The command suggestion UI for prompts ('/'), models ('@'), and knowledge ('#') was completely overhauled with a more responsive and keyboard-navigable interface. [Commit](https://github.com/open-webui/open-webui/commit/6b69c4da0fb9329ccf7024483960e070cf52ccab), [Commit](https://github.com/open-webui/open-webui/commit/06a6855f844456eceaa4d410c93379460e208202), [Commit](https://github.com/open-webui/open-webui/commit/c55f5578280b936cf581a743df3703e3db1afd54), [Commit](https://github.com/open-webui/open-webui/commit/f68d1ba394d4423d369f827894cde99d760b2402)
- 👥 User and channel suggestions were added to the mention system, enabling '@' mentions for users and models, and '#' mentions for channels with searchable user lookup and clickable navigation. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/bbd1d2b58c89b35daea234f1fc9208f2af840899), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/aef1e06f0bb72065a25579c982dd49157e320268), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/779db74d7e9b7b00d099b7d65cfbc8a831e74690)
- 📁 Folder functionality was enhanced with custom background image support, improved drag-and-drop capabilities for moving folders to root level, and better menu interactions. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/2a234829f5dfdfde27fdfd30591caa908340efb4), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/2b1ee8b0dc5f7c0caaafdd218f20705059fa72e2), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/b1e5bc8e490745f701909c19b6a444b67c04660e), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/3e584132686372dfeef187596a7c557aa5f48308)
- ☁️ OneDrive integration configuration now supports selecting between personal and work/school account types via ENABLE_ONEDRIVE_PERSONAL and ENABLE_ONEDRIVE_BUSINESS environment variables. [#17354](https://github.com/open-webui/open-webui/pull/17354), [Commit](https://github.com/open-webui/open-webui/commit/e1e3009a30f9808ce06582d81a60e391f5ca09ec), [Docs:#697](https://github.com/open-webui/docs/pull/697)
- ⚡ Mermaid.js is now dynamically loaded on demand, significantly reducing first-screen loading time and improving initial page performance. [#17476](https://github.com/open-webui/open-webui/issues/17476), [#17477](https://github.com/open-webui/open-webui/pull/17477)
- ⚡ Azure MSAL browser library is now dynamically loaded on demand, reducing initial bundle size by 730KB and improving first-screen loading speed. [#17479](https://github.com/open-webui/open-webui/pull/17479)
- ⚡ CodeEditor component is now dynamically loaded on demand, reducing initial bundle size by 1MB and improving first-screen loading speed. [#17498](https://github.com/open-webui/open-webui/pull/17498)
- ⚡ Hugging Face Transformers library is now dynamically loaded on demand, reducing initial bundle size by 1.9MB and improving first-screen loading speed. [#17499](https://github.com/open-webui/open-webui/pull/17499)
- ⚡ jsPDF and html2canvas-pro libraries are now dynamically loaded on demand, reducing initial bundle size by 980KB and improving first-screen loading speed. [#17502](https://github.com/open-webui/open-webui/pull/17502)
- ⚡ Leaflet mapping library is now dynamically loaded on demand, reducing initial bundle size by 454KB and improving first-screen loading speed. [#17503](https://github.com/open-webui/open-webui/pull/17503)
- 📊 OpenTelemetry metrics collection was enhanced to properly handle HTTP 500 errors and ensure metrics are recorded even during exceptions. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/b14617a653c6bdcfd3102c12f971924fd1faf572)
- 🔒 OAuth token retrieval logic was refactored, improving the reliability and consistency of authentication handling across the backend. [Commit](https://github.com/open-webui/open-webui/commit/6c0a5fa91cdbf6ffb74667ee61ca96bebfdfbc50)
- 💻 Code block output processing was improved to handle Python execution results more reliably, along with refined visual styling and button layouts. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/0e5320c39e308ff97f2ca9e289618af12479eb6e)
- ⚡ Message input processing was optimized to skip unnecessary text variable handling when input is empty, improving performance. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/e1386fe80b77126a12dabc4ad058abe9b024b275)
- 📄 Individual chat PDF export was added to the sidebar chat menu, allowing users to export single conversations as PDF documents with both stylized and plain text options. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/d041d58bb619689cd04a391b4f8191b23941ca62)
- 🛠️ Function validation was enhanced with improved valve validation and better error handling during function loading and synchronization. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/e66e0526ed6a116323285f79f44237538b6c75e6), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/8edfd29102e0a61777b23d3575eaa30be37b59a5)
- 🔔 Notification toast interaction was enhanced with drag detection to prevent accidental clicks and added keyboard support for accessibility. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/621e7679c427b6f0efa85f95235319238bf171ad)
- 🗓️ Improved date and time formatting dynamically adapts to the selected language, ensuring consistent localization across the UI. [#17409](https://github.com/open-webui/open-webui/pull/17409), [Commit](https://github.com/open-webui/open-webui/commit/2227f24bd6d861b1fad8d2cabacf7d62ce137d0c)
- 🔒 Feishu SSO integration was added, allowing users to authenticate via Feishu. [#17284](https://github.com/open-webui/open-webui/pull/17284), [Docs:#685](https://github.com/open-webui/docs/pull/685)
- 🔠 Toggle filters in the chat input options menu are now sorted alphabetically for easier navigation. [Commit](https://github.com/open-webui/open-webui/commit/ca853ca4656180487afcd84230d214f91db52533)
- 🎨 Long chat titles in the sidebar are now truncated to prevent text overflow and maintain a clean layout. [#17356](https://github.com/open-webui/open-webui/pull/17356)
- 🎨 Temporary chat interface design was refined with improved layout and visual consistency. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/67549dcadd670285d491bd41daf3d081a70fd094), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/2ca34217e68f3b439899c75881dfb050f49c9eb2), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/fb02ec52a5df3f58b53db4ab3a995c15f83503cd)
- 🎨 Download icon consistency was improved across the entire interface by standardizing the icon component used in menus, functions, tools, and export features. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/596be451ece7e11b5cd25465d49670c27a1cb33f)
- 🎨 Settings interface was enhanced with improved iconography and reorganized the 'Chats' section into 'Data Controls' for better clarity. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/8bf0b40fdd978b5af6548a6e1fb3aabd90bcd5cd)
- 🔄 Various improvements were implemented across the frontend and backend to enhance performance, stability, and security.
- 🌐 Translations for Finnish, German, Kabyle, Portuguese (Brazil), Simplified Chinese, Spanish (Spain), and Traditional Chinese (Taiwan) were enhanced and expanded.
### Fixed
- 📚 Knowledge base permission logic was corrected to ensure private collection owners can access their own content when embedding bypass is enabled. [#17432](https://github.com/open-webui/open-webui/issues/17432), [Commit](https://github.com/open-webui/open-webui/commit/a51f0c30ec1472d71487eab3e15d0351a2716b12)
- ⚙️ Connection URL editing in Admin Settings now properly saves changes instead of reverting to original values, fixing issues with both Ollama and OpenAI-compatible endpoints. [#17435](https://github.com/open-webui/open-webui/issues/17435), [Commit](https://github.com/open-webui/open-webui/commit/e4c864de7eb0d577843a80688677ce3659d1f81f)
- 📊 Usage information collection from Google models was corrected to handle providers that send usage data alongside content chunks instead of separately. [#17421](https://github.com/open-webui/open-webui/pull/17421), [Commit](https://github.com/open-webui/open-webui/commit/c2f98a4cd29ed738f395fef09c42ab8e73cd46a0)
- ⚙️ Settings modal scrolling issue was resolved by moving image compression controls to a dedicated modal, preventing the main settings from becoming scrollable out of view. [#17474](https://github.com/open-webui/open-webui/issues/17474), [Commit](https://github.com/open-webui/open-webui/commit/fed5615c19b0045a55b0be426b468a57bfda4b66)
- 📁 Folder click behavior was improved to prevent accidental actions by implementing proper double-click detection and timing delays for folder expansion and selection. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/19e3214997170eea6ee92452e8c778e04a28e396)
- 🔐 Access control component reliability was improved with better null checking and error handling for group permissions and private access scenarios. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/c8780a7f934c5e49a21b438f2f30232f83cf75d2), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/32015c392dbc6b7367a6a91d9e173e675ea3402c)
- 🔗 The citation modal now correctly displays and links to external web page sources in addition to internal documents. [Commit](https://github.com/open-webui/open-webui/commit/9208a84185a7e59524f00a7576667d493c3ac7d4)
- 🔗 Web and YouTube attachment handling was fixed, ensuring their content is now reliably processed and included in the chat context for retrieval. [Commit](https://github.com/open-webui/open-webui/commit/210197fd438b52080cda5d6ce3d47b92cdc264c8)
- 📂 Large file upload failures are resolved by correcting the processing logic for scenarios where document embedding is bypassed. [Commit](https://github.com/open-webui/open-webui/commit/051b6daa8299fd332503bd584563556e2ae6adab)
- 🌐 Rich text input placeholder text now correctly updates when the interface language is switched, ensuring proper localization. [#17473](https://github.com/open-webui/open-webui/pull/17473), [Commit](https://github.com/open-webui/open-webui/commit/77358031f5077e6efe5cc08d8d4e5831c7cd1cd9)
- 📊 Llama.cpp server timing metrics are now correctly parsed and displayed by fixing a typo in the response handling. [#17350](https://github.com/open-webui/open-webui/issues/17350), [Commit](https://github.com/open-webui/open-webui/commit/cf72f5503f39834b9da44ebbb426a3674dad0caa)
- 🛠️ Filter functions with file_handler configuration now properly handle messages without file attachments, preventing runtime errors. [#17423](https://github.com/open-webui/open-webui/pull/17423)
- 🔔 Channel notification delivery was fixed to properly handle background task execution and user access checking. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/1077b2ac8b96e49c2ad2620e76eb65bbb2a3a1f3)
### Changed
- 📝 Prompt template variables are now optional by default instead of being forced as required, allowing flexible workflows with optional metadata fields. [#17447](https://github.com/open-webui/open-webui/issues/17447), [Commit](https://github.com/open-webui/open-webui/commit/d5824b1b495fcf86e57171769bcec2a0f698b070), [Docs:#696](https://github.com/open-webui/docs/pull/696)
- 🛠️ Direct external tool servers now require explicit user selection from the input interface instead of being automatically included in conversations, providing better control over tool usage. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/0f04227c34ca32746c43a9323e2df32299fcb6af), [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/99bba12de279dd55c55ded35b2e4f819af1c9ab5)
- 📺 Widescreen mode option was removed from Channels interface, with all channel layouts now using full-width display. [Commit](https://github.com/open-webui/open-webui/pull/17420/commits/d46b7b8f1b99a8054b55031fe935c8a16d5ec956)
- 🎛️ The plain textarea input option was deprecated, and the custom text editor is now the standard for all chat inputs. [Commit](https://github.com/open-webui/open-webui/commit/153afd832ccd12a1e5fd99b085008d080872c161)
## [0.6.28] - 2025-09-10
### Added
- 🔍 The "@" command for model selection now supports real-time search and filtering, improving usability and aligning its behavior with other input commands. [#17307](https://github.com/open-webui/open-webui/issues/17307), [Commit](https://github.com/open-webui/open-webui/commit/f2a09c71499489ee71599af4a179e7518aaf658b)
- 🛠️ External tool server data handling is now more robust, automatically attempting to parse specifications as JSON before falling back to YAML, regardless of the URL extension. [Commit](https://github.com/open-webui/open-webui/commit/774c0056bde88ed4831422efa81506488e3d6641)
- 🎯 The "Title" field is now automatically focused when creating a new chat folder, streamlining the folder creation process. [#17315](https://github.com/open-webui/open-webui/issues/17315), [Commit](https://github.com/open-webui/open-webui/commit/c51a651a2d5e2a27546416666812e9b92205562d)
- 🔄 Various improvements were implemented across the frontend and backend to enhance performance, stability, and security.
- 🌐 Brazilian Portuguese and Simplified Chinese translations were expanded and refined.
### Fixed
- 🔊 A regression affecting Text-to-Speech for local providers using the OpenAI engine was fixed by reverting a URL joining change. [#17316](https://github.com/open-webui/open-webui/issues/17316), [Commit](https://github.com/open-webui/open-webui/commit/8339f59cdfc63f2d58c8e26933d1bf1438479d75)
- 🪧 A regression was fixed where the input modal for prompts with placeholders would not open, causing the raw prompt text to be pasted into the chat input field instead. [#17325](https://github.com/open-webui/open-webui/issues/17325), [Commit](https://github.com/open-webui/open-webui/commit/d5cb65527eaa4831459a4c7dbf187daa9c0525ae)
- 🔑 An issue was resolved where modified connection keys in the OpenAIConnection component did not take effect. [#17324](https://github.com/open-webui/open-webui/pull/17324)
## [0.6.27] - 2025-09-09
### Added
- 📁 Emoji folder icons were added, allowing users to personalize workspace organization with visual cues, including improved chevron display. [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/1588f42fe777ad5d807e3f2fc8dbbc47a8db87c0), [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/b70c0f36c0f5bbfc2a767429984d6fba1a7bb26c), [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/11dea8795bfce42aa5d8d58ef316ded05173bd87), [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/c0a47169fa059154d5f5a9ea6b94f9a66d82f255)
- 📁 The 'Search Collection' input field now dynamically displays the total number of files within the knowledge base. [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/fbbe1117ae4c9c8fec6499d790eee275818eccc5)
- ☁️ A provider toggle in connection settings now allows users to manually specify Azure OpenAI deployments. [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/5bdd334b74fbd154085f2d590f4afdba32469c8a)
- ⚡ Model list caching performance was optimized by fixing cache key generation to reduce redundant API calls. [#17158](https://github.com/open-webui/open-webui/pull/17158)
- 🎨 Azure OpenAI image generation is now supported, with configurations for IMAGES_OPENAI_API_VERSION via environment variable and admin UI. [#17147](https://github.com/open-webui/open-webui/pull/17147), [#16274](https://github.com/open-webui/open-webui/discussions/16274), [Docs:#679](https://github.com/open-webui/docs/pull/679)
- ⚡ Comprehensive N+1 query performance is optimized by reducing database queries from 1+N to 1+1 patterns across major listing endpoints. [#17165](https://github.com/open-webui/open-webui/pull/17165), [#17160](https://github.com/open-webui/open-webui/pull/17160), [#17161](https://github.com/open-webui/open-webui/pull/17161), [#17162](https://github.com/open-webui/open-webui/pull/17162), [#17159](https://github.com/open-webui/open-webui/pull/17159), [#17166](https://github.com/open-webui/open-webui/pull/17166)
- ⚡ The PDF.js library is now dynamically loaded, significantly reducing initial page load size and improving responsiveness. [#17222](https://github.com/open-webui/open-webui/pull/17222)
- ⚡ The heic2any library is now dynamically loaded across various message input components, including channels, for faster page loads. [#17225](https://github.com/open-webui/open-webui/pull/17225), [#17229](https://github.com/open-webui/open-webui/pull/17229)
- 📚 The knowledge API now supports a "delete_file" query parameter, allowing configurable file deletion behavior. [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/22c4ef4fb096498066b73befe993ae3a82f7a8e7)
- 📊 Llama.cpp timing statistics are now integrated into the usage field for comprehensive model performance metrics. [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/e830b4959ecd4b2795e29e53026984a58a7696a9)
- 🗄️ The PGVECTOR_CREATE_EXTENSION environment variable now allows control over automatic pgvector extension creation. [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/c2b4976c82d335ed524bd80dc914b5e2f5bfbd9e), [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/b45219c8b15b48d5ee3d42983e1107bbcefbab01), [Docs:#672](https://github.com/open-webui/docs/pull/672)
- 🔒 Comprehensive server-side OAuth token management was implemented, securely storing encrypted tokens in a new database table and introducing an automatic refresh mechanism, enabling seamless and secure forwarding of valid user-specific OAuth tokens to downstream services, including OpenAI-compatible endpoints and external tool servers via the new "system_oauth" authentication type, resolving long-standing issues such as large token size limitations, stale/expired tokens, and reliable token propagation, and enhancing overall security by minimizing client-side token exposure, configurable via "ENABLE_OAUTH_ID_TOKEN_COOKIE" and "OAUTH_SESSION_TOKEN_ENCRYPTION_KEY" environment variables. [Docs:#683](https://github.com/open-webui/docs/pull/683), [#17210](https://github.com/open-webui/open-webui/pull/17210), [#8957](https://github.com/open-webui/open-webui/discussions/8957), [#11029](https://github.com/open-webui/open-webui/discussions/11029), [#17178](https://github.com/open-webui/open-webui/issues/17178), [#17183](https://github.com/open-webui/open-webui/issues/17183), [Commit](https://github.com/open-webui/open-webui/commit/217f4daef09b36d3d4cc4681e11d3ebd9984a1a5), [Commit](https://github.com/open-webui/open-webui/commit/fc11e4384fe98fac659e10596f67c23483578867), [Commit](https://github.com/open-webui/open-webui/commit/f11bdc6ab5dd5682bb3e27166e77581f5b8af3e0), [Commit](https://github.com/open-webui/open-webui/commit/f71834720e623761d972d4d740e9bbd90a3a86c6), [Commit](https://github.com/open-webui/open-webui/commit/b5bb6ae177dcdc4e8274d7e5ffa50bc8099fd466), [Commit](https://github.com/open-webui/open-webui/commit/b786d1e3f3308ef4f0f95d7130ddbcaaca4fc927), [Commit](https://github.com/open-webui/open-webui/commit/8a9f8627017bd0a74cbd647891552b26e56aabb7), [Commit](https://github.com/open-webui/open-webui/commit/30d1dc2c60e303756120fe1c5538968c4e6139f4), [Commit](https://github.com/open-webui/open-webui/commit/2b2d123531eb3f42c0e940593832a64e2806240d), [Commit](https://github.com/open-webui/open-webui/commit/6f6412dd16c63c2bb4df79a96b814bf69cb3f880)
- 🔒 Conditional Permission Hardening for OpenShift Deployments: Added a build argument to enable optional permission hardening for OpenShift and container environments. [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/0ebe4f8f8490451ac8e85a4846f010854d9b54e5)
- 👥 Regex pattern support is added for OAuth blocked groups, allowing more flexible group filtering rules. [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/df66e21472646648d008ebb22b0e8d5424d491df)
- 💬 Web search result display was enhanced to include titles and favicons, providing a clearer overview of search sources. [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/33f04a771455e3fabf8f0e8ebb994ae7f41b8ed4), [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/0a85dd4bca23022729eafdbc82c8c139fa365af2), [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/16090bc2721fde492afa2c4af5927e2b668527e1), [#17197](https://github.com/open-webui/open-webui/pull/17197), [#14179](https://github.com/open-webui/open-webui/issues/14179), [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/1cdb7aed1ee9bf81f2fd0404be52dcfa64f8ed4f), [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/f2525ebc447c008cf7269ef20ce04fa456f302c4), [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/7f523de408ede4075349d8de71ae0214b7e1a62e), [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/3d37e4a42d344051ae715ab59bd7b5718e46c343), [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/cd5e2be27b613314aadda6107089331783987985), [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/6dc0df247347aede2762fe2065cf30275fd137ae)
- 💬 A new setting was added to control whether clicking a suggested prompt automatically sends the message or only inserts the text. [#17192](https://github.com/open-webui/open-webui/issues/17192), [Commit](https://github.com/open-webui/open-webui/commit/e023a98f11fc52feb21e4065ec707cc98e50c7d3)
- 🔄 Various improvements were implemented across the frontend and backend to enhance performance, stability, and security.
- 🌐 Translations for Portuguese (Brazil), Simplified Chinese, Catalan, and Spanish were enhanced and expanded.
### Fixed
- 🔍 Hybrid search functionality now correctly handles lexical-semantic weight labels and avoids errors when BM25 weight is zero. [#17049](https://github.com/open-webui/open-webui/pull/17049), [#17046](https://github.com/open-webui/open-webui/issues/17046)
- 🛑 Task stopping errors are prevented by gracefully handling multiple stop requests for the same task. [#17195](https://github.com/open-webui/open-webui/pull/17195)
- 🐍 Code execution package detection precision is improved in Pyodide to prevent unnecessary package inclusions. [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/bbe116795860a81a647d9567e0d9cb1950650095)
- 🛠️ Tool message format API compliance is fixed by ensuring content fields in tool call responses contain valid string values instead of null. [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/37bf0087e5b8a324009c9d06b304027df351ea6b)
- 📱 Mobile app config API authentication now supports Authorization header token verification with cookie fallback for iOS and Android requests. [#17175](https://github.com/open-webui/open-webui/pull/17175)
- 💾 Knowledge file save race conditions are prevented by serializing API calls and adding an "isSaving" guard. [#17137](https://github.com/open-webui/open-webui/pull/17137), [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/4ca936f0bf9813bee11ec8aea41d7e34fb6b16a9)
- 🔐 The SSO login button visibility is restored for OIDC PKCE authentication without a client secret. [#17012](https://github.com/open-webui/open-webui/pull/17012)
- 🔊 Text-to-Speech (TTS) API requests now use proper URL joining methods, ensuring reliable functionality regardless of trailing slashes in the base URL. [#17061](https://github.com/open-webui/open-webui/pull/17061)
- 🛡️ Admin account creation on Hugging Face Spaces now correctly detects the configured port, resolving issues with custom port deployments. [#17064](https://github.com/open-webui/open-webui/pull/17064)
- 📁 Unicode filename support is improved for external document loaders by properly URL-encoding filenames in HTTP headers. [#17013](https://github.com/open-webui/open-webui/pull/17013), [#17000](https://github.com/open-webui/open-webui/issues/17000)
- 🔗 Web page and YouTube attachments are now correctly processed by setting their type as "text" and using collection names for accurate content retrieval. [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/487979859a6ffcfd60468f523822cdf838fbef5b)
- ✍️ Message input composition event handling is fixed to properly manage text input for multilingual users using Input Method Editors (IME). [#17085](https://github.com/open-webui/open-webui/pull/17085)
- 💬 Follow-up tooltip duplication is removed, streamlining the user interface and preventing visual clutter. [#17186](https://github.com/open-webui/open-webui/pull/17186)
- 🎨 Chat button text display is corrected by preventing clipping of descending characters and removing unnecessary capitalization. [#17191](https://github.com/open-webui/open-webui/pull/17191)
- 🧠 RAG Loop/Error with Gemma 3.1 2B Instruct is fixed by correctly unwrapping unexpected single-item list responses from models. [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/1bc9711afd2b72cd07c4e539a83783868733767c), [#17213](https://github.com/open-webui/open-webui/issues/17213)
- 🖼️ HEIC conversion failures are resolved, improving robustness of image handling. [#17225](https://github.com/open-webui/open-webui/pull/17225)
- 📦 The slim Docker image size regression has been fixed by refining the build process to correctly exclude components when USE_SLIM=true. [#16997](https://github.com/open-webui/open-webui/issues/16997), [Commit](https://github.com/open-webui/open-webui/commit/be373e9fd42ac73b0302bdb487e16dbeae178b4e), [Commit](https://github.com/open-webui/open-webui/commit/0ebe4f8f8490451ac8e85a4846f010854d9b54e5)
- 📁 Knowledge base update validation errors are resolved, ensuring seamless management via UI or API. [#17244](https://github.com/open-webui/open-webui/issues/17244), [Commit](https://github.com/open-webui/open-webui/commit/9aac1489080a5c9441e89b1a56de0d3a672bc5fb)
- 🔐 Resolved a security issue where a global web search setting overrode model-specific restrictions, ensuring model-level settings are now correctly prioritized. [#17151](https://github.com/open-webui/open-webui/issues/17151), [Commit](https://github.com/open-webui/open-webui/commit/9368d0ac751ec3072d5a96712b80a9b20a642ce6)
- 🔐 OAuth redirect reliability is improved by robustly preserving the intended redirect path using session storage. [#17235](https://github.com/open-webui/open-webui/issues/17235), [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/4f2b821088367da18374027919594365c7a3f459), [#15575](https://github.com/open-webui/open-webui/pull/15575), [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/d9f97c832c556fae4b116759da0177bf4fe619de)
- 🔐 Fixed a security vulnerability where knowledge base access within chat folders persisted after permissions were revoked. [#17182](https://github.com/open-webui/open-webui/issues/17182), [Commit](https://github.com/open-webui/open-webui/commit/40e40d1dddf9ca937e99af41c8ca038dbc93a7e6)
- 🔒 OIDC access denied errors are now displayed as user-friendly toast notifications instead of raw JSON. [#17208](https://github.com/open-webui/open-webui/issues/17208), [Commit](https://github.com/open-webui/open-webui/commit/3d6d050ad82d360adc42d6e9f42e8faf8d13c9f4)
- 💬 Chat exception handling is enhanced to prevent system instability during message generation and ensure graceful error recovery. [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/f56889c5c7f0cf1a501c05d35dfa614e4f8b6958)
- 🔒 Static asset authentication is improved by adding crossorigin="use-credentials" attributes to all link elements, enabling proper cookie forwarding for proxy environments and authenticated requests to favicon, manifest, and stylesheet resources. [#17280](https://github.com/open-webui/open-webui/pull/17280), [Commit](https://github.com/open-webui/open-webui/commit/f17d8b5d19e1a05df7d63f53e939c99772a59c1e)
### Changed
- 🛠️ Renamed "Tools" to "External Tools" across the UI for clearer distinction between built-in and external functionalities. [Commit](https://github.com/open-webui/open-webui/pull/17070/commits/0bca4e230ef276bec468889e3be036242ad11086f)
- 🛡️ Default permission validation for message regeneration and deletion actions is enhanced to provide more restrictive access controls, improving chat security and user data protection. [#17285](https://github.com/open-webui/open-webui/pull/17285)
## [0.6.26] - 2025-08-28
### Added
- 🛂 **Granular Chat Interaction Permissions**: Added fine-grained permission controls for individual chat actions including "Continue Response", "Regenerate Response", "Rate Response", and "Delete Messages". Administrators can now configure these permissions per user group or set system defaults via environment variables, providing enhanced security and governance by preventing potential system prompt leakage through response continuation and enabling precise control over user interactions with AI responses.
- 🧠 **Custom Reasoning Tags Configuration**: Added configurable reasoning tag detection for AI model responses, allowing administrators and users to customize how the system identifies and processes reasoning content. Users can now define custom reasoning tag pairs, use default tags like "think" and "reasoning", or disable reasoning detection entirely through the Advanced Parameters interface, providing enhanced control over AI thought process visibility.
- 📱 **Pull-to-Refresh Support**: Added pull-to-refresh functionality allowing user to easily refresh the interface by pulling down on the navbar area. This resolves timeout issues that occurred when temporarily switching away from the app during long AI response generations, eliminating the need to close and relaunch the PWA.
- 📁 **Configurable File Upload Processing Mode**: Added "process_in_background" query parameter to the file upload API endpoint, allowing clients to choose between asynchronous (default) and synchronous file processing. Setting "process_in_background=false" forces the upload request to wait until extraction and embedding complete, returning immediately usable files and simplifying integration for backend API consumers that prefer blocking calls over polling workflows.
- 🔐 **Azure Document Intelligence DefaultAzureCredential Support**: Added support for authenticating with Azure Document Intelligence using DefaultAzureCredential in addition to API key authentication, enabling seamless integration with Azure Entra ID and managed identity authentication for enterprise Azure environments.
- 🔐 **Authentication Bootstrapping Enhancements**: Added "ENABLE_INITIAL_ADMIN_SIGNUP" environment variable and "?form=true" URL parameter to enable initial admin user creation and forced login form display in SSO-only deployments. This resolves bootstrap issues where administrators couldn't create the first user when login forms were disabled, allowing proper initialization of SSO-configured deployments without requiring temporary configuration changes.
- ⚡ **Query Generation Caching**: Added "ENABLE_QUERIES_CACHE" environment variable to enable request-scoped caching of generated search queries. When both web search and file retrieval are active, queries generated for web search are automatically reused for file retrieval, eliminating duplicate LLM API calls and reducing token usage and costs while maintaining search quality.
- 🔧 **Configurable Tool Call Retry Limit**: Added "CHAT_RESPONSE_MAX_TOOL_CALL_RETRIES" environment variable to control the maximum number of sequential tool calls allowed before safety stopping a chat session. This replaces the previous hardcoded limit of 10, enabling administrators to configure higher limits for complex workflows requiring extensive tool interactions.
- 📦 **Slim Docker Image Variant**: Added new slim Docker image option via "USE_SLIM" build argument that excludes embedded AI models and Ollama installation, reducing image size by approximately 1GB. This variant enables faster image pulls and deployments for environments where AI models are managed externally, particularly beneficial for auto-scaling clusters and distributed deployments.
- 🗂️ **Shift-to-Delete Functionality for Workspace Prompts**: Added keyboard shortcut support for quick prompt deletion on the Workspace Prompts page. Hold Shift and hover over any prompt to reveal a trash icon for instant deletion, bringing consistent interaction patterns across all workspace sections (Models, Tools, Functions, and now Prompts) and streamlining prompt management workflows.
- ♿ **Accessibility Enhancements**: Enhanced user interface accessibility with improved keyboard navigation, ARIA labels, and screen reader compatibility across key platform components.
- 📄 **Optimized PDF Export for Smaller File Size**: PDF exports are now significantly optimized, producing much smaller files for faster downloads and easier sharing or archiving of your chats and documents.
- 📦 **Slimmed Default Install with Optional Full Dependencies**: Installing Open WebUI via pip now defaults to a slimmer package; PostgreSQL support is no longer included by default—simply use 'pip install open-webui[all]' to include all optional dependencies for full feature compatibility.
- 🔄 **General Backend Refactoring**: Implemented various backend improvements to enhance performance, stability, and security, ensuring a more resilient and reliable platform for all users.
- 🌐 **Localization & Internationalization Improvements**: Enhanced and expanded translations for Finnish, Spanish, Japanese, Polish, Portuguese (Brazil), and Chinese, including missing translations and typo corrections, providing a more natural and professional user experience for speakers of these languages across the entire interface.
### Fixed
- ⚠️ **Chat Error Feedback Restored**: Fixed an issue where various backend errors (tool server failures, API connection issues, malformed responses) would cause chats to load indefinitely without providing user feedback. The system now properly displays error messages when failures occur during chat generation, allowing users to understand issues and retry as needed instead of waiting indefinitely.
- 🖼️ **Image Generation Steps Setting Visibility Fixed**: Fixed a UI issue where the "Set Steps" configuration option was incorrectly displayed for OpenAI and Gemini image generation engines that don't support this parameter. The setting is now only visible for compatible engines like ComfyUI and Automatic1111, eliminating user confusion about non-functional configuration options.
- 📄 **Datalab Marker API Document Loader Fixed**: Fixed broken Datalab Marker API document loader functionality by correcting URL path handling for both hosted Datalab services and self-hosted Marker servers. Removed hardcoded "/marker" paths from the loader code and restored proper default URL structure, ensuring PDF and document processing works correctly with both deployment types.
- 📋 **Citation Error Handling Improved**: Fixed an issue where malformed citation or source objects from external tools, pipes, or filters would cause JavaScript errors and make the chat interface completely unresponsive. The system now gracefully handles missing or undefined citation properties, allowing conversations to load properly even when tools generate defective citation events.
- 👥 **Group User Add API Endpoint Fixed**: Fixed an issue where the "/api/v1/groups/id/{group_id}/users/add" API endpoint would accept requests without errors but fail to actually add users to groups. The system now properly initializes and deduplicates user ID lists, ensuring users are correctly added to and removed from groups via API calls.
- 🛠️ **External Tool Server Error Handling Improved**: Fixed an issue where unreachable or misconfigured external tool servers would cause JavaScript errors and prevent the interface from loading properly. The system now gracefully handles connection failures, displays appropriate error messages, and filters out inaccessible servers while maintaining full functionality for working connections.
- 📋 **Code Block Copy Button Content Fixed**: Fixed an issue where the copy button in code blocks would copy the original AI-generated code instead of any user-edited content, ensuring the copy function always captures the currently displayed code as modified by users.
- 📄 **PDF Export Content Mismatch Fixed**: Resolved an issue where exporting a PDF of one chat while viewing another chat would incorrectly generate the PDF using the currently viewed chat's content instead of the intended chat's content. Additionally optimized the PDF generation algorithm with improved canvas slicing, better memory management, and enhanced image quality, while removing the problematic PDF export option from individual chat menus to prevent further confusion.
- 🖱️ **Windows Sidebar Cursor Icon Corrected**: Fixed confusing cursor icons on Windows systems where sidebar toggle buttons displayed resize cursors (ew-resize) instead of appropriate pointer cursors. The sidebar buttons now show standard pointer cursors on Windows, eliminating user confusion about whether the buttons expand/collapse the sidebar or resize it.
- 📝 **Safari IME Composition Bug Fixed**: Resolved an issue where pressing Enter while composing Chinese text using Input Method Editors (IMEs) on Safari would prematurely send messages instead of completing text composition. The system now properly detects composition states and ignores keydown events that occur immediately after composition ends, ensuring smooth multilingual text input across all browsers.
- 🔍 **Hybrid Search Parameter Handling Fixed**: Fixed an issue where the "hybrid" parameter in collection query requests was not being properly evaluated, causing the system to ignore user-specified hybrid search preferences and only check global configuration. Additionally resolved a division by zero error that occurred in hybrid search when BM25Retriever was called with empty document lists, ensuring robust search functionality across all collection states.
- 💬 **RTL Text Orientation in Messages Fixed**: Fixed text alignment issues in user messages and AI responses for Right-to-Left languages, ensuring proper text direction based on user language settings. Code blocks now consistently use Left-to-Right orientation regardless of the user's language preference, maintaining code readability across all supported languages.
- 📁 **File Content Preview in Modal Restored**: Fixed an issue where clicking on uploaded files would display an empty preview modal, even when the files were successfully processed and available for AI context. File content now displays correctly in the preview modal, ensuring users can verify and review their uploaded documents as intended.
- 🌐 **Playwright Timeout Configuration Corrected**: Fixed an issue where Playwright timeout values were incorrectly converted from milliseconds to seconds with an additional 1000x multiplier, causing excessively long web loading timeouts. The timeout parameter now correctly uses the configured millisecond values as intended, ensuring responsive web search and document loading operations.
### Changed
- 🔄 **Follow-Up Question Language Constraint Removed**: Follow-up question suggestions no longer strictly adhere to the chat's primary language setting, allowing for more flexible and diverse suggestion generation that may include questions in different languages based on conversation context and relevance rather than enforced language matching.
## [0.6.25] - 2025-08-22
### Fixed
- 🖼️ **Image Generation Reliability Restored**: Fixed a key issue causing image generation failures.
- 🏆 **Reranking Functionality Restored**: Resolved errors with rerank feature.
## [0.6.24] - 2025-08-21
### Added
- ♿ **High Contrast Mode in Chat Messages**: Implemented enhanced High Contrast Mode support for chat messages, making text and important details easier to read and improving accessibility for users with visual preferences or requirements.
- 🌎 **Localization & Internationalization Improvements**: Enhanced and expanded translations for a more natural and professional user experience for speakers of these languages across the entire interface.
### Fixed
- 🖼️ **ComfyUI Image Generation Restored**: Fixed a critical bug where ComfyUI-based image generation was not functioning, ensuring users can once again effortlessly create and interact with AI-generated visuals in their workflows.
- 🛠️ **Tool Server Loading and Visibility Restored**: Resolved an issue where connected tool servers were not loading or visible, restoring seamless integration and uninterrupted access to all external and custom tools directly within the platform.
- 🛡️ **Redis User Session Reliability**: Fixed a problem affecting the saving of user sessions in Redis, ensuring reliable login sessions, stable authentication, and secure multi-user environments.
## [0.6.23] - 2025-08-21
### Added
- ⚡ **Asynchronous Chat Payload Processing**: Refactored the chat completion pipeline to return a response immediately for streaming requests involving web search or tool calls. This enables users to stop ongoing generations promptly and preventing network timeouts during lengthy preprocessing phases, thus significantly improving user experience and responsiveness.
- 📁 **Asynchronous File Upload with Polling**: Implemented an asynchronous file upload process with frontend polling to resolve gateway timeouts and improve reliability when uploading large files. This ensures that even lengthy file processing, such as embedding or transcription, does not block the user interface or lead to connection timeouts, providing a smoother experience for all file operations.
- 📈 **Database Performance Indexes and Migration Script**: Introduced new database indexes on the "chat", "tag", and "function" tables to significantly enhance query performance for SQLite and PostgreSQL installations. For existing deployments, a new Alembic migration script is included to seamlessly apply these indexes, ensuring faster filtering and sorting operations across the platform.
- ✨ **Enhanced Database Performance Options**: Introduced new configurable options to significantly improve database performance, especially for SQLite. This includes "DATABASE_ENABLE_SQLITE_WAL" to enable SQLite WAL (Write-Ahead Logging) mode for concurrent operations, and "DATABASE_DEDUPLICATE_INTERVAL" which, in conjunction with a new deduplication mechanism, reduces redundant updates to "user.last_active_at", minimizing write conflicts across all database types.
- 💾 **Save Temporary Chats Button**: Introduced a new 'Save Chat' button for conversations initiated in temporary mode. This allows users to permanently save valuable temporary conversations to their chat history, providing greater flexibility and ensuring important discussions are not lost.
- 📂 **Chat Movement Options in Menu**: Added the ability to move chats directly to folders from the chat menu. This enhances chat organization and allows users to manage their conversations more efficiently by relocating them between folders with ease.
- 💬 **Language-Aware Follow-Up Suggestions**: Enhanced the AI's follow-up question generation to dynamically adapt to the primary language of the current chat. Follow-up prompts will now be suggested in the same language the user and AI are conversing in, ensuring more natural and contextually relevant interactions.
- 👤 **Expanded User Profile Details**: Introduced new user profile fields including username, bio, gender, and date of birth, allowing for more comprehensive user customization and information management. This enhancement includes corresponding updates to the database schema, API, and user interface for seamless integration.
- 👥 **Direct Navigation to User Groups from User Edit**: Enhanced the user edit modal to include a direct link to the associated user group. This allows administrators to quickly navigate from a user's profile to their group settings, streamlining user and group management workflows.
- 🔧 **Enhanced External Tool Server Compatibility**: Improved handling of responses from external tool servers, allowing both the backend and frontend to process plain text content in addition to JSON, ensuring greater flexibility and integration with diverse tool outputs.
- 🗣️ **Enhanced Audio Transcription Language Fallback and Deepgram Support**: Implemented a robust language fallback mechanism for both OpenAI and Deepgram Speech-to-Text (STT) API calls. If a specified language parameter is not supported by the model or provider, the system will now intelligently retry the transcription without the language parameter or with a default, ensuring greater reliability and preventing failed API calls. This also specifically adds and refines support for the audio language parameter in Deepgram API integrations.
- ⚡ **Optimized Hybrid Search Performance for BM25 Weight Configuration**: Enhanced hybrid search to significantly improve performance when the BM25 weight is set to 0 or less. This optimization intelligently disables unnecessary collection retrieval and BM25 ranking calculations, leading to faster search results without impacting accuracy for configurations that do not utilize lexical search contributions.
- 🔒 **Configurable Code Interpreter Module Blacklist**: Introduced the "CODE_INTERPRETER_BLACKLISTED_MODULES" environment variable, allowing administrators to specify Python modules that are forbidden from being imported or executed within the code interpreter. This significantly enhances the security posture by mitigating risks associated with arbitrary code execution, such as unauthorized data access, system manipulation, or outbound connections.
- 🔐 **Enhanced OAuth Role Claim Handling**: Improved compatibility with diverse OAuth providers by allowing role claims to be supplied as single strings or integers, in addition to arrays. The system now automatically normalizes these single-value claims into arrays for consistent processing, streamlining integration with identity providers that format role data differently.
- ⚙️ **Configurable Tool Call Timeout**: Introduced the "AIOHTTP_CLIENT_TIMEOUT" environment variable, allowing administrators to specify custom timeout durations for external tool calls, which is crucial for integrations with tools that have varying or extended response times.
- 🛠️ **Improved Tool Callable Generation for Google genai SDK**: Enhanced the creation of tool callables to directly support native function calling within the Google 'genai' SDK. This refactoring ensures proper signature inference and removes extraneous parameters, enabling seamless integration for advanced AI workflows using Google's generative AI models.
- ✨ **Dynamic Loading of 'kokoro-js'**: Implemented dynamic loading for the 'kokoro-js' library, preventing failures and improving compatibility on older iOS browsers that may not support direct imports or certain modern JavaScript APIs like 'DecompressionStream'.
- 🖥️ **Improved Command List Visibility on Small Screens**: Resolved an issue where the top items in command lists (e.g., Knowledge Base, Models, Prompts) were hidden or overlapped by the header on smaller screen sizes or specific browser zoom levels. The command option lists now dynamically adjust their height, ensuring all items are fully visible and accessible with proper scrolling.
- 📦 **Improved Docker Image Compatibility for Arbitrary UIDs**: Fixed issues preventing the Open WebUI container from running in environments with arbitrary User IDs (UIDs), such as OpenShift's restricted Security Context Constraints (SCC). The Dockerfile has been updated to correctly set file system permissions for "/app" and "/root" directories, ensuring they are writable by processes running with a supplemental GID 0, thus resolving permission errors for Python libraries and application caches.
- ♿ **Accessibility Enhancements**: Significantly improved the semantic structure of chat messages by using "section", "h2", "ul", and "li" HTML tags, and enhanced screen reader compatibility by explicitly hiding decorative images with "aria-hidden" attributes. This refactoring provides clearer structural context and improves overall accessibility and web standards compliance for the conversation flow.
- 🌐 **Localization & Internationalization Improvements**: Significantly expanded internationalization support throughout the user interface, translating numerous user-facing strings in toast messages, placeholders, and other UI elements. This, alongside continuous refinement and expansion of translations for languages including Brazilian Portuguese, Kabyle (Taqbaylit), Czech, Finnish, Chinese (Simplified), Chinese (Traditional), and German, and general fixes for several other translation files, further enhances linguistic coverage and user experience.
### Fixed
- 🛡️ **Resolved Critical OIDC SSO Login Failure**: Fixed a critical issue where OIDC Single Sign-On (SSO) logins failed due to an error in setting the authentication token as a cookie during the redirect process. This ensures reliable and seamless authentication for users utilizing OIDC providers, restoring full login functionality that was impacted by previous security hardening.
- ⚡ **Prevented UI Blocking by Unreachable Webhooks**: Resolved a critical performance and user experience issue where synchronous webhook calls to unreachable or slow endpoints would block the entire user interface for all users. Webhook requests are now processed asynchronously using "aiohttp", ensuring that the UI remains responsive and functional even if webhook delivery encounters delays or failures.
- 🔒 **Password Change Option Hidden for Externally Authenticated Users**: Resolved an issue where the password change dialog was visible to users authenticated via external methods (e.g., LDAP, OIDC, Trusted Header). The option to change a password in user settings is now correctly hidden for these users, as their passwords are managed externally, streamlining the user interface and preventing confusion.
- 💬 **Resolved Temporary Chat and Permission Enforcement Issues**: Fixed a bug where temporary chats (identified by "chat_id = local") incorrectly triggered database checks, leading to 404 errors. This also resolves the issue where the 'USER_PERMISSIONS_CHAT_TEMPORARY_ENFORCED' setting was not functioning as intended, ensuring temporary chat mode now works correctly for user roles.
- 🔐 **Admin Model Visibility for Administrators**: Private models remained visible and usable for administrators in the chat model selector, even when the intended privacy setting ("ENABLE_ADMIN_WORKSPACE_CONTENT_ACCESS" - now renamed to "BYPASS_ADMIN_ACCESS_CONTROL") was disabled. This ensures consistent enforcement of model access controls and adherence to the principle of least privilege.
- 🔍 **Clarified Web Search Engine Label for DDGS**: Addressed user confusion and inaccurate labeling by renaming "duckduckgo" to "DDGS" (Dux Distributed Global Search) in the web search engine selector. This clarifies that the system utilizes DDGS, a metasearch library that aggregates results from various search providers, accurately reflecting its underlying functionality rather than implying exclusive use of DuckDuckGo's search engine.
- 🛠️ **Improved Settings UI Reactivity and Visibility**: Resolved an issue where settings tabs for 'Connections' and 'Tools' did not dynamically update their visibility based on global administrative feature flags (e.g., 'enable_direct_connections'). The UI now reactively shows or hides these sections, ensuring a consistent and clear experience when administrators control feature availability.
- 🎚️ **Restored Model and Banner Reordering Functionality**: Fixed a bug that prevented administrators from reordering models in the Admin Panel's 'Models' settings and banners in the 'Interface' settings via drag-and-drop. The sortable functionality has been restored, allowing for proper customization of display order.
- 📝 **Restored Custom Pending User Overlay Visibility**: Fixed an issue where the custom title and description configured for pending users were not visible. The application now correctly exposes these UI configuration settings to pending users, ensuring that the custom onboarding messages are displayed as intended.
- 📥 **Fixed Community Function Import Compatibility**: Resolved an issue that prevented the successful import of function files downloaded from openwebui.com due to schema differences. The system now correctly processes these files, allowing for seamless integration of community-contributed functions.
- 📦 **Fixed Stale Ollama Version in Docker Images**: Resolved an issue where the Ollama installation within Docker images could become stale due to caching during the build process. The Dockerfile now includes a mechanism to invalidate the build cache for the Ollama installation step, ensuring that the latest version of Ollama is always installed.
- 🗄️ **Improved Milvus Query Handling for Large Datasets**: Fixed a "MilvusException" that occurred when attempting to query more than 16384 entries from a Milvus collection. The query logic has been refactored to use "query_iterator()", enabling efficient fetching of larger result sets in batches and resolving the previous limitation on the number of entries that could be retrieved.
- 🐛 **Restored Message Toolbar Icons for Empty Messages with Files**: Fixed an issue where the edit, copy, and delete icons were not displayed on user messages that contained an attached file but no text content. This ensures full interaction capabilities for all message types, allowing users to manage their messages consistently.
- 💬 **Resolved Streaming Interruption for Kimi-Dev Models**: Fixed an issue where streaming responses from Kimi-Dev models would halt prematurely upon encountering specific 'thinking' tokens (◁think▷, ◁/think▷). The system now correctly processes these tokens, ensuring uninterrupted streaming and proper handling of hidden or collapsible thinking sections.
- 🔍 **Enhanced Knowledge Base Search Functionality**: Improved the search capability within the 'Knowledge' section of the Workspace. Previously, searching for knowledge bases required exact term matches or starting with the first letter. Now, the search algorithm has been refined to allow broader, less exact matches, making it easier and more intuitive to find relevant knowledge bases.
- 📝 **Resolved Chinese Input 'Enter' Key Issue (macOS & iOS Safari)**: Fixed a bug where pressing the 'Enter' key during text composition with Input Method Editors (IMEs) on macOS and iOS Safari browsers would prematurely send the message. The system now robustly handles the composition state by addressing a 'compositionend' event bug specific to Safari, ensuring a smooth and expected typing experience for users of various languages, including Chinese and Korean.
- 🔐 **Resolved OAUTH_GROUPS_CLAIM Configuration Issue**: Fixed a bug where the "OAUTH_GROUPS_CLAIM" environment variable was not correctly parsed due to a typo in the configuration file. This ensures that OAuth group management features, including automatic group creation, now correctly utilize the specified claim from the identity provider, allowing for seamless integration with external user directories like Keycloak.
- 🗄️ **Resolved Azure PostgreSQL pgvector Extension Permissions**: Fixed an issue preventing the creation of "pgvector" and "pgcrypto" extensions on Azure PostgreSQL Flexible Servers due to permission limitations (e.g., 'Only members of "azure_pg_admin" are allowed to use "CREATE EXTENSION"'). The extension creation process now includes a conditional check, ensuring seamless deployment and compatibility with Azure PostgreSQL environments even with restricted database user permissions.
- 🛠️ **Improved Backend Path Resolution and Alembic Stability**: Fixed issues causing Alembic database migrations to fail due to incorrect path resolution within the application. By implementing canonical path resolution for core directories and refining Alembic configuration, the robustness and correctness of internal pathing have been significantly enhanced, ensuring reliable database operations.
- 📊 **Resolved Arena Model Identification in Feedback History**: Fixed an issue where the model used for feedback in arena settings was incorrectly reported as 'arena-model' in the evaluation history. The system now correctly logs and displays the actual model ID that received the feedback, restoring clarity and enabling proper analysis of model performance in arena environments.
- 🎨 **Resolved Icon Overlap in 'Her' Theme**: Fixed a visual glitch in the 'Her' theme where icons would overlap on the loading screen and certain icons appeared incongruous. The display has been corrected to ensure proper visual presentation and theme consistency.
- 🛠️ **Resolved Model Sorting TypeError with Null Names**: Fixed a "TypeError" that occurred in the "/api/models" endpoint when sorting models with null or missing names. The model sorting logic has been improved to gracefully handle such edge cases by ensuring that model IDs and names are treated as empty strings if their values are null or undefined, preventing comparison errors and improving API stability.
- 💬 **Resolved Silently Dropped Streaming Response Chunks**: Fixed an issue where the final partial chunks of streaming chat responses could be silently dropped, leading to incomplete message delivery. The system now reliably flush any pending delta data upon stream termination, early breaks (e.g., code interpreter tags), or connection closure, ensuring complete and accurate response delivery.
- 📱 **Disabled Overscroll for iOS Frontend**: Fixed an issue where overscrolling was enabled on iOS devices, causing unexpected scrolling behavior over fixed or sticky elements within the PWA. Overscroll has now been disabled, providing a more native application-like experience for iOS users.
- 📝 **Resolved Code Block Input Issue with Shift+Enter**: Fixed a bug where typing three backticks followed by a language and then pressing Shift+Enter would cause the code block prefix to disappear, preventing proper code formatting. The system now correctly preserves the code block syntax, ensuring consistent behavior for multi-line code input.
- 🛠️ **Improved OpenAI Model List Handling for Null Names**: Fixed an edge case where some OpenAI-compatible API providers might return models with a null value for their 'name' field. This could lead to issues like broken model list sorting. The system now gracefully handles these instances by removing the null 'name' key, ensuring stable model retrieval and display.
- 🔍 **Resolved DDGS Concurrent Request Configuration**: Fixed an issue where the configured number of concurrent requests was not being honored for the DDGS (Dux Distributed Global Search) metasearch engine. The system now correctly applies the specified concurrency setting, improving efficiency for web searches.
- 🛠️ **Improved Tool List Synchronization in Multi-Replica Deployments**: Resolved an issue where tool updates were not consistently reflected across all instances in multi-replica environments, leading to stale tool lists for users on other replicas. The tool list in the message input menu is now automatically refreshed each time it is accessed, ensuring all users always see the most current set of available tools.
- 🛠️ **Resolved Duplicate Tool Name Collision**: Fixed an issue where tools with identical names from different external servers were silently removed, preventing their simultaneous use. The system now correctly handles tool name collisions by internally prefixing tools with their server identifier, allowing multiple instances of similarly named tools from different servers to be active and usable by LLMs.
- 🖼️ **Resolved Image Generation API Size Parameter Issue**: Fixed a bug where the "/api/v1/images/generations" API endpoint did not correctly apply the 'size' parameter specified in the request payload for image generation. The system now properly honors the requested image dimensions (e.g., '1980x1080'), ensuring that generated images match the user's explicit size preference rather than defaulting to settings.
- 🗄️ **Resolved S3 Vector Upload Limitations**: Fixed an issue that prevented uploading more than 500 vectors to S3 Vector buckets due to API limitations, which resulted in a "ValidationException". S3 vector uploads are now batched in groups of 500, ensuring successful processing of larger datasets.
- 🛠️ **Fixed Tool Installation Error During Startup**: Resolved a "NoneType" error that occurred during tool installation at startup when 'tool.user' was unexpectedly null. The system now includes a check to ensure 'tool.user' exists before attempting to access its properties, preventing crashes and ensuring robust tool initialization.
- 🛠️ **Improved Azure OpenAI GPT-5 Parameter Handling**: Fixed an issue with Azure OpenAI SDK parameter handling to correctly support GPT-5 models. The 'max_tokens' parameter is now appropriately converted to 'max_completion_tokens' for GPT-5 models, ensuring consistent behavior and proper function execution similar to existing o-series models.
- 🐛 **Resolved Exception with Missing Group Permissions**: Fixed an exception that occurred in the access control logic when group permission objects were missing or null. The system now correctly handles cases where groups may not have explicit permission definitions, ensuring that 'None' checks prevent errors and maintain application stability when processing user permissions.
- 🛠️ **Improved OpenAI API Base URL Handling**: Fixed an issue where a trailing slash in the 'OPENAI_API_BASE_URL' configuration could lead to models not being detected or the endpoint failing. The system now automatically removes trailing slashes from the configured URL, ensuring robust and consistent connections to OpenAI-compatible APIs.
- 🖼️ **Resolved S3-Compatible Storage Upload Failures**: Fixed an issue where uploads to S3-compatible storage providers would fail with an "XAmzContentSHA256Mismatch" error. The system now correctly handles checksum calculations, ensuring reliable file and image uploads to S3-compatible services.
- 🌐 **Corrected 'Releases' Link**: Fixed an issue where the 'Releases' button in the user menu directed to an incorrect URL, now correctly linking to the Open WebUI GitHub releases page.
- 🛠️ **Resolved Model Sorting Errors with Null or Undefined Names**: Fixed multiple "TypeError" instances that occurred when attempting to sort model lists where model names were null or undefined. The sorting logic across various UI components (including Ollama model selection, leaderboard, and admin model settings) has been made more robust by gracefully handling absent model names, preventing crashes and ensuring consistent alphabetical sorting based on available name or ID.
- 🎨 **Resolved Banner Dismissal Issue with Iteration IDs**: Fixed a bug where dismissing banners could lead to unintended multiple banner dismissals or other incorrect behavior, especially when banners lacked unique iteration IDs. Unique IDs are now assigned during banner iteration, ensuring proper individual dismissal and consistent display behavior.
### Changed
- 🛂 **Environment Variable for Admin Access Control**: The environment variable "ENABLE_ADMIN_WORKSPACE_CONTENT_ACCESS" has been renamed to "BYPASS_ADMIN_ACCESS_CONTROL". This new name more accurately reflects its function as a control to allow administrators to bypass model access restrictions. Users are encouraged to update their configurations to use the new variable name; existing configurations using the old name will still be honored for backward compatibility.
- 🗂️ **Core Directory Path Resolution Updated**: The internal mechanism for resolving core application directory paths ("OPEN_WEBUI_DIR", "BACKEND_DIR", "BASE_DIR") has been updated to use canonical resolution via "Path().resolve()". This change improves path reliability but may require adjustments for any external scripts or configurations that previously relied on specific non-canonical path interpretations.
- 🗃️ **Database Performance Options**: New database performance options, "DATABASE_ENABLE_SQLITE_WAL" and "DATABASE_DEDUPLICATE_INTERVAL", are now available. If "DATABASE_ENABLE_SQLITE_WAL" is enabled, SQLite will operate in WAL mode, which may alter SQLite's file locking behavior. If "DATABASE_DEDUPLICATE_INTERVAL" is set to a non-zero value, the "user.last_active_at" timestamp will be updated less frequently, leading to slightly less real-time accuracy for this specific field but significantly reducing database write conflicts and improving overall performance. Both options are disabled by default.
- 🌐 **Renamed Web Search Concurrency Setting**: The environment variable "WEB_SEARCH_CONCURRENT_REQUESTS" has been renamed to "WEB_LOADER_CONCURRENT_REQUESTS". This change clarifies its scope, explicitly applying to the concurrency of the web loader component (which fetches content from search results) rather than the initial search engine query. Users relying on the old environment variable name for configuring web search concurrency must update their configurations to use "WEB_LOADER_CONCURRENT_REQUESTS".
## [0.6.22] - 2025-08-11
### Added
- 🔗 **OpenAI API '/v1' Endpoint Compatibility**: Enhanced API compatibility by supporting requests to paths like '/v1/models', '/v1/embeddings', and '/v1/chat/completions'. This allows Open WebUI to integrate more seamlessly with tools that expect OpenAI's '/v1' API structure.
- 🪄 **Toggle for Guided Response Regeneration Menu**: Introduced a new setting in 'Interface' settings, providing the ability to enable or disable the expanded guided response regeneration menu. This offers users more control over their chat workflow and interface preferences.
- ✨ **General UI/UX Enhancements**: Implemented various user interface and experience improvements, including more rounded corners for cards in the Knowledge, Prompts, and Tools sections, and minor layout adjustments within the chat Navbar for improved visual consistency.
- 🌐 **Localization & Internationalization Improvements**: Introduced support for the Kabyle (Taqbaylit) language, refined and expanded translations for Chinese, expanding the platform's linguistic coverage.
### Fixed
- 🐞 **OpenAI Error Message Propagation**: Resolved an issue where specific OpenAI API errors (e.g., 'Organization Not Verified') were obscured by generic 'JSONResponse' iterable errors. The system now correctly propagates detailed and actionable error messages from OpenAI to the user.
- 🌲 **Pinecone Insert Issue**: Fixed a bug that prevented proper insertion of items into Pinecone vector databases.
- 📦 **S3 Vector Issue**: Resolved a bug where s3vector functionality failed due to incorrect import paths.
- 🏠 **Landing Page Option Setting Not Working**: Fixed an issue where the landing page option in settings was not functioning as intended.
## [0.6.21] - 2025-08-10
### Added
- 👥 **User Groups in Edit Modal**: Added display of user groups information in the user edit modal, allowing administrators to view and manage group memberships directly when editing a user.
### Fixed
- 🐞 **Chat Completion 'model_id' Error**: Resolved a critical issue where chat completions failed with an "undefined model_id" error after upgrading to version 0.6.20, ensuring all models now function correctly and reliably.
- 🛠️ **Audit Log User Information Logging**: Fixed an issue where user information was not being correctly logged in the audit trail due to an unreflected function prototype change, ensuring complete logging for administrative oversight.
- 🛠️ **OpenTelemetry Configuration Consistency**: Fixed an issue where OpenTelemetry metric and log exporters' 'insecure' settings did not correctly default to the general OpenTelemetry 'insecure' flag, ensuring consistent security configurations across all OpenTelemetry exports.
- 📝 **Reply Input Content Display**: Fixed an issue where replying to a message incorrectly displayed '{{INPUT_CONTENT}}' instead of the actual message content, ensuring proper content display in replies.
- 🌐 **Localization & Internationalization Improvements**: Refined and expanded translations for Catalan, Korean, Spanish and Irish, ensuring a more fluent and native experience for global users.
## [0.6.20] - 2025-08-10
### Fixed
- 🛠️ **Quick Actions "Add" Behavior**: Fixed a bug where using the "Add" button in Quick Actions would add the resulting message as the very first message in the chat, instead of appending it to the latest message.
## [0.6.19] - 2025-08-09
### Added
- ✨ **Modernized Sidebar and Major UI Refinements**: The main navigation sidebar has been completely redesigned with a modern, cleaner aesthetic, featuring a sticky header and footer to keep key controls accessible. Core sidebar logic, like the pinned models list, was also refactored into dedicated components for better performance and maintainability.
- 🪄 **Guided Response Regeneration**: The "Regenerate" button has been transformed into a powerful new menu. You can now guide the AI's next attempt by suggesting changes in a text prompt, or use one-click options like "Try Again," "Add Details," or "More Concise" to instantly refine and reshape the response to better fit your needs.
- 🛠️ **Improved Tool Call Handling for GPT-OSS Models**: Implemented robust handling for tool calls specifically for GPT-OSS models, ensuring proper function execution and integration.
- 🛑 **Stop Button for Merge Responses**: Added a dedicated stop button to immediately halt the generation of merged AI responses, providing users with more control over ongoing outputs.
- 🔄 **Experimental SCIM 2.0 Support**: Implemented SCIM 2.0 (System for Cross-domain Identity Management) protocol support, enabling enterprise-grade automated user and group provisioning from identity providers like Okta, Azure AD, and Google Workspace for seamless user lifecycle management. Configuration is managed securely via environment variables.
- 🗂️ **Amazon S3 Vector Support**: You can now use Amazon S3 Vector as a high-performance vector database for your Retrieval-Augmented Generation (RAG) workflows. This provides a scalable, cloud-native storage option for users deeply integrated into the AWS ecosystem, simplifying infrastructure and enabling enterprise-scale knowledge management.
- 🗄️ **Oracle 23ai Vector Search Support**: Added support for Oracle 23ai's new vector search capabilities as a supported vector database, providing a robust and scalable option for managing large-scale documents and integrating vector search with existing business data at the database level.
- ⚡ **Qdrant Performance and Configuration Enhancements**: The Qdrant client has been significantly improved with faster data retrieval logic for 'get' and 'query' operations. New environment variables ('QDRANT_TIMEOUT', 'QDRANT_HNSW_M') provide administrators with finer control over query timeouts and HNSW index parameters, enabling better performance tuning for large-scale deployments.
- 🔐 **Encrypted SQLite Database with SQLCipher**: You can now encrypt your entire SQLite database at rest using SQLCipher. By setting the 'DATABASE_TYPE' to 'sqlite+sqlcipher' and providing a 'DATABASE_PASSWORD', all data is transparently encrypted, providing an essential security layer for protecting sensitive information in self-hosted deployments. Note that this requires additional system libraries and the 'sqlcipher3-wheels' Python package.
- 🚀 **Efficient Redis Connection Management**: Implemented a shared connection pool cache to reuse Redis connections, dramatically reducing the number of active clients. This prevents connection exhaustion errors, improves performance, and ensures greater stability in high-concurrency deployments and those using Redis Sentinel.
- ⚡ **Batched Response Streaming for High Performance**: Dramatically improve performance and stability during high-speed response streaming by batching multiple tokens together before sending them to the client. A new 'Stream Delta Chunk Size' advanced parameter can be set per-model or in user/chat settings, significantly reducing CPU load on the server, Redis, and client, and preventing connection issues in high-concurrency environments.
- ⚙️ **Global Batched Streaming Configuration**: Administrators can now set a system-wide default for response streaming using the new 'CHAT_RESPONSE_STREAM_DELTA_CHUNK_SIZE' environment variable. This allows for global performance tuning, while still letting per-model or per-chat settings override the default for more granular control.
- 🔎 **Advanced Chat Search with Status Filters**: Quickly find any conversation with powerful new search filters. You can now instantly narrow down your chats using prefixes like 'pinned:true', 'shared:true', and 'archived:true' directly in the search bar. An intelligent dropdown menu assists you by suggesting available filter options as you type, streamlining your workflow and making chat management more efficient than ever.
- 🛂 **Granular Chat Controls Permissions**: Administrators can now manage chat settings with greater detail. The main "Chat Controls" permission now acts as a master switch, while new granular toggles for "Valves", "System Prompts", and "Advanced Parameters" allow for more specific control over which sections are visible to users inside the panel.
- ✍️ **Formatting Toolbar for Chat Input**: Introduced a dedicated formatting toolbar for the rich text chat input field, providing users with more accessible options for text styling and editing, configurable via interface settings.
- 📑 **Tabbed View for Multi-Model Responses**: You can now enable a new tabbed interface to view responses from multiple models. Instead of side-scrolling cards, this compact view organizes each model's response into its own tab, making it easier to compare outputs and saving vertical space. This feature can be toggled on or off in Interface settings.
- ↕️ **Reorder Pinned Models via Drag-and-Drop**: You can now organize your pinned models in the sidebar by simply dragging and dropping them into your preferred order. This custom layout is saved automatically, giving you more flexible control over your workspace.
- 📌 **Quick Model Unpin Shortcut**: You can now quickly unpin a model by holding the Shift key and hovering over it to reveal an instant unpin button, streamlining your workspace customization.
- ⚡ **Improved Chat Input Performance**: The chat input is now significantly more responsive, especially when pasting or typing large amounts of text. This was achieved by implementing a debounce mechanism for the auto-save feature, which prevents UI lag and ensures a smooth, uninterrupted typing experience.
- ✍️ **Customizable Floating Quick Actions with Tool Support**: Take full control of your text interaction workflow with new customizable floating quick actions. In Settings, you can create, edit, or disable these actions and even integrate tools using the '{{TOOL:tool_id}}' syntax in your prompts, enabling powerful one-click automations on selected text. This is in addition to using placeholders like '{{CONTENT}}' and '{{INPUT_CONTENT}}' for custom text transformations.
- 🔒 **Admin Workspace Privacy Control**: Introduced the 'ENABLE_ADMIN_WORKSPACE_CONTENT_ACCESS' environment variable (defaults to 'True') allowing administrators to control their access privileges to workspace items (Knowledge, Models, Prompts, Tools). When disabled, administrators adhere to the same access control rules as regular users, enhancing data separation for multi-tenant deployments.
- 🗄️ **Comprehensive Model Configuration Management**: Administrators can now export the entire model configuration to a file and use a new declarative sync endpoint to manage models in bulk. This powerful feature enables seamless backups, migrations, and state replication across multiple instances.
- 📦 **Native Redis Cluster Mode Support**: Added full support for connecting to Redis in cluster mode, allowing for scalable and highly available Redis deployments beyond Sentinel-managed setups. New environment variables 'REDIS_CLUSTER' and 'WEBSOCKET_REDIS_CLUSTER' enable the use of 'redis.cluster.RedisCluster' clients.
- 📊 **Granular OpenTelemetry Metrics Configuration**: Introduced dedicated environment variables and enhanced configuration options for OpenTelemetry metrics, allowing for separate OTLP endpoints, basic authentication credentials, and protocol (HTTP/gRPC) specifically for metrics export, independent of trace settings. This provides greater flexibility for integrating with diverse observability stacks.
- 🪵 **Granular OpenTelemetry Logging Configuration**: Enhanced the OpenTelemetry logging integration by introducing dedicated environment variables for logs, allowing separate OTLP endpoints, basic authentication credentials, and protocol (HTTP/gRPC) specifically for log export, independent of general OTel settings. The application's default Python logger now leverages this configuration to automatically send logs to your OTel endpoint when enabled via 'ENABLE_OTEL_LOGS'.
- 📁 **Enhanced Folder Chat Management with Sorting and Time Blocks**: The chat list within folders now supports comprehensive sorting options by title and updated time, along with intelligent time-based grouping (e.g., "Today," "Yesterday") similar to the main chat view, making navigation and organization of project-specific conversations significantly easier.
- ⚙️ **Configurable Datalab Marker API & Advanced Processing Options**: Enhanced Datalab Marker API integration, allowing administrators to configure custom API base URLs for self-hosting and to specify comprehensive processing options via a new 'additional_config' JSON parameter. This replaces the deprecated language selection feature and provides granular control over document extraction, with streamlined API endpoint resolution for more robust self-hosted deployments.
- 🧑‍💼 **Export All Users to CSV**: Administrators can now export a complete list of all users to a CSV file directly from the Admin Panel's database settings. This provides a simple, one-click way to generate user data for auditing, reporting, or management purposes.
- 🛂 **Customizable OAuth 'sub' Claim**: Administrators can now use the 'OAUTH_SUB_CLAIM_OVERRIDE' environment variable to specify which claim from the identity provider should be used as the unique user identifier ('sub'). This provides greater flexibility and control for complex enterprise authentication setups where modifying the IDP's default claims is not possible.
- 👁️ **Password Visibility Toggle for Input Fields**: Password fields across the application (login, registration, user management, and account settings) now utilize a new 'SensitiveInput' component, providing a consistent toggle to reveal/hide passwords for improved usability and security.
- 🛂 **Optional "Confirm Password" on Sign-Up**: To help prevent password typos during account creation, administrators can now enable a "Confirm Password" field on the sign-up page. This feature is disabled by default and can be activated via an environment variable for enhanced user experience.
- 💬 **View Full Chat from User Feedback**: Administrators can now easily navigate to the full conversation associated with a user feedback entry directly from the feedback modal, streamlining the review and troubleshooting process.
- 🎚️ **Intuitive Hybrid Search BM25-Weight Slider**: The numerical input for the BM25-Weight parameter in Hybrid Search has been replaced with an interactive slider, offering a more intuitive way to adjust the balance between lexical and semantic search. A "Default/Custom" toggle and clearer labels enhance usability and understanding of this key parameter.
- ⚙️ **Enhanced Bulk Function Synchronization**: The API endpoint for synchronizing functions has been significantly improved to reliably handle bulk updates. This ensures that importing and managing large libraries of functions is more robust and error-free for administrators.
- 🖼️ **Option to Disable Image Compression in Channels**: Introduced a new setting under Interface options to allow users to force-disable image compression specifically for images posted in channels, ensuring higher resolution for critical visual content.
- 🔗 **Custom CORS Scheme Support**: Introduced a new environment variable 'CORS_ALLOW_CUSTOM_SCHEME' that allows administrators to define custom URL schemes (e.g., 'app://') for CORS origins, enabling greater flexibility for local development or desktop client integrations.
- ♿ **Translatable and Accessible Banners**: Enhanced banner elements with translatable badge text and proper ARIA attributes (aria-label, aria-hidden) for SVG icons, significantly improving accessibility and screen reader compatibility.
- ⚠️ **OAuth Configuration Warning for Missing OPENID_PROVIDER_URL**: Added a proactive startup warning that notifies administrators when OAuth providers (Google, Microsoft, or GitHub) are configured but the essential 'OPENID_PROVIDER_URL' environment variable is missing. This prevents silent OAuth logout failures and guides administrators to complete their setup correctly.
- ♿ **Major Accessibility Enhancements**: Key parts of the interface have been made significantly more accessible. The user profile menu is now fully navigable via keyboard, essential controls in the Playground now include proper ARIA labels for screen readers, and decorative images have been hidden from assistive technologies to reduce audio clutter. Menu buttons also feature enhanced accessibility with 'aria-label', 'aria-hidden' for SVGs, and 'aria-pressed' for toggle buttons.
- ⚙️ **General Backend Refactoring**: Implemented various backend improvements to enhance performance, stability, and security, ensuring a more resilient and reliable platform for all users, including refining logging output to be cleaner and more efficient by conditionally including 'extra_json' fields and improving consistent metadata handling in vector database operations, and laying preliminary scaffolding for future analytics features.
- 🌐 **Localization & Internationalization Improvements**: Refined and expanded translations for Catalan, Danish, Korean, Persian, Polish, Simplified Chinese, and Spanish, ensuring a more fluent and native experience for global users across all supported languages.
### Fixed
- 🛡️ **Hardened Channel Message Security**: Fixed a key permission flaw that allowed users with channel access to edit or delete messages belonging to others. The system now correctly enforces that users can only modify their own messages, protecting data integrity in shared channels.
- 🛡️ **Hardened OAuth Security by Removing JWT from URL**: Fixed a critical security vulnerability where the authentication token was exposed in the URL after a successful OAuth login. The token is now transferred via a browser cookie, preventing potential leaks through browser history or server logs and protecting user sessions.
- 🛡️ **Hardened Chat Completion API Security**: The chat completion API endpoint now includes an explicit ownership check, ensuring non-admin users cannot access chats that do not belong to them and preventing potential unauthorized access.
- 🛠️ **Resilient Model Loading**: Fixed an issue where a failure in loading the model list (e.g., from a misconfigured provider) would prevent the entire user interface, including the admin panel, from loading. The application now gracefully handles these errors, ensuring the UI remains accessible.
- 🔒 **Resolved FIPS Self-Test Failure**: Fixed a critical issue that prevented Open WebUI from running on FIPS-compliant systems, specifically resolving the "FATAL FIPS SELFTEST FAILURE" error related to OpenSSL and SentenceTransformers, restoring compatibility with secure environments.
- 📦 **Redis Cluster Connection Restored**: Fixed an issue where the backend was unable to connect to Redis in cluster mode, now ensuring seamless integration with scalable Redis cluster deployments.
- 📦 **PGVector Connection Stability**: Fixed an issue where read-only operations could leave database transactions idle, preventing potential connection errors and improving overall database stability and resource management.
- 🛠️ **OpenAPI Tool Integration for Array Parameters Fixed**: Resolved a critical bug where external tools using array parameters (e.g., for tags) would fail when used with OpenAI models. The system now correctly generates the required 'items' property in the function schema, restoring functionality and preventing '400 Bad Request' errors.
- 🛠️ **Tool Creation for Users Restored**: Fixed a bug in the code editor where status messages were incorrectly prepended to tool scripts, causing a syntax error upon saving. All authorized users can now reliably create and save new tools.
- 📁 **Folder Knowledge Processing Restored**: Fixed a bug where files uploaded to folder and model knowledge bases were not being extracted or analyzed for Retrieval-Augmented Generation (RAG) when the 'Max Upload Count' setting was empty, ensuring seamless document processing and knowledge augmentation.
- 🧠 **Custom Model Knowledge Base Updates Recognized**: Fixed a bug where custom models linked to to knowledge bases did not automatically recognize newly added files to those knowledge bases. Models now correctly incorporate the latest information from updated knowledge collections.
- 📦 **Comprehensive Redis Key Prefixing**: Corrected hardcoded prefixes to ensure the REDIS_KEY_PREFIX is now respected across all WebSocket and task management keys. This prevents data collisions in multi-instance deployments and improves compatibility with Redis cluster mode.
- ✨ **More Descriptive OpenAI Router Errors**: The OpenAI-compatible API router now propagates detailed upstream error messages instead of returning a generic 'Bad Request'. This provides clear, actionable feedback for developers and API users, making it significantly easier to debug and resolve issues with model requests.
- 🔐 **Hardened OIDC Signout Flow**: The OpenID Connect signout process now verifies that the 'OPENID_PROVIDER_URL' is configured before attempting to communicate with it, preventing potential errors and ensuring a more reliable logout experience.
- 🍓 **Raspberry Pi Compatibility Restored**: Pinned the pyarrow library to version 20.0.0, resolving an "Illegal Instruction" crash on ARM-based devices like the Raspberry Pi and ensuring stable operation on this hardware.
- 📁 **Folder System Prompt Variables Restored**: Fixed a bug where prompt variables (e.g., '{{CURRENT_DATETIME}}') were not being rendered in Folder-level System Prompts. This restores an important capability for creating dynamic, context-aware instructions for all chats within a project folder.
- 📝 **Note Access in Knowledge Retrieval Fixed**: Corrected a permission oversight in knowledge retrieval, ensuring users can always use their own notes as a source for RAG without needing explicit sharing permissions.
- 🤖 **Title Generation Compatibility for GPT-5 Models**: Added support for 'gpt-5' models in the payload handler, which correctly converts the deprecated 'max_tokens' parameter to 'max_completion_tokens'. This resolves title generation failures and ensures seamless operation with the latest generation of models.
- ⚙️ **Correct API 'finish_reason' in Streaming Responses**: Fixed an issue where intermediate 'reasoning_content' chunks in streaming API responses incorrectly reported a 'finish_reason' of 'stop'. The 'finish_reason' is now correctly set to 'null' for these chunks, ensuring compatibility with third-party applications that rely on this field.
- 📈 **Evaluation Pages Stability**: Resolved a crash on the Leaderboard and Feedbacks pages when processing legacy feedback entries that were missing a 'rating' field. The system now gracefully handles this older data, ensuring both pages load reliably for all users.
- 🤝 **Reliable Collaborative Session Cleanup**: Fixed an asynchronous bug in the real-time collaboration engine that prevented document sessions from being properly cleaned up after all users had left. This ensures greater stability and resource management for features like Collaborative Notes.
- 🧠 **Enhanced Memory Stability and Security**: Refactored memory update and delete operations to strictly enforce user ownership, preventing potential data integrity issues. Additionally, improved error handling for memory queries now provides clearer feedback when no memories exists.
- 🧑‍⚖️ **Restored Admin Access to User Feedback**: Fixed a permission issue that blocked administrators from viewing or editing user feedback they didn't create, ensuring they can properly manage all evaluations across the platform.
- 🔐 **PGVector Encryption Fix for Metadata**: Corrected a SQL syntax error in the experimental 'PGVECTOR_PGCRYPTO' feature that prevented encrypted metadata from being saved. Document uploads to encrypted PGVector collections now work as intended.
- 🔍 **Serply Web Search Integration Restored**: Fixed an issue where incorrect parameters were passed to the Serply web search engine, restoring its functionality for RAG and web search workflows.
- 🔍 **Resilient Web Search Processing**: Web search retrieval now gracefully handles search results that are missing a 'snippet', preventing crashes and ensuring that RAG workflows complete successfully even with incomplete data from search engines.
- 🖼️ **Table Pasting in Rich Text Input Displayed Correctly**: Fixed an issue where pasting table text into the rich text input would incorrectly display it as code. Tables are now properly rendered as expected, improving content formatting and user experience.
- ✍️ **Rich Text Input TypeError Resolution**: Addressed a potential 'TypeError: ue.getWordAtDocPos is not a function' in 'MessageInput.svelte' by refactoring how the 'getWordAtDocPos' function is accessed and referenced from 'RichTextInput.svelte', ensuring stable rich text input behavior, especially after production restarts.
- ✏️ **Manual Code Block Creation in Chat Restored**: Fixed an issue where typing three backticks and then pressing Shift+Enter would incorrectly remove the backticks when "Enter to Send" mode was active. This ensures users can reliably create multi-line code blocks manually.
- 🎨 **Consistent Dark Mode Background**: Fixed an issue where the application background could incorrectly flash or remain white during page loads and refreshes in dark mode, ensuring a seamless and consistent visual experience.
- 🎨 **'Her' Theme Rendering Fixed**: Corrected a bug that caused the "Her" theme to incorrectly render as a dark theme in some situations. The theme now reliably applies its intended light appearance across all sessions.
- 📜 **Corrected Markdown Table Line Break Rendering**: Fixed an issue where line breaks ('<br>') within Markdown tables were displayed as raw HTML instead of being rendered correctly. This ensures that tables with multi-line cell content are now displayed as intended.
- 🚦 **Corrected App Configuration for Pending Users**: Fixed an issue where users awaiting approval could incorrectly load the full application interface, leading to a confusing or broken UI. This ensures that only fully approved users receive the standard app 'config', resulting in a smoother and more reliable onboarding experience.
- 🔄 **Chat Cloning Now Includes Tags, Folder Status, and Pinned Status**: When cloning a chat or shared chat, its associated tags, folder organization, and pinned status are now correctly replicated, ensuring consistent chat management.
- ⚙️ **Enhanced Backend Reliability**: Resolved a potential crash in knowledge base retrieval when referencing a deleted note. Additionally, chat processing was refactored to ensure model information is saved more reliably, enhancing overall system stability.
- ⚙️ **Floating 'Ask/Explain' Modal Stability**: Fixed an issue that spammed the console with errors when navigating away while a model was generating a response in the floating 'Ask' or 'Explain' modals. In-flight requests are now properly cancelled, improving application stability.
- ⚡ **Optimized User Count Checks**: Improved performance for user count and existence checks across the application by replacing resource-intensive 'COUNT' queries with more efficient 'EXISTS' queries, reducing database load.
- 🔐 **Hardened OpenTelemetry Exporter Configuration**: The OTLP HTTP exporter no longer uses a potentially insecure explicit flag, improving security by relying on the connection URL's protocol (HTTP/HTTPS) to ensure transport safety.
- 📱 **Mobile User Menu Closing Behavior Fixed**: Resolved an issue where the user menu would remain open on mobile devices after selecting an option, ensuring the menu correctly closes and returns focus to the main interface for a smoother mobile experience.
- 📱 **OnBoarding Page Display Fixed on Mobile**: Resolved an issue where buttons on the OnBoarding page were not consistently visible on certain mobile browsers, ensuring a functional and complete user experience across devices.
- ↕️ **Improved Pinned Models Drag-and-Drop Behavior**: The drag-and-drop functionality for reordering pinned models is now explicitly disabled on mobile devices, ensuring better usability and preventing potential UI conflicts or unexpected behavior.
- 📱 **PWA Rotation Behavior Corrected**: The Progressive Web App now correctly respects the device's screen orientation lock, preventing unwanted rotation and ensuring a more native mobile experience.
- ✏️ **Improved Chat Title Editing Behavior**: Changes to a chat title are now reliably saved when the user clicks away or presses Enter, replacing a less intuitive behavior that could accidentally discard edits. This makes renaming chats a smoother and more predictable experience.
- ✏️ **Underscores Allowed in Prompt Commands**: Fixed the validation for prompt commands to correctly allow the use of underscores ('\_'), aligning with documentation examples and improving flexibility in naming custom prompts.
- 💡 **Title Generation Button Behavior Fixed**: Resolved an issue where clicking the "Generate Title" button while editing a chat or note title would incorrectly save the title before generation could start. The focus is now managed correctly, ensuring a smooth and predictable user experience.
- ✏️ **Consistent Chat Input Height**: Fixed a minor visual bug where the chat input field's height would change slightly when toggling the "Rich Text Input for Chat" setting, ensuring a more stable and consistent layout.
- 🙈 **Admin UI Toggle Stability**: Fixed a visual glitch in the Admin settings where toggle switches could briefly display an incorrect state on page load, ensuring the UI always accurately reflects the saved settings.
- 🙈 **Community Sharing Button Visibility**: The "Share to Community" button on the feedback page is now correctly hidden when the Enable Community Sharing feature is disabled in the admin settings, ensuring the UI respects the configured sharing policy.
- 🙈 **"Help Us Translate" Link Visibility**: The "Help us translate" link in settings is now correctly hidden in deployments with specific license configurations, ensuring a cleaner interface for enterprise users.
- 🔗 **Robust Tool Server URL Handling**: Fixed an issue where providing a full URL for a tool server's OpenAPI specification resulted in an invalid path. The system now correctly handles both absolute URLs and relative paths, improving configuration flexibility.
- 🔧 **Improved Azure URL Detection**: The logic for identifying Azure OpenAI endpoints has been made more robust, ensuring all valid Azure URLs are now correctly detected for a smoother connection setup.
- ⚙️ **Corrected Direct Connection Save Logic**: Fixed a bug in the Admin Connections settings page by removing a redundant save action for 'Direct Connections', leading to more reliable and predictable behavior when updating settings.
- 🔗 **Corrected "Discover" Links**: The "Discover" links for models, prompts, tools, and functions now point to their specific, relevant pages on openwebui.com, improving content discovery for users.
- ⏱️ **Refined Display of AI Thought Duration**: Adjusted the display logic for AI thought (reasoning) durations to more accurately show very short thought times as "less than a second," improving clarity in AI process feedback.
- 📜 **Markdown Line Break Rendering Refinement**: Improved handling of line breaks within Markdown rendering for better visual consistency.
- 🛠️ **Corrected OpenTelemetry Docker Compose Example**: The docker-compose.otel.yaml file has been fixed and enhanced by removing duplicates, adding necessary environment variables, and hardening security settings, ensuring a more reliable out-of-box observability setup.
- 🛠️ **Development Script CORS Fix**: Corrected the CORS origin URL in the local development script (dev.sh) by removing the trailing slash, ensuring a more reliable and consistent setup for developers.
- ⬆️ **OpenTelemetry Libraries Updated**: Upgraded all OpenTelemetry-related libraries to their latest versions, ensuring better performance, stability, and compatibility for observability.
### Changed
- ❗ **Docling Integration Upgraded to v1 API (Breaking Change)**: The integration with the Docling document processing engine has been updated to its new, stable '/v1' API. This is required for compatibility with Docling version 1.0.0 and newer. As a result, older versions of Docling are no longer supported. Users who rely on Docling for document ingestion **must upgrade** their docling-serve instance to ensure continued functionality.
- 🗣️ **Admin-First Whisper Language Priority**: The global WHISPER_LANGUAGE setting now acts as a strict override for audio transcriptions. If set, it will be used for all speech-to-text tasks, ignoring any language specified by the user on a per-request basis. This gives administrators more control over transcription consistency.
- ✂️ **Datalab Marker API Language Selection Removed**: The separate language selection option for the Datalab Marker API has been removed, as its functionality is now integrated and superseded by the more comprehensive 'additional_config' parameter. Users should transition to using 'additional_config' for relevant language and processing settings.
- 📄 **Documentation and Releases Links Visibility**: The "Documentation" and "Releases" links in the user menu are now visible only to admin users, streamlining the user interface for non-admin roles.
## [0.6.18] - 2025-07-19
### Fixed
- 🚑 **Users Not Loading in Groups**: Resolved an issue where user list was not displaying within user groups, restoring full visibility and management of group memberships for teams and admins.
## [0.6.17] - 2025-07-19
### Added
- 📂 **Dedicated Folder View with Chat List**: Clicking a folder now reveals a brand-new landing page showcasing a list of all chats within that folder, making navigation simpler and giving teams immediate visibility into project-specific conversations.
- 🆕 **Streamlined Folder Creation Modal**: Creating a new folder is now a seamless, unified experience with a dedicated modal that visually and functionally matches the edit folder flow, making workspace organization more intuitive and error-free for all users.
- 🗃️ **Direct File Uploads to Folder Knowledge**: You can now upload files straight to a folder’s knowledge—empowering you to enrich project spaces by adding resources and documents directly, without the need to pre-create knowledge bases beforehand.
- 🔎 **Chat Preview in Search**: When searching chats, instantly preview results in context without having to open them—making discovery, auditing, and recall dramatically quicker, especially in large, active teams.
- 🖼️ **Image Upload and Inline Insertion in Notes**: Notes now support inserting images directly among your text, letting you create rich, visually structured documentation, brainstorms, or reports in a more natural and engaging way—no more images just as attachments.
- 📱 **Enhanced Note Selection Editing and Q&A**: Select any portion of your notes to either edit just the highlighted part or ask focused questions about that content—streamlining workflows, boosting productivity, and making reviews or AI-powered enhancements more targeted.
- 📝 **Copy Notes as Rich Text**: Copy entire notes—including all formatting, images, and structure—directly as rich text for seamless pasting into emails, reports, or other tools, maintaining clarity and consistency outside the WebUI.
- ⚡ **Fade-In Streaming Text Experience**: Live-generated responses now elegantly fade in as the AI streams them, creating a more natural and visually engaging reading experience; easily toggled off in Interface settings if you prefer static displays.
- 🔄 **Settings for Follow-Up Prompts**: Fine-tune your follow-up prompt experience—with new controls, you can choose to keep them visible or have them inserted directly into the message input instead of auto-submitting, giving you more flexibility and control over your workflow.
- 🔗 **Prompt Variable Documentation Quick Link**: Access documentation for prompt variables in one click from the prompt editor modal—shortening the learning curve and making advanced prompt-building more accessible.
- 📈 **Active and Total User Metrics for Telemetry**: Gain valuable insights into usage patterns and platform engagement with new metrics tracking active and total users—enhancing auditability and planning for large organizations.
- 🏷️ **Traceability with Log Trace and Span IDs**: Each log entry now carries detailed trace and span IDs, making it much easier for admins to pinpoint and resolve issues across distributed systems or in complex troubleshooting.
- 👥 **User Group Add/Remove Endpoints**: Effortlessly add or remove users from groups with new, improved endpoints—giving admins and team leads faster, clearer control over collaboration and permissions.
- ⚙️ **Note Settings and Controls Streamlined**: The main “Settings” for notes are now simply called “Controls”, and note files now reside in a dedicated controls section, decluttering navigation and making it easier to find and configure note-related options.
- 🚀 **Faster Admin User Page Loads**: The user list endpoint for admins has been optimized to exclude heavy profile images, speeding up load times for large teams and reducing waiting during administrative tasks.
- 📡 **Chat ID Header Forwarding**: Ollama and OpenAI router requests now include the chat ID in request headers, enabling better request correlation and debugging capabilities across AI model integrations.
- 🧠 **Enhanced Reasoning Tag Processing**: Improved and expanded reasoning tag parsing to handle various tag formats more robustly, including standard XML-style tags and custom delimiters, ensuring better AI reasoning transparency and debugging capabilities.
- 🔐 **OAuth Token Endpoint Authentication Method**: Added configurable OAuth token endpoint authentication method support, providing enhanced flexibility and security options for enterprise OAuth integrations and identity provider compatibility.
- 🛡️ **Redis Sentinel High Availability Support**: Comprehensive Redis Sentinel failover implementation with automatic master discovery, intelligent retry logic for connection failures, and seamless operation during master node outages—eliminating single points of failure and ensuring continuous service availability in production deployments.
- 🌐 **Localization & Internationalization Improvements**: Refined and expanded translations for Simplified Chinese, Traditional Chinese, French, German, Korean, and Polish, ensuring a more fluent and native experience for global users across all supported languages.
### Fixed
- 🏷️ **Hybrid Search Functionality Restored**: Hybrid search now works seamlessly again—enabling more accurate, relevant, and comprehensive knowledge discovery across all RAG-powered workflows.
- 🚦 **Note Chat - Edit Button Disabled During AI Generation**: The edit button when chatting with a note is now disabled while the AI is responding—preventing accidental edits and ensuring workflow clarity during chat sessions.
- 🧹 **Cleaner Database Credentials**: Database connection no longer duplicates ‘@’ in credentials, preventing potential connection issues and ensuring smoother, more reliable integrations.
- 🧑‍💻 **File Deletion Now Removes Related Vector Data**: When files are deleted from storage, they are now purged from the vector database as well, ensuring clean data management and preventing clutter or stale search results.
- 📁 **Files Modal Translation Issues Fixed**: All modal dialog strings—including “Using Entire Document” and “Using Focused Retrieval”—are now fully translated for a more consistent and localized UI experience.
- 🚫 **Drag-and-Drop File Upload Disabled for Unsupported Models**: File upload by drag-and-drop is disabled when using models that do not support attachments—removing confusion and preventing workflow interruptions.
- 🔑 **Ollama Tool Calls Now Reliable**: Fixed issues with Ollama-based tool calls, ensuring uninterrupted AI augmentation and tool use for every chat.
- 📄 **MIME Type Help String Correction**: Cleaned up mimetype help text by removing extraneous characters, providing clearer guidance for file upload configurations.
- 📝 **Note Editor Permission Fix**: Removed unnecessary admin-only restriction from note chat functionality, allowing all authorized users to access note editing features as intended.
- 📋 **Chat Sources Handling Improved**: Fixed sources handling logic to prevent duplicate source assignments in chat messages, ensuring cleaner and more accurate source attribution during conversations.
- 😀 **Emoji Generation Error Handling**: Improved error handling in audio router and fixed metadata structure for emoji generation tasks, preventing crashes and ensuring more reliable emoji generation functionality.
- 🔒 **Folder System Prompt Permission Enforcement**: System prompt fields in folder edit modal are now properly hidden for users without system prompt permissions, ensuring consistent security policy enforcement across all folder management interfaces.
- 🌐 **WebSocket Redis Lock Timeout Type Conversion**: Fixed proper integer type conversion for WebSocket Redis lock timeout configuration with robust error handling, preventing potential configuration errors and ensuring stable WebSocket connections.
- 📦 **PostHog Dependency Added**: Added PostHog 5.4.0 library to resolve ChromaDB compatibility issues, ensuring stable vector database operations and preventing library version conflicts during deployment.
### Changed
- 👀 **Tiptap Editor Upgraded to v3**: The underlying rich text editor has been updated for future-proofing, though some supporting libraries remain on v2 for compatibility. For now, please install dependencies using 'npm install --force' to avoid installation errors.
- 🚫 **Removed Redundant or Unused Strings and Elements**: Miscellaneous unused, duplicate, or obsolete code and translations have been cleaned up to maintain a streamlined and high-performance experience.
## [0.6.16] - 2025-07-14
### Added
- 🗂️ **Folders as Projects**: Organize your workflow with folder-based projects—set folder-level system prompts and associate custom knowledge, bringing seamless, context-rich management to teams and users handling multiple initiatives or clients.
- 📁 **Instant Folder-Based Chat Creation**: Start a new chat directly from any folder; just click and your new conversation is automatically embedded in the right project context—no more manual dragging or setup, saving time and eliminating mistakes.
- 🧩 **Prompt Variables with Automatic Input Modal**: Prompts containing variables now display a clean, auto-generated input modal that **autofocuses on the first field** for instant value entry—just select the prompt and fill in exactly what’s needed, reducing friction and guesswork.
- 🔡 **Variable Input Typing in Prompts**: Define input types for prompt variables (e.g., text, textarea, number, select, color, date, map and more), giving everyone a clearer and more precise prompt-building experience for advanced automation or workflows.
- 🚀 **Base Model List Caching**: Cache your base model list to speed up model selection and reduce repeated API calls; toggle this in Admin Settings > Connections for responsive model management even in large or multi-provider setups.
- ⏱️ **Configurable Model List Cache TTL**: Take control over model list caching with the new MODEL_LIST_CACHE_TTL environment variable. Set a custom cache duration in seconds to balance performance and freshness, reducing API requests in stable environments or ensuring rapid updates when models change frequently.
- 🔖 **Reference Notes as Knowledge or in Chats**: Use any note as knowledge for a model or folder, or reference it directly from chat—integrate living documentation into your Retrieval Augmented Generation workflows or discussions, bridging knowledge and action.
- 📝 **Chat Directly with Notes (Experimental)**: Ask questions about any note, and directly edit or update notes from within a chat—unlock direct AI-powered brainstorming, summarization, and cleanup, like having your own collaborative AI canvas.
- 🤝 **Collaborative Notes with Multi-User Editing**: Share notes with others and collaborate live—multiple users can edit a note in real-time, boosting cooperative knowledge building and workflow documentation.
- 🛡️ **Collaborative Note Permissions**: Control who can view or edit each note with robust sharing permissions, ensuring privacy or collaboration per your organizational needs.
- 🔗 **Copy Link to Notes**: Quickly copy and share direct links to notes for easier knowledge transfer within your team or external collaborators.
- 📋 **Task List Support in Notes**: Add, organize, and manage checklists or tasks inside your notes—plan projects, track to-dos, and keep everything actionable in a single space.
- 🧠 **AI-Generated Note Titles**: Instantly generate relevant and concise titles for your notes using AI—keep your knowledge library organized without tedious manual editing.
- 🔄 **Full Undo/Redo Support in Notes**: Effortlessly undo or redo your latest note changes—never fear mistakes or accidental edits while collaborating or writing.
- 📝 **Enhanced Note Word/Character Counter**: Always know the size of your notes with built-in counters, making it easier to adhere to length guidelines for shared or published content.
- 🖊️ **Floating & Bubble Formatting Menus in Note Editor**: Access text formatting tools through both a floating menu and an intuitive bubble menu directly in the note editor—making rich text editing faster, more discoverable, and easier than ever.
- ✍️ **Rich Text Prompt Insertion**: A new setting allows prompts to be inserted directly into the chat box as fully-formatted rich text, preserving Markdown elements like headings, lists, and bold text for a more intuitive and visually consistent editing experience.
- 🌐 **Configurable Database URL**: WebUI now supports more flexible database configuration via new environment variables—making deployment and scaling simpler across various infrastructure setups.
- 🎛️ **Completely Frontend-Handled File Upload in Temporary Chats**: When using temporary chats, file extraction now occurs fully in your browser with zero files sent to the backend, further strengthening privacy and giving you instant feedback.
- 🔄 **Enhanced Banner and Chat Command Visibility**: Banner handling and command feedback in chat are now clearer and more contextually visible, making alerts, suggestions, and automation easier to spot and interact with for all users.
- 📱 **Mobile Experience Polished**: The "new chat" button is back in mobile, plus core navigation and input controls have been smoothed out for better usability on phones and tablets.
- 📄 **OpenDocument Text (.odt) Support**: Seamlessly upload and process .odt files from open-source office suites like LibreOffice and OpenOffice, expanding your ability to build knowledge from a wider range of document formats.
- 📑 **Enhanced Markdown Document Splitting**: Improve knowledge retrieval from Markdown files with a new header-aware splitting strategy. This method intelligently chunks documents based on their header structure, preserving the original context and hierarchy for more accurate and relevant RAG results.
- 📚 **Full Context Mode for Knowledge Bases**: When adding a knowledge base to a folder or custom model, you can now toggle full context mode for the entire knowledge base. This bypasses the usual chunking and retrieval process, making it perfect for leaner knowledge bases.
- 🕰️ **Configurable OAuth Timeout**: Enhance login reliability by setting a custom timeout (OAUTH_TIMEOUT) for all OAuth providers (Google, Microsoft, GitHub, OIDC), preventing authentication failures on slow or restricted networks.
- 🎨 **Accessibility & High-Contrast Theme Enhancements**: Major accessibility overhaul with significant updates to the high-contrast theme. Improved focus visibility, ARIA labels, and semantic HTML ensure core components like the chat interface and model selector are fully compliant and readable for visually impaired users.
- ↕️ **Resizable System Prompt Fields**: Conveniently resize system prompt input fields to comfortably view and edit lengthy or complex instructions, improving the user experience for advanced model configuration.
- 🔧 **Granular Update Check Control**: Gain finer control over outbound connections with the new ENABLE_VERSION_UPDATE_CHECK flag. This allows administrators to disable version update checks independently of the full OFFLINE_MODE, perfect for environments with restricted internet access that still need to download embedding models.
- 🗃️ **Configurable Qdrant Collection Prefix**: Enhance scalability by setting a custom QDRANT_COLLECTION_PREFIX. This allows multiple Open WebUI instances to share a single Qdrant cluster safely, ensuring complete data isolation between separate deployments without conflicts.
- ⚙️ **Improved Default Database Performance**: Enhanced out-of-the-box performance by setting smarter database connection pooling defaults, reducing API response times for users on non-SQLite databases without requiring manual configuration.
- 🔧 **Configurable Redis Key Prefix**: Added support for the REDIS_KEY_PREFIX environment variable, allowing multiple Open WebUI instances to share a Redis cluster with isolated key namespaces for improved multi-tenancy.
- ➡️ **Forward User Context to Reranker**: For advanced RAG integrations, user information (ID, name, email, role) can now be forwarded as HTTP headers to external reranking services, enabling personalized results or per-user access control.
- ⚙️ **PGVector Connection Pooling**: Enhance performance and stability for PGVector-based RAG by enabling and configuring the database connection pool. New environment variables allow fine-tuning of pool size, timeout, and overflow settings to handle high-concurrency workloads efficiently.
- ⚙️ **General Backend Refactoring**: Extensive refactoring delivers a faster, more reliable, and robust backend experience—improving chat speed, model management, and day-to-day reliability.
- 🌍 **Expanded & Improved Translations**: Enjoy a more accessible and intuitive experience thanks to comprehensive updates and enhancements for Chinese (Simplified and Traditional), German, French, Catalan, Irish, and Spanish translations throughout the interface.
### Fixed
- 🛠️ **Rich Text Input Stability and Performance**: Multiple improvements ensure faster, cleaner text editing and rendering with reduced glitches—especially supporting links, color picking, checkbox controls, and code blocks in notes and chats.
- 📷 **Seamless iPhone Image Uploads**: Effortlessly upload photos from iPhones and other devices using HEIC format—images are now correctly recognized and processed, eliminating compatibility issues.
- 🔄 **Audio MIME Type Registration**: Issues with audio file content types have been resolved, guaranteeing smoother, error-free uploads and playback for transcription or note attachments.
- 🖍️ **Input Commands Now Always Visible**: Input commands (like prompts or knowledge) dynamically adjust their height on small screens, ensuring nothing is cut off and every tool remains easily accessible.
- 🛑 **Tool Result Rendering**: Fixed display problems with tool results, providing fast, clear feedback when using external or internal tools.
- 🗂️ **Table Alignment in Markdown**: Markdown tables are now rendered and aligned as expected, keeping reports and documentation readable.
- 🖼️ **Thread Image Handling**: Fixed an issue where messages containing only images in threads weren’t displayed correctly.
- 🗝️ **Note Access Control Security**: Tightened access control logic for notes to guarantee that shared or collaborative notes respect all user permissions and privacy safeguards.
- 🧾 **Ollama API Compatibility**: Fixed model parameter naming in the API to ensure uninterrupted compatibility for all Ollama endpoints.
- 🛠️ **Detection for 'text/html' Files**: Files loaded with docling/tika are now reliably detected as the correct type, improving knowledge ingestion and document parsing.
- 🔐 **OAuth Login Stability**: Resolved a critical OAuth bug that caused login failures on subsequent attempts after logging out. The user session is now completely cleared on logout, ensuring reliable and secure authentication across all supported providers (Google, Microsoft, GitHub, OIDC).
- 🚪 **OAuth Logout and Redirect Reliability**: The OAuth logout process has been made more robust. Logout requests now correctly use proxy environment variables, ensuring they succeed in corporate networks. Additionally, the custom WEBUI_AUTH_SIGNOUT_REDIRECT_URL is now properly respected for all OAuth/OIDC configurations, ensuring a seamless sign-out experience.
- 📜 **Banner Newline Rendering**: Banners now correctly render newline characters, ensuring that multi-line announcements and messages are displayed with their intended formatting.
- ℹ️ **Consistent Model Description Rendering**: Model descriptions now render Markdown correctly in the main chat interface, matching the formatting seen in the model selection dropdown for a consistent user experience.
- 🔄 **Offline Mode Update Check Display**: Corrected a UI bug where the "Checking for Updates..." message would display indefinitely when the application was set to offline mode.
- 🛠️ **Tool Result Encoding**: Fixed a bug where tool calls returning non-ASCII characters would fail, ensuring robust handling of international text and special characters in tool outputs.
## [0.6.15] - 2025-06-16
### Added

64
Dockerfile
View file

@ -3,6 +3,8 @@
# use build args in the docker build command with --build-arg="BUILDARG=true"
ARG USE_CUDA=false
ARG USE_OLLAMA=false
ARG USE_SLIM=false
ARG USE_PERMISSION_HARDENING=false
# Tested with cu117 for CUDA 11 and cu121 for CUDA 12 (default)
ARG USE_CUDA_VER=cu128
# any sentence transformer model; models to use can be found at https://huggingface.co/models?library=sentence-transformers
@ -24,13 +26,16 @@ ARG GID=0
FROM --platform=$BUILDPLATFORM node:22-alpine3.20 AS build
ARG BUILD_HASH
# Set Node.js options (heap limit Allocation failed - JavaScript heap out of memory)
# ENV NODE_OPTIONS="--max-old-space-size=4096"
WORKDIR /app
# to store git revision in build
RUN apk add --no-cache git
COPY package.json package-lock.json ./
RUN npm ci
RUN npm ci --force
COPY . .
ENV APP_BUILD_HASH=${BUILD_HASH}
@ -43,6 +48,8 @@ FROM python:3.11-slim-bookworm AS base
ARG USE_CUDA
ARG USE_OLLAMA
ARG USE_CUDA_VER
ARG USE_SLIM
ARG USE_PERMISSION_HARDENING
ARG USE_EMBEDDING_MODEL
ARG USE_RERANKING_MODEL
ARG UID
@ -54,6 +61,7 @@ ENV ENV=prod \
# pass build args to the build
USE_OLLAMA_DOCKER=${USE_OLLAMA} \
USE_CUDA_DOCKER=${USE_CUDA} \
USE_SLIM_DOCKER=${USE_SLIM} \
USE_CUDA_DOCKER_VER=${USE_CUDA_VER} \
USE_EMBEDDING_MODEL_DOCKER=${USE_EMBEDDING_MODEL} \
USE_RERANKING_MODEL_DOCKER=${USE_RERANKING_MODEL}
@ -108,29 +116,13 @@ RUN echo -n 00000000-0000-0000-0000-000000000000 > $HOME/.cache/chroma/telemetry
# Make sure the user has access to the app and root directory
RUN chown -R $UID:$GID /app $HOME
RUN if [ "$USE_OLLAMA" = "true" ]; then \
apt-get update && \
# Install pandoc and netcat
apt-get install -y --no-install-recommends git build-essential pandoc netcat-openbsd curl && \
apt-get install -y --no-install-recommends gcc python3-dev && \
# for RAG OCR
apt-get install -y --no-install-recommends ffmpeg libsm6 libxext6 && \
# install helper tools
apt-get install -y --no-install-recommends curl jq && \
# install ollama
curl -fsSL https://ollama.com/install.sh | sh && \
# cleanup
rm -rf /var/lib/apt/lists/*; \
else \
apt-get update && \
# Install pandoc, netcat and gcc
apt-get install -y --no-install-recommends git build-essential pandoc gcc netcat-openbsd curl jq && \
apt-get install -y --no-install-recommends gcc python3-dev && \
# for RAG OCR
apt-get install -y --no-install-recommends ffmpeg libsm6 libxext6 && \
# cleanup
rm -rf /var/lib/apt/lists/*; \
fi
# Install common system dependencies
RUN apt-get update && \
apt-get install -y --no-install-recommends \
git build-essential pandoc gcc netcat-openbsd curl jq \
python3-dev \
ffmpeg libsm6 libxext6 \
&& rm -rf /var/lib/apt/lists/*
# install python dependencies
COPY --chown=$UID:$GID ./backend/requirements.txt ./requirements.txt
@ -146,13 +138,22 @@ RUN pip3 install --no-cache-dir uv && \
else \
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu --no-cache-dir && \
uv pip install --system -r requirements.txt --no-cache-dir && \
if [ "$USE_SLIM" != "true" ]; then \
python -c "import os; from sentence_transformers import SentenceTransformer; SentenceTransformer(os.environ['RAG_EMBEDDING_MODEL'], device='cpu')" && \
python -c "import os; from faster_whisper import WhisperModel; WhisperModel(os.environ['WHISPER_MODEL'], device='cpu', compute_type='int8', download_root=os.environ['WHISPER_MODEL_DIR'])"; \
python -c "import os; import tiktoken; tiktoken.get_encoding(os.environ['TIKTOKEN_ENCODING_NAME'])"; \
fi; \
chown -R $UID:$GID /app/backend/data/
fi; \
mkdir -p /app/backend/data && chown -R $UID:$GID /app/backend/data/ && \
rm -rf /var/lib/apt/lists/*;
# Install Ollama if requested
RUN if [ "$USE_OLLAMA" = "true" ]; then \
date +%s > /tmp/ollama_build_hash && \
echo "Cache broken at timestamp: `cat /tmp/ollama_build_hash`" && \
curl -fsSL https://ollama.com/install.sh | sh && \
rm -rf /var/lib/apt/lists/*; \
fi
# copy embedding weight from build
# RUN mkdir -p /root/.cache/chroma/onnx_models/all-MiniLM-L6-v2
@ -170,6 +171,17 @@ EXPOSE 8080
HEALTHCHECK CMD curl --silent --fail http://localhost:${PORT:-8080}/health | jq -ne 'input.status == true' || exit 1
# Minimal, atomic permission hardening for OpenShift (arbitrary UID):
# - Group 0 owns /app and /root
# - Directories are group-writable and have SGID so new files inherit GID 0
RUN if [ "$USE_PERMISSION_HARDENING" = "true" ]; then \
set -eux; \
chgrp -R 0 /app /root || true; \
chmod -R g+rwX /app /root || true; \
find /app -type d -exec chmod g+s {} + || true; \
find /root -type d -exec chmod g+s {} + || true; \
fi
USER $UID:$GID
ARG BUILD_HASH

53
LICENSE_HISTORY Normal file
View file

@ -0,0 +1,53 @@
All code and materials created before commit `60d84a3aae9802339705826e9095e272e3c83623` are subject to the following copyright and license:
Copyright (c) 2023-2025 Timothy Jaeryang Baek
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
1. Redistributions of source code must retain the above copyright notice, this
list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright notice,
this list of conditions and the following disclaimer in the documentation
and/or other materials provided with the distribution.
3. Neither the name of the copyright holder nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
All code and materials created before commit `a76068d69cd59568b920dfab85dc573dbbb8f131` are subject to the following copyright and license:
MIT License
Copyright (c) 2023 Timothy Jaeryang Baek
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

11
LICENSE_NOTICE Normal file
View file

@ -0,0 +1,11 @@
# Open WebUI Multi-License Notice
This repository contains code governed by multiple licenses based on the date and origin of contribution:
1. All code committed prior to commit a76068d69cd59568b920dfab85dc573dbbb8f131 is licensed under the MIT License (see LICENSE_HISTORY).
2. All code committed from commit a76068d69cd59568b920dfab85dc573dbbb8f131 up to and including commit 60d84a3aae9802339705826e9095e272e3c83623 is licensed under the BSD 3-Clause License (see LICENSE_HISTORY).
3. All code contributed or modified after commit 60d84a3aae9802339705826e9095e272e3c83623 is licensed under the Open WebUI License (see LICENSE).
For details on which commits are covered by which license, refer to LICENSE_HISTORY.

65
README.md
View file

@ -12,10 +12,12 @@
**Open WebUI is an [extensible](https://docs.openwebui.com/features/plugin/), feature-rich, and user-friendly self-hosted AI platform designed to operate entirely offline.** It supports various LLM runners like **Ollama** and **OpenAI-compatible APIs**, with **built-in inference engine** for RAG, making it a **powerful AI deployment solution**.
Passionate about open-source AI? [Join our team →](https://careers.openwebui.com/)
![Open WebUI Demo](./demo.gif)
> [!TIP]
> **Looking for an [Enterprise Plan](https://docs.openwebui.com/enterprise)?** – **[Speak with Our Sales Team Today!](mailto:sales@openwebui.com)**
> **Looking for an [Enterprise Plan](https://docs.openwebui.com/enterprise)?** – **[Speak with Our Sales Team Today!](https://docs.openwebui.com/enterprise)**
>
> Get **enhanced capabilities**, including **custom theming and branding**, **Service Level Agreement (SLA) support**, **Long-Term Support (LTS) versions**, and **more!**
@ -35,24 +37,38 @@ For more information, be sure to check out our [Open WebUI Documentation](https:
- ✒️🔢 **Full Markdown and LaTeX Support**: Elevate your LLM experience with comprehensive Markdown and LaTeX capabilities for enriched interaction.
- 🎤📹 **Hands-Free Voice/Video Call**: Experience seamless communication with integrated hands-free voice and video call features, allowing for a more dynamic and interactive chat environment.
- 🎤📹 **Hands-Free Voice/Video Call**: Experience seamless communication with integrated hands-free voice and video call features using multiple Speech-to-Text providers (Local Whisper, OpenAI, Deepgram, Azure) and Text-to-Speech engines (Azure, ElevenLabs, OpenAI, Transformers, WebAPI), allowing for dynamic and interactive chat environments.
- 🛠️ **Model Builder**: Easily create Ollama models via the Web UI. Create and add custom characters/agents, customize chat elements, and import models effortlessly through [Open WebUI Community](https://openwebui.com/) integration.
- 🐍 **Native Python Function Calling Tool**: Enhance your LLMs with built-in code editor support in the tools workspace. Bring Your Own Function (BYOF) by simply adding your pure Python functions, enabling seamless integration with LLMs.
- 📚 **Local RAG Integration**: Dive into the future of chat interactions with groundbreaking Retrieval Augmented Generation (RAG) support. This feature seamlessly integrates document interactions into your chat experience. You can load documents directly into the chat or add files to your document library, effortlessly accessing them using the `#` command before a query.
- 💾 **Persistent Artifact Storage**: Built-in key-value storage API for artifacts, enabling features like journals, trackers, leaderboards, and collaborative tools with both personal and shared data scopes across sessions.
- 🔍 **Web Search for RAG**: Perform web searches using providers like `SearXNG`, `Google PSE`, `Brave Search`, `serpstack`, `serper`, `Serply`, `DuckDuckGo`, `TavilySearch`, `SearchApi` and `Bing` and inject the results directly into your chat experience.
- 📚 **Local RAG Integration**: Dive into the future of chat interactions with groundbreaking Retrieval Augmented Generation (RAG) support using your choice of 9 vector databases and multiple content extraction engines (Tika, Docling, Document Intelligence, Mistral OCR, External loaders). Load documents directly into chat or add files to your document library, effortlessly accessing them using the `#` command before a query.
- 🔍 **Web Search for RAG**: Perform web searches using 15+ providers including `SearXNG`, `Google PSE`, `Brave Search`, `Kagi`, `Mojeek`, `Tavily`, `Perplexity`, `serpstack`, `serper`, `Serply`, `DuckDuckGo`, `SearchApi`, `SerpApi`, `Bing`, `Jina`, `Exa`, `Sougou`, `Azure AI Search`, and `Ollama Cloud`, injecting results directly into your chat experience.
- 🌐 **Web Browsing Capability**: Seamlessly integrate websites into your chat experience using the `#` command followed by a URL. This feature allows you to incorporate web content directly into your conversations, enhancing the richness and depth of your interactions.
- 🎨 **Image Generation Integration**: Seamlessly incorporate image generation capabilities using options such as AUTOMATIC1111 API or ComfyUI (local), and OpenAI's DALL-E (external), enriching your chat experience with dynamic visual content.
- 🎨 **Image Generation & Editing Integration**: Create and edit images using multiple engines including OpenAI's DALL-E, Gemini, ComfyUI (local), and AUTOMATIC1111 (local), with support for both generation and prompt-based editing workflows.
- ⚙️ **Many Models Conversations**: Effortlessly engage with various models simultaneously, harnessing their unique strengths for optimal responses. Enhance your experience by leveraging a diverse set of models in parallel.
- 🔐 **Role-Based Access Control (RBAC)**: Ensure secure access with restricted permissions; only authorized individuals can access your Ollama, and exclusive model creation/pulling rights are reserved for administrators.
- 🗄️ **Flexible Database & Storage Options**: Choose from SQLite (with optional encryption), PostgreSQL, or configure cloud storage backends (S3, Google Cloud Storage, Azure Blob Storage) for scalable deployments.
- 🔍 **Advanced Vector Database Support**: Select from 9 vector database options including ChromaDB, PGVector, Qdrant, Milvus, Elasticsearch, OpenSearch, Pinecone, S3Vector, and Oracle 23ai for optimal RAG performance.
- 🔐 **Enterprise Authentication**: Full support for LDAP/Active Directory integration, SCIM 2.0 automated provisioning, and SSO via trusted headers alongside OAuth providers. Enterprise-grade user and group provisioning through SCIM 2.0 protocol, enabling seamless integration with identity providers like Okta, Azure AD, and Google Workspace for automated user lifecycle management.
- ☁️ **Cloud-Native Integration**: Native support for Google Drive and OneDrive/SharePoint file picking, enabling seamless document import from enterprise cloud storage.
- 📊 **Production Observability**: Built-in OpenTelemetry support for traces, metrics, and logs, enabling comprehensive monitoring with your existing observability stack.
- ⚖️ **Horizontal Scalability**: Redis-backed session management and WebSocket support for multi-worker and multi-node deployments behind load balancers.
- 🌐🌍 **Multilingual Support**: Experience Open WebUI in your preferred language with our internationalization (i18n) support. Join us in expanding our supported languages! We're actively seeking contributors!
- 🧩 **Pipelines, Open WebUI Plugin Support**: Seamlessly integrate custom logic and Python libraries into Open WebUI using [Pipelines Plugin Framework](https://github.com/open-webui/pipelines). Launch your Pipelines instance, set the OpenAI URL to the Pipelines URL, and explore endless possibilities. [Examples](https://github.com/open-webui/pipelines/tree/main/examples) include **Function Calling**, User **Rate Limiting** to control access, **Usage Monitoring** with tools like Langfuse, **Live Translation with LibreTranslate** for multilingual support, **Toxic Message Filtering** and much more.
@ -61,43 +77,6 @@ For more information, be sure to check out our [Open WebUI Documentation](https:
Want to learn more about Open WebUI's features? Check out our [Open WebUI documentation](https://docs.openwebui.com/features) for a comprehensive overview!
## Sponsors 🙌
#### Emerald
<table>
<tr>
<td>
<a href="https://n8n.io/" target="_blank">
<img src="https://docs.openwebui.com/sponsors/logos/n8n.png" alt="n8n" style="width: 8rem; height: 8rem; border-radius: .75rem;" />
</a>
</td>
<td>
<a href="https://n8n.io/">n8n</a> • Does your interface have a backend yet?<br>Try <a href="https://n8n.io/">n8n</a>
</td>
</tr>
<tr>
<td>
<a href="https://warp.dev/open-webui" target="_blank">
<img src="https://docs.openwebui.com/sponsors/logos/warp.png" alt="Warp" style="width: 8rem; height: 8rem; border-radius: .75rem;" />
</a>
</td>
<td>
<a href="https://warp.dev/open-webui">Warp</a> • The intelligent terminal for developers
</td>
</tr>
<tr>
<td>
<a href="https://tailscale.com/blog/self-host-a-local-ai-stack/?utm_source=OpenWebUI&utm_medium=paid-ad-placement&utm_campaign=OpenWebUI-Docs" target="_blank">
<img src="https://docs.openwebui.com/sponsors/logos/tailscale.png" alt="Tailscale" style="width: 8rem; height: 8rem; border-radius: .75rem;" />
</a>
</td>
<td>
<a href="https://tailscale.com/blog/self-host-a-local-ai-stack/?utm_source=OpenWebUI&utm_medium=paid-ad-placement&utm_campaign=OpenWebUI-Docs">Tailscale</a> • Connect self-hosted AI to any device with Tailscale
</td>
</tr>
</table>
---
We are incredibly grateful for the generous support of our sponsors. Their contributions help us to maintain and improve our project, ensuring we can continue to deliver quality work to our community. Thank you!
@ -244,7 +223,7 @@ Discover upcoming features on our roadmap in the [Open WebUI Documentation](http
## License 📜
This project is licensed under the [Open WebUI License](LICENSE), a revised BSD-3-Clause license. You receive all the same rights as the classic BSD-3 license: you can use, modify, and distribute the software, including in proprietary and commercial products, with minimal restrictions. The only additional requirement is to preserve the "Open WebUI" branding, as detailed in the LICENSE file. For full terms, see the [LICENSE](LICENSE) document. 📄
This project contains code under multiple licenses. The current codebase includes components licensed under the Open WebUI License with an additional requirement to preserve the "Open WebUI" branding, as well as prior contributions under their respective original licenses. For a detailed record of license changes and the applicable terms for each section of the code, please refer to [LICENSE_HISTORY](./LICENSE_HISTORY). For complete and updated licensing details, please see the [LICENSE](./LICENSE) and [LICENSE_HISTORY](./LICENSE_HISTORY) files.
## Support 💬

1
backend/dev.sh
View file

@ -1,2 +1,3 @@
export CORS_ALLOW_ORIGIN="http://localhost:5173;http://localhost:8080"
PORT="${PORT:-8080}"
uvicorn open_webui.main:app --port $PORT --host 0.0.0.0 --forwarded-allow-ips '*' --reload

2
backend/open_webui/alembic.ini
View file

@ -10,7 +10,7 @@ script_location = migrations
# sys.path path, will be prepended to sys.path if present.
# defaults to the current working directory.
prepend_sys_path = .
prepend_sys_path = ..
# timezone to use when rendering the date within the migration file
# as well as the filename.

1005
backend/open_webui/config.py
View file

File diff suppressed because it is too large Load diff

7
backend/open_webui/constants.py
View file

@ -38,13 +38,14 @@ class ERROR_MESSAGES(str, Enum):
ID_TAKEN = "Uh-oh! This id is already registered. Please choose another id string."
MODEL_ID_TAKEN = "Uh-oh! This model id is already registered. Please choose another model id string."
NAME_TAG_TAKEN = "Uh-oh! This name tag is already registered. Please choose another name tag string."
MODEL_ID_TOO_LONG = "The model id is too long. Please make sure your model id is less than 256 characters long."
INVALID_TOKEN = (
"Your session has expired or the token is invalid. Please sign in again."
)
INVALID_CRED = "The email or password provided is incorrect. Please check for typos and try logging in again."
INVALID_EMAIL_FORMAT = "The email format you entered is invalid. Please double-check and make sure you're using a valid email address (e.g., yourname@example.com)."
INVALID_PASSWORD = (
INCORRECT_PASSWORD = (
"The password provided is incorrect. Please check for typos and try again."
)
INVALID_TRUSTED_HEADER = "Your provider has not provided a trusted header. Please contact your administrator for assistance."
@ -104,6 +105,10 @@ class ERROR_MESSAGES(str, Enum):
)
FILE_NOT_PROCESSED = "Extracted content is not available for this file. Please ensure that the file is processed before proceeding."
INVALID_PASSWORD = lambda err="": (
err if err else "The password does not meet the required validation criteria."
)
class TASKS(str, Enum):
def __str__(self) -> str:

348
backend/open_webui/env.py
View file

@ -7,6 +7,9 @@ import sys
import shutil
from uuid import uuid4
from pathlib import Path
from cryptography.hazmat.primitives import serialization
import re
import markdown
from bs4 import BeautifulSoup
@ -16,14 +19,17 @@ from open_webui.constants import ERROR_MESSAGES
# Load .env file
####################################
OPEN_WEBUI_DIR = Path(__file__).parent # the path containing this file
print(OPEN_WEBUI_DIR)
# Use .resolve() to get the canonical path, removing any '..' or '.' components
ENV_FILE_PATH = Path(__file__).resolve()
BACKEND_DIR = OPEN_WEBUI_DIR.parent # the path containing this file
BASE_DIR = BACKEND_DIR.parent # the path containing the backend/
# OPEN_WEBUI_DIR should be the directory where env.py resides (open_webui/)
OPEN_WEBUI_DIR = ENV_FILE_PATH.parent
print(BACKEND_DIR)
print(BASE_DIR)
# BACKEND_DIR is the parent of OPEN_WEBUI_DIR (backend/)
BACKEND_DIR = OPEN_WEBUI_DIR.parent
# BASE_DIR is the parent of BACKEND_DIR (open-webui-dev/)
BASE_DIR = BACKEND_DIR.parent
try:
from dotenv import find_dotenv, load_dotenv
@ -131,6 +137,9 @@ else:
PACKAGE_DATA = {"version": "0.0.0"}
VERSION = PACKAGE_DATA["version"]
DEPLOYMENT_ID = os.environ.get("DEPLOYMENT_ID", "")
INSTANCE_ID = os.environ.get("INSTANCE_ID", str(uuid4()))
@ -199,6 +208,7 @@ CHANGELOG = changelog_json
SAFE_MODE = os.environ.get("SAFE_MODE", "false").lower() == "true"
####################################
# ENABLE_FORWARD_USER_INFO_HEADERS
####################################
@ -207,6 +217,11 @@ ENABLE_FORWARD_USER_INFO_HEADERS = (
os.environ.get("ENABLE_FORWARD_USER_INFO_HEADERS", "False").lower() == "true"
)
# Experimental feature, may be removed in future
ENABLE_STAR_SESSIONS_MIDDLEWARE = (
os.environ.get("ENABLE_STAR_SESSIONS_MIDDLEWARE", "False").lower() == "true"
)
####################################
# WEBUI_BUILD_HASH
####################################
@ -266,21 +281,43 @@ else:
DATABASE_URL = os.environ.get("DATABASE_URL", f"sqlite:///{DATA_DIR}/webui.db")
DATABASE_TYPE = os.environ.get("DATABASE_TYPE")
DATABASE_USER = os.environ.get("DATABASE_USER")
DATABASE_PASSWORD = os.environ.get("DATABASE_PASSWORD")
DATABASE_CRED = ""
if DATABASE_USER:
DATABASE_CRED += f"{DATABASE_USER}"
if DATABASE_PASSWORD:
DATABASE_CRED += f":{DATABASE_PASSWORD}"
DB_VARS = {
"db_type": DATABASE_TYPE,
"db_cred": DATABASE_CRED,
"db_host": os.environ.get("DATABASE_HOST"),
"db_port": os.environ.get("DATABASE_PORT"),
"db_name": os.environ.get("DATABASE_NAME"),
}
if all(DB_VARS.values()):
DATABASE_URL = f"{DB_VARS['db_type']}://{DB_VARS['db_cred']}@{DB_VARS['db_host']}:{DB_VARS['db_port']}/{DB_VARS['db_name']}"
elif DATABASE_TYPE == "sqlite+sqlcipher" and not os.environ.get("DATABASE_URL"):
# Handle SQLCipher with local file when DATABASE_URL wasn't explicitly set
DATABASE_URL = f"sqlite+sqlcipher:///{DATA_DIR}/webui.db"
# Replace the postgres:// with postgresql://
if "postgres://" in DATABASE_URL:
DATABASE_URL = DATABASE_URL.replace("postgres://", "postgresql://")
DATABASE_SCHEMA = os.environ.get("DATABASE_SCHEMA", None)
DATABASE_POOL_SIZE = os.environ.get("DATABASE_POOL_SIZE", 0)
DATABASE_POOL_SIZE = os.environ.get("DATABASE_POOL_SIZE", None)
if DATABASE_POOL_SIZE == "":
DATABASE_POOL_SIZE = 0
else:
if DATABASE_POOL_SIZE != None:
try:
DATABASE_POOL_SIZE = int(DATABASE_POOL_SIZE)
except Exception:
DATABASE_POOL_SIZE = 0
DATABASE_POOL_SIZE = None
DATABASE_POOL_MAX_OVERFLOW = os.environ.get("DATABASE_POOL_MAX_OVERFLOW", 0)
@ -312,6 +349,21 @@ else:
except Exception:
DATABASE_POOL_RECYCLE = 3600
DATABASE_ENABLE_SQLITE_WAL = (
os.environ.get("DATABASE_ENABLE_SQLITE_WAL", "False").lower() == "true"
)
DATABASE_USER_ACTIVE_STATUS_UPDATE_INTERVAL = os.environ.get(
"DATABASE_USER_ACTIVE_STATUS_UPDATE_INTERVAL", None
)
if DATABASE_USER_ACTIVE_STATUS_UPDATE_INTERVAL is not None:
try:
DATABASE_USER_ACTIVE_STATUS_UPDATE_INTERVAL = float(
DATABASE_USER_ACTIVE_STATUS_UPDATE_INTERVAL
)
except Exception:
DATABASE_USER_ACTIVE_STATUS_UPDATE_INTERVAL = 0.0
RESET_CONFIG_ON_START = (
os.environ.get("RESET_CONFIG_ON_START", "False").lower() == "true"
)
@ -320,14 +372,29 @@ ENABLE_REALTIME_CHAT_SAVE = (
os.environ.get("ENABLE_REALTIME_CHAT_SAVE", "False").lower() == "true"
)
ENABLE_QUERIES_CACHE = os.environ.get("ENABLE_QUERIES_CACHE", "False").lower() == "true"
####################################
# REDIS
####################################
REDIS_URL = os.environ.get("REDIS_URL", "")
REDIS_CLUSTER = os.environ.get("REDIS_CLUSTER", "False").lower() == "true"
REDIS_KEY_PREFIX = os.environ.get("REDIS_KEY_PREFIX", "open-webui")
REDIS_SENTINEL_HOSTS = os.environ.get("REDIS_SENTINEL_HOSTS", "")
REDIS_SENTINEL_PORT = os.environ.get("REDIS_SENTINEL_PORT", "26379")
# Maximum number of retries for Redis operations when using Sentinel fail-over
REDIS_SENTINEL_MAX_RETRY_COUNT = os.environ.get("REDIS_SENTINEL_MAX_RETRY_COUNT", "2")
try:
REDIS_SENTINEL_MAX_RETRY_COUNT = int(REDIS_SENTINEL_MAX_RETRY_COUNT)
if REDIS_SENTINEL_MAX_RETRY_COUNT < 1:
REDIS_SENTINEL_MAX_RETRY_COUNT = 2
except ValueError:
REDIS_SENTINEL_MAX_RETRY_COUNT = 2
####################################
# UVICORN WORKERS
####################################
@ -347,6 +414,14 @@ except ValueError:
####################################
WEBUI_AUTH = os.environ.get("WEBUI_AUTH", "True").lower() == "true"
ENABLE_INITIAL_ADMIN_SIGNUP = (
os.environ.get("ENABLE_INITIAL_ADMIN_SIGNUP", "False").lower() == "true"
)
ENABLE_SIGNUP_PASSWORD_CONFIRMATION = (
os.environ.get("ENABLE_SIGNUP_PASSWORD_CONFIRMATION", "False").lower() == "true"
)
WEBUI_AUTH_TRUSTED_EMAIL_HEADER = os.environ.get(
"WEBUI_AUTH_TRUSTED_EMAIL_HEADER", None
)
@ -356,6 +431,17 @@ WEBUI_AUTH_TRUSTED_GROUPS_HEADER = os.environ.get(
)
ENABLE_PASSWORD_VALIDATION = (
os.environ.get("ENABLE_PASSWORD_VALIDATION", "False").lower() == "true"
)
PASSWORD_VALIDATION_REGEX_PATTERN = os.environ.get(
"PASSWORD_VALIDATION_REGEX_PATTERN",
"^(?=.*[a-z])(?=.*[A-Z])(?=.*\d)(?=.*[^\w\s]).{8,}$",
)
PASSWORD_VALIDATION_REGEX_PATTERN = re.compile(PASSWORD_VALIDATION_REGEX_PATTERN)
BYPASS_MODEL_ACCESS_CONTROL = (
os.environ.get("BYPASS_MODEL_ACCESS_CONTROL", "False").lower() == "true"
)
@ -396,18 +482,186 @@ WEBUI_AUTH_COOKIE_SECURE = (
if WEBUI_AUTH and WEBUI_SECRET_KEY == "":
raise ValueError(ERROR_MESSAGES.ENV_VAR_NOT_FOUND)
ENABLE_COMPRESSION_MIDDLEWARE = (
os.environ.get("ENABLE_COMPRESSION_MIDDLEWARE", "True").lower() == "true"
)
####################################
# OAUTH Configuration
####################################
ENABLE_OAUTH_EMAIL_FALLBACK = (
os.environ.get("ENABLE_OAUTH_EMAIL_FALLBACK", "False").lower() == "true"
)
ENABLE_OAUTH_ID_TOKEN_COOKIE = (
os.environ.get("ENABLE_OAUTH_ID_TOKEN_COOKIE", "True").lower() == "true"
)
OAUTH_CLIENT_INFO_ENCRYPTION_KEY = os.environ.get(
"OAUTH_CLIENT_INFO_ENCRYPTION_KEY", WEBUI_SECRET_KEY
)
OAUTH_SESSION_TOKEN_ENCRYPTION_KEY = os.environ.get(
"OAUTH_SESSION_TOKEN_ENCRYPTION_KEY", WEBUI_SECRET_KEY
)
####################################
# SCIM Configuration
####################################
ENABLE_SCIM = (
os.environ.get("ENABLE_SCIM", os.environ.get("SCIM_ENABLED", "False")).lower()
== "true"
)
SCIM_TOKEN = os.environ.get("SCIM_TOKEN", "")
####################################
# LICENSE_KEY
####################################
LICENSE_KEY = os.environ.get("LICENSE_KEY", "")
LICENSE_BLOB = None
LICENSE_BLOB_PATH = os.environ.get("LICENSE_BLOB_PATH", DATA_DIR / "l.data")
if LICENSE_BLOB_PATH and os.path.exists(LICENSE_BLOB_PATH):
with open(LICENSE_BLOB_PATH, "rb") as f:
LICENSE_BLOB = f.read()
LICENSE_PUBLIC_KEY = os.environ.get("LICENSE_PUBLIC_KEY", "")
pk = None
if LICENSE_PUBLIC_KEY:
pk = serialization.load_pem_public_key(
f"""
-----BEGIN PUBLIC KEY-----
{LICENSE_PUBLIC_KEY}
-----END PUBLIC KEY-----
""".encode(
"utf-8"
)
)
####################################
# MODELS
####################################
MODELS_CACHE_TTL = os.environ.get("MODELS_CACHE_TTL", "1")
if MODELS_CACHE_TTL == "":
MODELS_CACHE_TTL = None
else:
try:
MODELS_CACHE_TTL = int(MODELS_CACHE_TTL)
except Exception:
MODELS_CACHE_TTL = 1
####################################
# CHAT
####################################
ENABLE_CHAT_RESPONSE_BASE64_IMAGE_URL_CONVERSION = (
os.environ.get("ENABLE_CHAT_RESPONSE_BASE64_IMAGE_URL_CONVERSION", "False").lower()
== "true"
)
CHAT_RESPONSE_STREAM_DELTA_CHUNK_SIZE = os.environ.get(
"CHAT_RESPONSE_STREAM_DELTA_CHUNK_SIZE", "1"
)
if CHAT_RESPONSE_STREAM_DELTA_CHUNK_SIZE == "":
CHAT_RESPONSE_STREAM_DELTA_CHUNK_SIZE = 1
else:
try:
CHAT_RESPONSE_STREAM_DELTA_CHUNK_SIZE = int(
CHAT_RESPONSE_STREAM_DELTA_CHUNK_SIZE
)
except Exception:
CHAT_RESPONSE_STREAM_DELTA_CHUNK_SIZE = 1
CHAT_RESPONSE_MAX_TOOL_CALL_RETRIES = os.environ.get(
"CHAT_RESPONSE_MAX_TOOL_CALL_RETRIES", "30"
)
if CHAT_RESPONSE_MAX_TOOL_CALL_RETRIES == "":
CHAT_RESPONSE_MAX_TOOL_CALL_RETRIES = 30
else:
try:
CHAT_RESPONSE_MAX_TOOL_CALL_RETRIES = int(CHAT_RESPONSE_MAX_TOOL_CALL_RETRIES)
except Exception:
CHAT_RESPONSE_MAX_TOOL_CALL_RETRIES = 30
CHAT_STREAM_RESPONSE_CHUNK_MAX_BUFFER_SIZE = os.environ.get(
"CHAT_STREAM_RESPONSE_CHUNK_MAX_BUFFER_SIZE", ""
)
if CHAT_STREAM_RESPONSE_CHUNK_MAX_BUFFER_SIZE == "":
CHAT_STREAM_RESPONSE_CHUNK_MAX_BUFFER_SIZE = None
else:
try:
CHAT_STREAM_RESPONSE_CHUNK_MAX_BUFFER_SIZE = int(
CHAT_STREAM_RESPONSE_CHUNK_MAX_BUFFER_SIZE
)
except Exception:
CHAT_STREAM_RESPONSE_CHUNK_MAX_BUFFER_SIZE = None
####################################
# WEBSOCKET SUPPORT
####################################
ENABLE_WEBSOCKET_SUPPORT = (
os.environ.get("ENABLE_WEBSOCKET_SUPPORT", "True").lower() == "true"
)
WEBSOCKET_MANAGER = os.environ.get("WEBSOCKET_MANAGER", "")
WEBSOCKET_REDIS_OPTIONS = os.environ.get("WEBSOCKET_REDIS_OPTIONS", "")
if WEBSOCKET_REDIS_OPTIONS == "":
log.debug("No WEBSOCKET_REDIS_OPTIONS provided, defaulting to None")
WEBSOCKET_REDIS_OPTIONS = None
else:
try:
WEBSOCKET_REDIS_OPTIONS = json.loads(WEBSOCKET_REDIS_OPTIONS)
except Exception:
log.warning("Invalid WEBSOCKET_REDIS_OPTIONS, defaulting to None")
WEBSOCKET_REDIS_OPTIONS = None
WEBSOCKET_REDIS_URL = os.environ.get("WEBSOCKET_REDIS_URL", REDIS_URL)
WEBSOCKET_REDIS_LOCK_TIMEOUT = os.environ.get("WEBSOCKET_REDIS_LOCK_TIMEOUT", 60)
WEBSOCKET_REDIS_CLUSTER = (
os.environ.get("WEBSOCKET_REDIS_CLUSTER", str(REDIS_CLUSTER)).lower() == "true"
)
websocket_redis_lock_timeout = os.environ.get("WEBSOCKET_REDIS_LOCK_TIMEOUT", "60")
try:
WEBSOCKET_REDIS_LOCK_TIMEOUT = int(websocket_redis_lock_timeout)
except ValueError:
WEBSOCKET_REDIS_LOCK_TIMEOUT = 60
WEBSOCKET_SENTINEL_HOSTS = os.environ.get("WEBSOCKET_SENTINEL_HOSTS", "")
WEBSOCKET_SENTINEL_PORT = os.environ.get("WEBSOCKET_SENTINEL_PORT", "26379")
WEBSOCKET_SERVER_LOGGING = (
os.environ.get("WEBSOCKET_SERVER_LOGGING", "False").lower() == "true"
)
WEBSOCKET_SERVER_ENGINEIO_LOGGING = (
os.environ.get("WEBSOCKET_SERVER_LOGGING", "False").lower() == "true"
)
WEBSOCKET_SERVER_PING_TIMEOUT = os.environ.get("WEBSOCKET_SERVER_PING_TIMEOUT", "20")
try:
WEBSOCKET_SERVER_PING_TIMEOUT = int(WEBSOCKET_SERVER_PING_TIMEOUT)
except ValueError:
WEBSOCKET_SERVER_PING_TIMEOUT = 20
WEBSOCKET_SERVER_PING_INTERVAL = os.environ.get("WEBSOCKET_SERVER_PING_INTERVAL", "25")
try:
WEBSOCKET_SERVER_PING_INTERVAL = int(WEBSOCKET_SERVER_PING_INTERVAL)
except ValueError:
WEBSOCKET_SERVER_PING_INTERVAL = 25
AIOHTTP_CLIENT_TIMEOUT = os.environ.get("AIOHTTP_CLIENT_TIMEOUT", "")
@ -506,19 +760,32 @@ else:
# OFFLINE_MODE
####################################
ENABLE_VERSION_UPDATE_CHECK = (
os.environ.get("ENABLE_VERSION_UPDATE_CHECK", "true").lower() == "true"
)
OFFLINE_MODE = os.environ.get("OFFLINE_MODE", "false").lower() == "true"
if OFFLINE_MODE:
os.environ["HF_HUB_OFFLINE"] = "1"
ENABLE_VERSION_UPDATE_CHECK = False
####################################
# AUDIT LOGGING
####################################
# Where to store log file
AUDIT_LOGS_FILE_PATH = f"{DATA_DIR}/audit.log"
# Defaults to the DATA_DIR/audit.log. To set AUDIT_LOGS_FILE_PATH you need to
# provide the whole path, like: /app/audit.log
AUDIT_LOGS_FILE_PATH = os.getenv("AUDIT_LOGS_FILE_PATH", f"{DATA_DIR}/audit.log")
# Maximum size of a file before rotating into a new log file
AUDIT_LOG_FILE_ROTATION_SIZE = os.getenv("AUDIT_LOG_FILE_ROTATION_SIZE", "10MB")
# Comma separated list of logger names to use for audit logging
# Default is "uvicorn.access" which is the access log for Uvicorn
# You can add more logger names to this list if you want to capture more logs
AUDIT_UVICORN_LOGGER_NAMES = os.getenv(
"AUDIT_UVICORN_LOGGER_NAMES", "uvicorn.access"
).split(",")
# METADATA | REQUEST | REQUEST_RESPONSE
AUDIT_LOG_LEVEL = os.getenv("AUDIT_LOG_LEVEL", "NONE").upper()
try:
@ -539,10 +806,34 @@ AUDIT_EXCLUDED_PATHS = [path.lstrip("/") for path in AUDIT_EXCLUDED_PATHS]
####################################
ENABLE_OTEL = os.environ.get("ENABLE_OTEL", "False").lower() == "true"
ENABLE_OTEL_TRACES = os.environ.get("ENABLE_OTEL_TRACES", "False").lower() == "true"
ENABLE_OTEL_METRICS = os.environ.get("ENABLE_OTEL_METRICS", "False").lower() == "true"
ENABLE_OTEL_LOGS = os.environ.get("ENABLE_OTEL_LOGS", "False").lower() == "true"
OTEL_EXPORTER_OTLP_ENDPOINT = os.environ.get(
"OTEL_EXPORTER_OTLP_ENDPOINT", "http://localhost:4317"
)
OTEL_METRICS_EXPORTER_OTLP_ENDPOINT = os.environ.get(
"OTEL_METRICS_EXPORTER_OTLP_ENDPOINT", OTEL_EXPORTER_OTLP_ENDPOINT
)
OTEL_LOGS_EXPORTER_OTLP_ENDPOINT = os.environ.get(
"OTEL_LOGS_EXPORTER_OTLP_ENDPOINT", OTEL_EXPORTER_OTLP_ENDPOINT
)
OTEL_EXPORTER_OTLP_INSECURE = (
os.environ.get("OTEL_EXPORTER_OTLP_INSECURE", "False").lower() == "true"
)
OTEL_METRICS_EXPORTER_OTLP_INSECURE = (
os.environ.get(
"OTEL_METRICS_EXPORTER_OTLP_INSECURE", str(OTEL_EXPORTER_OTLP_INSECURE)
).lower()
== "true"
)
OTEL_LOGS_EXPORTER_OTLP_INSECURE = (
os.environ.get(
"OTEL_LOGS_EXPORTER_OTLP_INSECURE", str(OTEL_EXPORTER_OTLP_INSECURE)
).lower()
== "true"
)
OTEL_SERVICE_NAME = os.environ.get("OTEL_SERVICE_NAME", "open-webui")
OTEL_RESOURCE_ATTRIBUTES = os.environ.get(
"OTEL_RESOURCE_ATTRIBUTES", ""
@ -550,6 +841,33 @@ OTEL_RESOURCE_ATTRIBUTES = os.environ.get(
OTEL_TRACES_SAMPLER = os.environ.get(
"OTEL_TRACES_SAMPLER", "parentbased_always_on"
).lower()
OTEL_BASIC_AUTH_USERNAME = os.environ.get("OTEL_BASIC_AUTH_USERNAME", "")
OTEL_BASIC_AUTH_PASSWORD = os.environ.get("OTEL_BASIC_AUTH_PASSWORD", "")
OTEL_METRICS_BASIC_AUTH_USERNAME = os.environ.get(
"OTEL_METRICS_BASIC_AUTH_USERNAME", OTEL_BASIC_AUTH_USERNAME
)
OTEL_METRICS_BASIC_AUTH_PASSWORD = os.environ.get(
"OTEL_METRICS_BASIC_AUTH_PASSWORD", OTEL_BASIC_AUTH_PASSWORD
)
OTEL_LOGS_BASIC_AUTH_USERNAME = os.environ.get(
"OTEL_LOGS_BASIC_AUTH_USERNAME", OTEL_BASIC_AUTH_USERNAME
)
OTEL_LOGS_BASIC_AUTH_PASSWORD = os.environ.get(
"OTEL_LOGS_BASIC_AUTH_PASSWORD", OTEL_BASIC_AUTH_PASSWORD
)
OTEL_OTLP_SPAN_EXPORTER = os.environ.get(
"OTEL_OTLP_SPAN_EXPORTER", "grpc"
).lower() # grpc or http
OTEL_METRICS_OTLP_SPAN_EXPORTER = os.environ.get(
"OTEL_METRICS_OTLP_SPAN_EXPORTER", OTEL_OTLP_SPAN_EXPORTER
).lower() # grpc or http
OTEL_LOGS_OTLP_SPAN_EXPORTER = os.environ.get(
"OTEL_LOGS_OTLP_SPAN_EXPORTER", OTEL_OTLP_SPAN_EXPORTER
).lower() # grpc or http
####################################
# TOOLS/FUNCTIONS PIP OPTIONS

130
backend/open_webui/functions.py
View file

@ -19,6 +19,7 @@ from fastapi import (
from starlette.responses import Response, StreamingResponse
from open_webui.constants import ERROR_MESSAGES
from open_webui.socket.main import (
get_event_call,
get_event_emitter,
@ -47,7 +48,7 @@ from open_webui.utils.misc import (
)
from open_webui.utils.payload import (
apply_model_params_to_body_openai,
apply_model_system_prompt_to_body,
apply_system_prompt_to_body,
)
@ -60,8 +61,20 @@ def get_function_module_by_id(request: Request, pipe_id: str):
function_module, _, _ = get_function_module_from_cache(request, pipe_id)
if hasattr(function_module, "valves") and hasattr(function_module, "Valves"):
Valves = function_module.Valves
valves = Functions.get_function_valves_by_id(pipe_id)
function_module.valves = function_module.Valves(**(valves if valves else {}))
if valves:
try:
function_module.valves = Valves(
**{k: v for k, v in valves.items() if v is not None}
)
except Exception as e:
log.exception(f"Error loading valves for function {pipe_id}: {e}")
raise e
else:
function_module.valves = Valves()
return function_module
@ -70,65 +83,75 @@ async def get_function_models(request):
pipe_models = []
for pipe in pipes:
function_module = get_function_module_by_id(request, pipe.id)
try:
function_module = get_function_module_by_id(request, pipe.id)
# Check if function is a manifold
if hasattr(function_module, "pipes"):
sub_pipes = []
has_user_valves = False
if hasattr(function_module, "UserValves"):
has_user_valves = True
# Handle pipes being a list, sync function, or async function
try:
if callable(function_module.pipes):
if asyncio.iscoroutinefunction(function_module.pipes):
sub_pipes = await function_module.pipes()
else:
sub_pipes = function_module.pipes()
else:
sub_pipes = function_module.pipes
except Exception as e:
log.exception(e)
# Check if function is a manifold
if hasattr(function_module, "pipes"):
sub_pipes = []
log.debug(
f"get_function_models: function '{pipe.id}' is a manifold of {sub_pipes}"
)
# Handle pipes being a list, sync function, or async function
try:
if callable(function_module.pipes):
if asyncio.iscoroutinefunction(function_module.pipes):
sub_pipes = await function_module.pipes()
else:
sub_pipes = function_module.pipes()
else:
sub_pipes = function_module.pipes
except Exception as e:
log.exception(e)
sub_pipes = []
for p in sub_pipes:
sub_pipe_id = f'{pipe.id}.{p["id"]}'
sub_pipe_name = p["name"]
log.debug(
f"get_function_models: function '{pipe.id}' is a manifold of {sub_pipes}"
)
if hasattr(function_module, "name"):
sub_pipe_name = f"{function_module.name}{sub_pipe_name}"
for p in sub_pipes:
sub_pipe_id = f'{pipe.id}.{p["id"]}'
sub_pipe_name = p["name"]
pipe_flag = {"type": pipe.type}
if hasattr(function_module, "name"):
sub_pipe_name = f"{function_module.name}{sub_pipe_name}"
pipe_flag = {"type": pipe.type}
pipe_models.append(
{
"id": sub_pipe_id,
"name": sub_pipe_name,
"object": "model",
"created": pipe.created_at,
"owned_by": "openai",
"pipe": pipe_flag,
"has_user_valves": has_user_valves,
}
)
else:
pipe_flag = {"type": "pipe"}
log.debug(
f"get_function_models: function '{pipe.id}' is a single pipe {{ 'id': {pipe.id}, 'name': {pipe.name} }}"
)
pipe_models.append(
{
"id": sub_pipe_id,
"name": sub_pipe_name,
"id": pipe.id,
"name": pipe.name,
"object": "model",
"created": pipe.created_at,
"owned_by": "openai",
"pipe": pipe_flag,
"has_user_valves": has_user_valves,
}
)
else:
pipe_flag = {"type": "pipe"}
log.debug(
f"get_function_models: function '{pipe.id}' is a single pipe {{ 'id': {pipe.id}, 'name': {pipe.name} }}"
)
pipe_models.append(
{
"id": pipe.id,
"name": pipe.name,
"object": "model",
"created": pipe.created_at,
"owned_by": "openai",
"pipe": pipe_flag,
}
)
except Exception as e:
log.exception(e)
continue
return pipe_models
@ -219,6 +242,16 @@ async def generate_function_chat_completion(
__task__ = metadata.get("task", None)
__task_body__ = metadata.get("task_body", None)
oauth_token = None
try:
if request.cookies.get("oauth_session_id", None):
oauth_token = await request.app.state.oauth_manager.get_oauth_token(
user.id,
request.cookies.get("oauth_session_id", None),
)
except Exception as e:
log.error(f"Error getting OAuth token: {e}")
extra_params = {
"__event_emitter__": __event_emitter__,
"__event_call__": __event_call__,
@ -230,9 +263,10 @@ async def generate_function_chat_completion(
"__files__": files,
"__user__": user.model_dump() if isinstance(user, UserModel) else {},
"__metadata__": metadata,
"__oauth_token__": oauth_token,
"__request__": request,
}
extra_params["__tools__"] = get_tools(
extra_params["__tools__"] = await get_tools(
request,
tool_ids,
user,
@ -253,9 +287,7 @@ async def generate_function_chat_completion(
if params:
system = params.pop("system", None)
form_data = apply_model_params_to_body_openai(params, form_data)
form_data = apply_model_system_prompt_to_body(
system, form_data, metadata, user
)
form_data = apply_system_prompt_to_body(system, form_data, metadata, user)
pipe_id = get_pipe_id(form_data)
function_module = get_function_module_by_id(request, pipe_id)

78
backend/open_webui/internal/db.py
View file

@ -1,3 +1,4 @@
import os
import json
import logging
from contextlib import contextmanager
@ -13,9 +14,10 @@ from open_webui.env import (
DATABASE_POOL_RECYCLE,
DATABASE_POOL_SIZE,
DATABASE_POOL_TIMEOUT,
DATABASE_ENABLE_SQLITE_WAL,
)
from peewee_migrate import Router
from sqlalchemy import Dialect, create_engine, MetaData, types
from sqlalchemy import Dialect, create_engine, MetaData, event, types
from sqlalchemy.ext.declarative import declarative_base
from sqlalchemy.orm import scoped_session, sessionmaker
from sqlalchemy.pool import QueuePool, NullPool
@ -62,6 +64,9 @@ def handle_peewee_migration(DATABASE_URL):
except Exception as e:
log.error(f"Failed to initialize the database connection: {e}")
log.warning(
"Hint: If your database password contains special characters, you may need to URL-encode it."
)
raise
finally:
# Properly closing the database connection
@ -76,25 +81,68 @@ handle_peewee_migration(DATABASE_URL)
SQLALCHEMY_DATABASE_URL = DATABASE_URL
if "sqlite" in SQLALCHEMY_DATABASE_URL:
# Handle SQLCipher URLs
if SQLALCHEMY_DATABASE_URL.startswith("sqlite+sqlcipher://"):
database_password = os.environ.get("DATABASE_PASSWORD")
if not database_password or database_password.strip() == "":
raise ValueError(
"DATABASE_PASSWORD is required when using sqlite+sqlcipher:// URLs"
)
# Extract database path from SQLCipher URL
db_path = SQLALCHEMY_DATABASE_URL.replace("sqlite+sqlcipher://", "")
if db_path.startswith("/"):
db_path = db_path[1:] # Remove leading slash for relative paths
# Create a custom creator function that uses sqlcipher3
def create_sqlcipher_connection():
import sqlcipher3
conn = sqlcipher3.connect(db_path, check_same_thread=False)
conn.execute(f"PRAGMA key = '{database_password}'")
return conn
engine = create_engine(
"sqlite://", # Dummy URL since we're using creator
creator=create_sqlcipher_connection,
echo=False,
)
log.info("Connected to encrypted SQLite database using SQLCipher")
elif "sqlite" in SQLALCHEMY_DATABASE_URL:
engine = create_engine(
SQLALCHEMY_DATABASE_URL, connect_args={"check_same_thread": False}
)
def on_connect(dbapi_connection, connection_record):
cursor = dbapi_connection.cursor()
if DATABASE_ENABLE_SQLITE_WAL:
cursor.execute("PRAGMA journal_mode=WAL")
else:
cursor.execute("PRAGMA journal_mode=DELETE")
cursor.close()
event.listen(engine, "connect", on_connect)
else:
if DATABASE_POOL_SIZE > 0:
engine = create_engine(
SQLALCHEMY_DATABASE_URL,
pool_size=DATABASE_POOL_SIZE,
max_overflow=DATABASE_POOL_MAX_OVERFLOW,
pool_timeout=DATABASE_POOL_TIMEOUT,
pool_recycle=DATABASE_POOL_RECYCLE,
pool_pre_ping=True,
poolclass=QueuePool,
)
if isinstance(DATABASE_POOL_SIZE, int):
if DATABASE_POOL_SIZE > 0:
engine = create_engine(
SQLALCHEMY_DATABASE_URL,
pool_size=DATABASE_POOL_SIZE,
max_overflow=DATABASE_POOL_MAX_OVERFLOW,
pool_timeout=DATABASE_POOL_TIMEOUT,
pool_recycle=DATABASE_POOL_RECYCLE,
pool_pre_ping=True,
poolclass=QueuePool,
)
else:
engine = create_engine(
SQLALCHEMY_DATABASE_URL, pool_pre_ping=True, poolclass=NullPool
)
else:
engine = create_engine(
SQLALCHEMY_DATABASE_URL, pool_pre_ping=True, poolclass=NullPool
)
engine = create_engine(SQLALCHEMY_DATABASE_URL, pool_pre_ping=True)
SessionLocal = sessionmaker(

56
backend/open_webui/internal/wrappers.py
View file

@ -1,4 +1,5 @@
import logging
import os
from contextvars import ContextVar
from open_webui.env import SRC_LOG_LEVELS
@ -43,24 +44,47 @@ class ReconnectingPostgresqlDatabase(CustomReconnectMixin, PostgresqlDatabase):
def register_connection(db_url):
db = connect(db_url, unquote_user=True, unquote_password=True)
if isinstance(db, PostgresqlDatabase):
# Enable autoconnect for SQLite databases, managed by Peewee
# Check if using SQLCipher protocol
if db_url.startswith("sqlite+sqlcipher://"):
database_password = os.environ.get("DATABASE_PASSWORD")
if not database_password or database_password.strip() == "":
raise ValueError(
"DATABASE_PASSWORD is required when using sqlite+sqlcipher:// URLs"
)
from playhouse.sqlcipher_ext import SqlCipherDatabase
# Parse the database path from SQLCipher URL
# Convert sqlite+sqlcipher:///path/to/db.sqlite to /path/to/db.sqlite
db_path = db_url.replace("sqlite+sqlcipher://", "")
if db_path.startswith("/"):
db_path = db_path[1:] # Remove leading slash for relative paths
# Use Peewee's native SqlCipherDatabase with encryption
db = SqlCipherDatabase(db_path, passphrase=database_password)
db.autoconnect = True
db.reuse_if_open = True
log.info("Connected to PostgreSQL database")
log.info("Connected to encrypted SQLite database using SQLCipher")
# Get the connection details
connection = parse(db_url, unquote_user=True, unquote_password=True)
# Use our custom database class that supports reconnection
db = ReconnectingPostgresqlDatabase(**connection)
db.connect(reuse_if_open=True)
elif isinstance(db, SqliteDatabase):
# Enable autoconnect for SQLite databases, managed by Peewee
db.autoconnect = True
db.reuse_if_open = True
log.info("Connected to SQLite database")
else:
raise ValueError("Unsupported database connection")
# Standard database connection (existing logic)
db = connect(db_url, unquote_user=True, unquote_password=True)
if isinstance(db, PostgresqlDatabase):
# Enable autoconnect for SQLite databases, managed by Peewee
db.autoconnect = True
db.reuse_if_open = True
log.info("Connected to PostgreSQL database")
# Get the connection details
connection = parse(db_url, unquote_user=True, unquote_password=True)
# Use our custom database class that supports reconnection
db = ReconnectingPostgresqlDatabase(**connection)
db.connect(reuse_if_open=True)
elif isinstance(db, SqliteDatabase):
# Enable autoconnect for SQLite databases, managed by Peewee
db.autoconnect = True
db.reuse_if_open = True
log.info("Connected to SQLite database")
else:
raise ValueError("Unsupported database connection")
return db

835
backend/open_webui/main.py
View file

File diff suppressed because it is too large Load diff

41
backend/open_webui/migrations/env.py
View file

@ -2,8 +2,8 @@ from logging.config import fileConfig
from alembic import context
from open_webui.models.auths import Auth
from open_webui.env import DATABASE_URL
from sqlalchemy import engine_from_config, pool
from open_webui.env import DATABASE_URL, DATABASE_PASSWORD
from sqlalchemy import engine_from_config, pool, create_engine
# this is the Alembic Config object, which provides
# access to the values within the .ini file in use.
@ -62,11 +62,38 @@ def run_migrations_online() -> None:
and associate a connection with the context.
"""
connectable = engine_from_config(
config.get_section(config.config_ini_section, {}),
prefix="sqlalchemy.",
poolclass=pool.NullPool,
)
# Handle SQLCipher URLs
if DB_URL and DB_URL.startswith("sqlite+sqlcipher://"):
if not DATABASE_PASSWORD or DATABASE_PASSWORD.strip() == "":
raise ValueError(
"DATABASE_PASSWORD is required when using sqlite+sqlcipher:// URLs"
)
# Extract database path from SQLCipher URL
db_path = DB_URL.replace("sqlite+sqlcipher://", "")
if db_path.startswith("/"):
db_path = db_path[1:] # Remove leading slash for relative paths
# Create a custom creator function that uses sqlcipher3
def create_sqlcipher_connection():
import sqlcipher3
conn = sqlcipher3.connect(db_path, check_same_thread=False)
conn.execute(f"PRAGMA key = '{DATABASE_PASSWORD}'")
return conn
connectable = create_engine(
"sqlite://", # Dummy URL since we're using creator
creator=create_sqlcipher_connection,
echo=False,
)
else:
# Standard database connection (existing logic)
connectable = engine_from_config(
config.get_section(config.config_ini_section, {}),
prefix="sqlalchemy.",
poolclass=pool.NullPool,
)
with connectable.connect() as connection:
context.configure(connection=connection, target_metadata=target_metadata)

46
backend/open_webui/migrations/versions/018012973d35_add_indexes.py Normal file
View file

@ -0,0 +1,46 @@
"""Add indexes
Revision ID: 018012973d35
Revises: d31026856c01
Create Date: 2025-08-13 03:00:00.000000
"""
from alembic import op
import sqlalchemy as sa
revision = "018012973d35"
down_revision = "d31026856c01"
branch_labels = None
depends_on = None
def upgrade():
# Chat table indexes
op.create_index("folder_id_idx", "chat", ["folder_id"])
op.create_index("user_id_pinned_idx", "chat", ["user_id", "pinned"])
op.create_index("user_id_archived_idx", "chat", ["user_id", "archived"])
op.create_index("updated_at_user_id_idx", "chat", ["updated_at", "user_id"])
op.create_index("folder_id_user_id_idx", "chat", ["folder_id", "user_id"])
# Tag table index
op.create_index("user_id_idx", "tag", ["user_id"])
# Function table index
op.create_index("is_global_idx", "function", ["is_global"])
def downgrade():
# Chat table indexes
op.drop_index("folder_id_idx", table_name="chat")
op.drop_index("user_id_pinned_idx", table_name="chat")
op.drop_index("user_id_archived_idx", table_name="chat")
op.drop_index("updated_at_user_id_idx", table_name="chat")
op.drop_index("folder_id_user_id_idx", table_name="chat")
# Tag table index
op.drop_index("user_id_idx", table_name="tag")
# Function table index
op.drop_index("is_global_idx", table_name="function")

146
backend/open_webui/migrations/versions/37f288994c47_add_group_member_table.py Normal file
View file

@ -0,0 +1,146 @@
"""add_group_member_table
Revision ID: 37f288994c47
Revises: a5c220713937
Create Date: 2025-11-17 03:45:25.123939
"""
import uuid
import time
import json
from typing import Sequence, Union
from alembic import op
import sqlalchemy as sa
# revision identifiers, used by Alembic.
revision: str = "37f288994c47"
down_revision: Union[str, None] = "a5c220713937"
branch_labels: Union[str, Sequence[str], None] = None
depends_on: Union[str, Sequence[str], None] = None
def upgrade() -> None:
# 1. Create new table
op.create_table(
"group_member",
sa.Column("id", sa.Text(), primary_key=True, unique=True, nullable=False),
sa.Column(
"group_id",
sa.Text(),
sa.ForeignKey("group.id", ondelete="CASCADE"),
nullable=False,
),
sa.Column(
"user_id",
sa.Text(),
sa.ForeignKey("user.id", ondelete="CASCADE"),
nullable=False,
),
sa.Column("created_at", sa.BigInteger(), nullable=True),
sa.Column("updated_at", sa.BigInteger(), nullable=True),
sa.UniqueConstraint("group_id", "user_id", name="uq_group_member_group_user"),
)
connection = op.get_bind()
# 2. Read existing group with user_ids JSON column
group_table = sa.Table(
"group",
sa.MetaData(),
sa.Column("id", sa.Text()),
sa.Column("user_ids", sa.JSON()), # JSON stored as text in SQLite + PG
)
results = connection.execute(
sa.select(group_table.c.id, group_table.c.user_ids)
).fetchall()
print(results)
# 3. Insert members into group_member table
gm_table = sa.Table(
"group_member",
sa.MetaData(),
sa.Column("id", sa.Text()),
sa.Column("group_id", sa.Text()),
sa.Column("user_id", sa.Text()),
sa.Column("created_at", sa.BigInteger()),
sa.Column("updated_at", sa.BigInteger()),
)
now = int(time.time())
for group_id, user_ids in results:
if not user_ids:
continue
if isinstance(user_ids, str):
try:
user_ids = json.loads(user_ids)
except Exception:
continue # skip invalid JSON
if not isinstance(user_ids, list):
continue
rows = [
{
"id": str(uuid.uuid4()),
"group_id": group_id,
"user_id": uid,
"created_at": now,
"updated_at": now,
}
for uid in user_ids
]
if rows:
connection.execute(gm_table.insert(), rows)
# 4. Optionally drop the old column
with op.batch_alter_table("group") as batch:
batch.drop_column("user_ids")
def downgrade():
# Reverse: restore user_ids column
with op.batch_alter_table("group") as batch:
batch.add_column(sa.Column("user_ids", sa.JSON()))
connection = op.get_bind()
gm_table = sa.Table(
"group_member",
sa.MetaData(),
sa.Column("group_id", sa.Text()),
sa.Column("user_id", sa.Text()),
sa.Column("created_at", sa.BigInteger()),
sa.Column("updated_at", sa.BigInteger()),
)
group_table = sa.Table(
"group",
sa.MetaData(),
sa.Column("id", sa.Text()),
sa.Column("user_ids", sa.JSON()),
)
# Build JSON arrays again
results = connection.execute(sa.select(group_table.c.id)).fetchall()
for (group_id,) in results:
members = connection.execute(
sa.select(gm_table.c.user_id).where(gm_table.c.group_id == group_id)
).fetchall()
member_ids = [m[0] for m in members]
connection.execute(
group_table.update()
.where(group_table.c.id == group_id)
.values(user_ids=member_ids)
)
# Drop the new table
op.drop_table("group_member")

52
backend/open_webui/migrations/versions/38d63c18f30f_add_oauth_session_table.py Normal file
View file

@ -0,0 +1,52 @@
"""Add oauth_session table
Revision ID: 38d63c18f30f
Revises: 3af16a1c9fb6
Create Date: 2025-09-08 14:19:59.583921
"""
from typing import Sequence, Union
from alembic import op
import sqlalchemy as sa
# revision identifiers, used by Alembic.
revision: str = "38d63c18f30f"
down_revision: Union[str, None] = "3af16a1c9fb6"
branch_labels: Union[str, Sequence[str], None] = None
depends_on: Union[str, Sequence[str], None] = None
def upgrade() -> None:
# Create oauth_session table
op.create_table(
"oauth_session",
sa.Column("id", sa.Text(), nullable=False),
sa.Column("user_id", sa.Text(), nullable=False),
sa.Column("provider", sa.Text(), nullable=False),
sa.Column("token", sa.Text(), nullable=False),
sa.Column("expires_at", sa.BigInteger(), nullable=False),
sa.Column("created_at", sa.BigInteger(), nullable=False),
sa.Column("updated_at", sa.BigInteger(), nullable=False),
sa.PrimaryKeyConstraint("id"),
sa.ForeignKeyConstraint(["user_id"], ["user.id"], ondelete="CASCADE"),
)
# Create indexes for better performance
op.create_index("idx_oauth_session_user_id", "oauth_session", ["user_id"])
op.create_index("idx_oauth_session_expires_at", "oauth_session", ["expires_at"])
op.create_index(
"idx_oauth_session_user_provider", "oauth_session", ["user_id", "provider"]
)
def downgrade() -> None:
# Drop indexes first
op.drop_index("idx_oauth_session_user_provider", table_name="oauth_session")
op.drop_index("idx_oauth_session_expires_at", table_name="oauth_session")
op.drop_index("idx_oauth_session_user_id", table_name="oauth_session")
# Drop the table
op.drop_table("oauth_session")

32
backend/open_webui/migrations/versions/3af16a1c9fb6_update_user_table.py Normal file
View file

@ -0,0 +1,32 @@
"""update user table
Revision ID: 3af16a1c9fb6
Revises: 018012973d35
Create Date: 2025-08-21 02:07:18.078283
"""
from typing import Sequence, Union
from alembic import op
import sqlalchemy as sa
# revision identifiers, used by Alembic.
revision: str = "3af16a1c9fb6"
down_revision: Union[str, None] = "018012973d35"
branch_labels: Union[str, Sequence[str], None] = None
depends_on: Union[str, Sequence[str], None] = None
def upgrade() -> None:
op.add_column("user", sa.Column("username", sa.String(length=50), nullable=True))
op.add_column("user", sa.Column("bio", sa.Text(), nullable=True))
op.add_column("user", sa.Column("gender", sa.Text(), nullable=True))
op.add_column("user", sa.Column("date_of_birth", sa.Date(), nullable=True))
def downgrade() -> None:
op.drop_column("user", "username")
op.drop_column("user", "bio")
op.drop_column("user", "gender")
op.drop_column("user", "date_of_birth")

34
backend/open_webui/migrations/versions/a5c220713937_add_reply_to_id_column_to_message.py Normal file
View file

@ -0,0 +1,34 @@
"""Add reply_to_id column to message
Revision ID: a5c220713937
Revises: 38d63c18f30f
Create Date: 2025-09-27 02:24:18.058455
"""
from typing import Sequence, Union
from alembic import op
import sqlalchemy as sa
# revision identifiers, used by Alembic.
revision: str = "a5c220713937"
down_revision: Union[str, None] = "38d63c18f30f"
branch_labels: Union[str, Sequence[str], None] = None
depends_on: Union[str, Sequence[str], None] = None
def upgrade() -> None:
# Add 'reply_to_id' column to the 'message' table for replying to messages
op.add_column(
"message",
sa.Column("reply_to_id", sa.Text(), nullable=True),
)
pass
def downgrade() -> None:
# Remove 'reply_to_id' column from the 'message' table
op.drop_column("message", "reply_to_id")
pass

23
backend/open_webui/migrations/versions/d31026856c01_update_folder_table_data.py Normal file
View file

@ -0,0 +1,23 @@
"""Update folder table data
Revision ID: d31026856c01
Revises: 9f0c9cd09105
Create Date: 2025-07-13 03:00:00.000000
"""
from alembic import op
import sqlalchemy as sa
revision = "d31026856c01"
down_revision = "9f0c9cd09105"
branch_labels = None
depends_on = None
def upgrade():
op.add_column("folder", sa.Column("data", sa.JSON(), nullable=True))
def downgrade():
op.drop_column("folder", "data")

14
backend/open_webui/models/auths.py
View file

@ -7,7 +7,6 @@ from open_webui.models.users import UserModel, Users
from open_webui.env import SRC_LOG_LEVELS
from pydantic import BaseModel
from sqlalchemy import Boolean, Column, String, Text
from open_webui.utils.auth import verify_password
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["MODELS"])
@ -20,7 +19,7 @@ log.setLevel(SRC_LOG_LEVELS["MODELS"])
class Auth(Base):
__tablename__ = "auth"
id = Column(String, primary_key=True)
id = Column(String, primary_key=True, unique=True)
email = Column(String)
password = Column(Text)
active = Column(Boolean)
@ -73,11 +72,6 @@ class ProfileImageUrlForm(BaseModel):
profile_image_url: str
class UpdateProfileForm(BaseModel):
profile_image_url: str
name: str
class UpdatePasswordForm(BaseModel):
password: str
new_password: str
@ -127,7 +121,9 @@ class AuthsTable:
else:
return None
def authenticate_user(self, email: str, password: str) -> Optional[UserModel]:
def authenticate_user(
self, email: str, verify_password: callable
) -> Optional[UserModel]:
log.info(f"authenticate_user: {email}")
user = Users.get_user_by_email(email)
@ -138,7 +134,7 @@ class AuthsTable:
with get_db() as db:
auth = db.query(Auth).filter_by(id=user.id, active=True).first()
if auth:
if verify_password(password, auth.password):
if verify_password(auth.password):
return user
else:
return None

7
backend/open_webui/models/channels.py
View file

@ -19,7 +19,7 @@ from sqlalchemy.sql import exists
class Channel(Base):
__tablename__ = "channel"
id = Column(Text, primary_key=True)
id = Column(Text, primary_key=True, unique=True)
user_id = Column(Text)
type = Column(Text, nullable=True)
@ -57,6 +57,11 @@ class ChannelModel(BaseModel):
####################
class ChannelResponse(ChannelModel):
write_access: bool = False
user_count: Optional[int] = None
class ChannelForm(BaseModel):
name: str
description: Optional[str] = None

379
backend/open_webui/models/chats.py
View file

@ -6,12 +6,14 @@ from typing import Optional
from open_webui.internal.db import Base, get_db
from open_webui.models.tags import TagModel, Tag, Tags
from open_webui.models.folders import Folders
from open_webui.env import SRC_LOG_LEVELS
from pydantic import BaseModel, ConfigDict
from sqlalchemy import BigInteger, Boolean, Column, String, Text, JSON
from sqlalchemy import BigInteger, Boolean, Column, String, Text, JSON, Index
from sqlalchemy import or_, func, select, and_, text
from sqlalchemy.sql import exists
from sqlalchemy.sql.expression import bindparam
####################
# Chat DB Schema
@ -24,7 +26,7 @@ log.setLevel(SRC_LOG_LEVELS["MODELS"])
class Chat(Base):
__tablename__ = "chat"
id = Column(String, primary_key=True)
id = Column(String, primary_key=True, unique=True)
user_id = Column(String)
title = Column(Text)
chat = Column(JSON)
@ -39,6 +41,20 @@ class Chat(Base):
meta = Column(JSON, server_default="{}")
folder_id = Column(Text, nullable=True)
__table_args__ = (
# Performance indexes for common queries
# WHERE folder_id = ...
Index("folder_id_idx", "folder_id"),
# WHERE user_id = ... AND pinned = ...
Index("user_id_pinned_idx", "user_id", "pinned"),
# WHERE user_id = ... AND archived = ...
Index("user_id_archived_idx", "user_id", "archived"),
# WHERE user_id = ... ORDER BY updated_at DESC
Index("updated_at_user_id_idx", "updated_at", "user_id"),
# WHERE folder_id = ... AND user_id = ...
Index("folder_id_user_id_idx", "folder_id", "user_id"),
)
class ChatModel(BaseModel):
model_config = ConfigDict(from_attributes=True)
@ -66,12 +82,18 @@ class ChatModel(BaseModel):
class ChatForm(BaseModel):
chat: dict
folder_id: Optional[str] = None
class ChatImportForm(ChatForm):
meta: Optional[dict] = {}
pinned: Optional[bool] = False
folder_id: Optional[str] = None
created_at: Optional[int] = None
updated_at: Optional[int] = None
class ChatsImportForm(BaseModel):
chats: list[ChatImportForm]
class ChatTitleMessagesForm(BaseModel):
@ -105,6 +127,43 @@ class ChatTitleIdResponse(BaseModel):
class ChatTable:
def _clean_null_bytes(self, obj):
"""
Recursively remove actual null bytes (\x00) and unicode escape \\u0000
from strings inside dict/list structures.
Safe for JSON objects.
"""
if isinstance(obj, str):
return obj.replace("\x00", "").replace("\u0000", "")
elif isinstance(obj, dict):
return {k: self._clean_null_bytes(v) for k, v in obj.items()}
elif isinstance(obj, list):
return [self._clean_null_bytes(v) for v in obj]
return obj
def _sanitize_chat_row(self, chat_item):
"""
Clean a Chat SQLAlchemy model's title + chat JSON,
and return True if anything changed.
"""
changed = False
# Clean title
if chat_item.title:
cleaned = self._clean_null_bytes(chat_item.title)
if cleaned != chat_item.title:
chat_item.title = cleaned
changed = True
# Clean JSON
if chat_item.chat:
cleaned = self._clean_null_bytes(chat_item.chat)
if cleaned != chat_item.chat:
chat_item.chat = cleaned
changed = True
return changed
def insert_new_chat(self, user_id: str, form_data: ChatForm) -> Optional[ChatModel]:
with get_db() as db:
id = str(uuid.uuid4())
@ -112,59 +171,76 @@ class ChatTable:
**{
"id": id,
"user_id": user_id,
"title": (
"title": self._clean_null_bytes(
form_data.chat["title"]
if "title" in form_data.chat
else "New Chat"
),
"chat": form_data.chat,
"created_at": int(time.time()),
"updated_at": int(time.time()),
}
)
result = Chat(**chat.model_dump())
db.add(result)
db.commit()
db.refresh(result)
return ChatModel.model_validate(result) if result else None
def import_chat(
self, user_id: str, form_data: ChatImportForm
) -> Optional[ChatModel]:
with get_db() as db:
id = str(uuid.uuid4())
chat = ChatModel(
**{
"id": id,
"user_id": user_id,
"title": (
form_data.chat["title"]
if "title" in form_data.chat
else "New Chat"
),
"chat": form_data.chat,
"meta": form_data.meta,
"pinned": form_data.pinned,
"chat": self._clean_null_bytes(form_data.chat),
"folder_id": form_data.folder_id,
"created_at": int(time.time()),
"updated_at": int(time.time()),
}
)
result = Chat(**chat.model_dump())
db.add(result)
chat_item = Chat(**chat.model_dump())
db.add(chat_item)
db.commit()
db.refresh(result)
return ChatModel.model_validate(result) if result else None
db.refresh(chat_item)
return ChatModel.model_validate(chat_item) if chat_item else None
def _chat_import_form_to_chat_model(
self, user_id: str, form_data: ChatImportForm
) -> ChatModel:
id = str(uuid.uuid4())
chat = ChatModel(
**{
"id": id,
"user_id": user_id,
"title": self._clean_null_bytes(
form_data.chat["title"] if "title" in form_data.chat else "New Chat"
),
"chat": self._clean_null_bytes(form_data.chat),
"meta": form_data.meta,
"pinned": form_data.pinned,
"folder_id": form_data.folder_id,
"created_at": (
form_data.created_at if form_data.created_at else int(time.time())
),
"updated_at": (
form_data.updated_at if form_data.updated_at else int(time.time())
),
}
)
return chat
def import_chats(
self, user_id: str, chat_import_forms: list[ChatImportForm]
) -> list[ChatModel]:
with get_db() as db:
chats = []
for form_data in chat_import_forms:
chat = self._chat_import_form_to_chat_model(user_id, form_data)
chats.append(Chat(**chat.model_dump()))
db.add_all(chats)
db.commit()
return [ChatModel.model_validate(chat) for chat in chats]
def update_chat_by_id(self, id: str, chat: dict) -> Optional[ChatModel]:
try:
with get_db() as db:
chat_item = db.get(Chat, id)
chat_item.chat = chat
chat_item.title = chat["title"] if "title" in chat else "New Chat"
chat_item.chat = self._clean_null_bytes(chat)
chat_item.title = (
self._clean_null_bytes(chat["title"])
if "title" in chat
else "New Chat"
)
chat_item.updated_at = int(time.time())
db.commit()
db.refresh(chat_item)
@ -209,7 +285,7 @@ class ChatTable:
return chat.chat.get("title", "New Chat")
def get_messages_by_chat_id(self, id: str) -> Optional[dict]:
def get_messages_map_by_chat_id(self, id: str) -> Optional[dict]:
chat = self.get_chat_by_id(id)
if chat is None:
return None
@ -232,6 +308,10 @@ class ChatTable:
if chat is None:
return None
# Sanitize message content for null characters before upserting
if isinstance(message.get("content"), str):
message["content"] = message["content"].replace("\x00", "")
chat = chat.chat
history = chat.get("history", {})
@ -266,6 +346,27 @@ class ChatTable:
chat["history"] = history
return self.update_chat_by_id(id, chat)
def add_message_files_by_id_and_message_id(
self, id: str, message_id: str, files: list[dict]
) -> list[dict]:
chat = self.get_chat_by_id(id)
if chat is None:
return None
chat = chat.chat
history = chat.get("history", {})
message_files = []
if message_id in history.get("messages", {}):
message_files = history["messages"][message_id].get("files", [])
message_files = message_files + files
history["messages"][message_id]["files"] = message_files
chat["history"] = history
self.update_chat_by_id(id, chat)
return message_files
def insert_shared_chat_by_chat_id(self, chat_id: str) -> Optional[ChatModel]:
with get_db() as db:
# Get the existing chat to share
@ -280,6 +381,9 @@ class ChatTable:
"user_id": f"shared-{chat_id}",
"title": chat.title,
"chat": chat.chat,
"meta": chat.meta,
"pinned": chat.pinned,
"folder_id": chat.folder_id,
"created_at": chat.created_at,
"updated_at": int(time.time()),
}
@ -311,7 +415,9 @@ class ChatTable:
shared_chat.title = chat.title
shared_chat.chat = chat.chat
shared_chat.meta = chat.meta
shared_chat.pinned = chat.pinned
shared_chat.folder_id = chat.folder_id
shared_chat.updated_at = int(time.time())
db.commit()
db.refresh(shared_chat)
@ -330,6 +436,15 @@ class ChatTable:
except Exception:
return False
def unarchive_all_chats_by_user_id(self, user_id: str) -> bool:
try:
with get_db() as db:
db.query(Chat).filter_by(user_id=user_id).update({"archived": False})
db.commit()
return True
except Exception:
return False
def update_chat_share_id_by_id(
self, id: str, share_id: Optional[str]
) -> Optional[ChatModel]:
@ -360,6 +475,7 @@ class ChatTable:
with get_db() as db:
chat = db.get(Chat, id)
chat.archived = not chat.archived
chat.folder_id = None
chat.updated_at = int(time.time())
db.commit()
db.refresh(chat)
@ -395,7 +511,10 @@ class ChatTable:
order_by = filter.get("order_by")
direction = filter.get("direction")
if order_by and direction and getattr(Chat, order_by):
if order_by and direction:
if not getattr(Chat, order_by, None):
raise ValueError("Invalid order_by field")
if direction.lower() == "asc":
query = query.order_by(getattr(Chat, order_by).asc())
elif direction.lower() == "desc":
@ -456,12 +575,19 @@ class ChatTable:
self,
user_id: str,
include_archived: bool = False,
include_folders: bool = False,
include_pinned: bool = False,
skip: Optional[int] = None,
limit: Optional[int] = None,
) -> list[ChatTitleIdResponse]:
with get_db() as db:
query = db.query(Chat).filter_by(user_id=user_id).filter_by(folder_id=None)
query = query.filter(or_(Chat.pinned == False, Chat.pinned == None))
query = db.query(Chat).filter_by(user_id=user_id)
if not include_folders:
query = query.filter_by(folder_id=None)
if not include_pinned:
query = query.filter(or_(Chat.pinned == False, Chat.pinned == None))
if not include_archived:
query = query.filter_by(archived=False)
@ -506,8 +632,15 @@ class ChatTable:
def get_chat_by_id(self, id: str) -> Optional[ChatModel]:
try:
with get_db() as db:
chat = db.get(Chat, id)
return ChatModel.model_validate(chat)
chat_item = db.get(Chat, id)
if chat_item is None:
return None
if self._sanitize_chat_row(chat_item):
db.commit()
db.refresh(chat_item)
return ChatModel.model_validate(chat_item)
except Exception:
return None
@ -580,7 +713,7 @@ class ChatTable:
"""
Filters chats based on a search query using Python, allowing pagination using skip and limit.
"""
search_text = search_text.lower().strip()
search_text = search_text.replace("\u0000", "").lower().strip()
if not search_text:
return self.get_chat_list_by_user_id(
@ -596,8 +729,45 @@ class ChatTable:
if word.startswith("tag:")
]
# Extract folder names - handle spaces and case insensitivity
folders = Folders.search_folders_by_names(
user_id,
[
word.replace("folder:", "")
for word in search_text_words
if word.startswith("folder:")
],
)
folder_ids = [folder.id for folder in folders]
is_pinned = None
if "pinned:true" in search_text_words:
is_pinned = True
elif "pinned:false" in search_text_words:
is_pinned = False
is_archived = None
if "archived:true" in search_text_words:
is_archived = True
elif "archived:false" in search_text_words:
is_archived = False
is_shared = None
if "shared:true" in search_text_words:
is_shared = True
elif "shared:false" in search_text_words:
is_shared = False
search_text_words = [
word for word in search_text_words if not word.startswith("tag:")
word
for word in search_text_words
if (
not word.startswith("tag:")
and not word.startswith("folder:")
and not word.startswith("pinned:")
and not word.startswith("archived:")
and not word.startswith("shared:")
)
]
search_text = " ".join(search_text_words)
@ -605,30 +775,41 @@ class ChatTable:
with get_db() as db:
query = db.query(Chat).filter(Chat.user_id == user_id)
if not include_archived:
if is_archived is not None:
query = query.filter(Chat.archived == is_archived)
elif not include_archived:
query = query.filter(Chat.archived == False)
if is_pinned is not None:
query = query.filter(Chat.pinned == is_pinned)
if is_shared is not None:
if is_shared:
query = query.filter(Chat.share_id.isnot(None))
else:
query = query.filter(Chat.share_id.is_(None))
if folder_ids:
query = query.filter(Chat.folder_id.in_(folder_ids))
query = query.order_by(Chat.updated_at.desc())
# Check if the database dialect is either 'sqlite' or 'postgresql'
dialect_name = db.bind.dialect.name
if dialect_name == "sqlite":
# SQLite case: using JSON1 extension for JSON searching
sqlite_content_sql = (
"EXISTS ("
" SELECT 1 "
" FROM json_each(Chat.chat, '$.messages') AS message "
" WHERE LOWER(message.value->>'content') LIKE '%' || :content_key || '%'"
")"
)
sqlite_content_clause = text(sqlite_content_sql)
query = query.filter(
(
Chat.title.ilike(
f"%{search_text}%"
) # Case-insensitive search in title
| text(
"""
EXISTS (
SELECT 1
FROM json_each(Chat.chat, '$.messages') AS message
WHERE LOWER(message.value->>'content') LIKE '%' || :search_text || '%'
)
"""
)
).params(search_text=search_text)
or_(
Chat.title.ilike(bindparam("title_key")), sqlite_content_clause
).params(title_key=f"%{search_text}%", content_key=search_text)
)
# Check if there are any tags to filter, it should have all the tags
@ -662,23 +843,32 @@ class ChatTable:
)
elif dialect_name == "postgresql":
# PostgreSQL relies on proper JSON query for search
query = query.filter(
(
Chat.title.ilike(
f"%{search_text}%"
) # Case-insensitive search in title
| text(
"""
EXISTS (
SELECT 1
FROM json_array_elements(Chat.chat->'messages') AS message
WHERE LOWER(message->>'content') LIKE '%' || :search_text || '%'
)
"""
)
).params(search_text=search_text)
# PostgreSQL doesn't allow null bytes in text. We filter those out by checking
# the JSON representation for \u0000 before attempting text extraction
# Safety filter: JSON field must not contain \u0000
query = query.filter(text("Chat.chat::text NOT LIKE '%\\\\u0000%'"))
# Safety filter: title must not contain actual null bytes
query = query.filter(text("Chat.title::text NOT LIKE '%\\x00%'"))
postgres_content_sql = """
EXISTS (
SELECT 1
FROM json_array_elements(Chat.chat->'messages') AS message
WHERE json_typeof(message->'content') = 'string'
AND LOWER(message->>'content') LIKE '%' || :content_key || '%'
)
"""
postgres_content_clause = text(postgres_content_sql)
query = query.filter(
or_(
Chat.title.ilike(bindparam("title_key")),
postgres_content_clause,
)
).params(title_key=f"%{search_text}%", content_key=search_text.lower())
# Check if there are any tags to filter, it should have all the tags
if "none" in tag_ids:
@ -723,7 +913,7 @@ class ChatTable:
return [ChatModel.model_validate(chat) for chat in all_chats]
def get_chats_by_folder_id_and_user_id(
self, folder_id: str, user_id: str
self, folder_id: str, user_id: str, skip: int = 0, limit: int = 60
) -> list[ChatModel]:
with get_db() as db:
query = db.query(Chat).filter_by(folder_id=folder_id, user_id=user_id)
@ -732,6 +922,11 @@ class ChatTable:
query = query.order_by(Chat.updated_at.desc())
if skip:
query = query.offset(skip)
if limit:
query = query.limit(limit)
all_chats = query.all()
return [ChatModel.model_validate(chat) for chat in all_chats]
@ -861,6 +1056,16 @@ class ChatTable:
return count
def count_chats_by_folder_id_and_user_id(self, folder_id: str, user_id: str) -> int:
with get_db() as db:
query = db.query(Chat).filter_by(user_id=user_id)
query = query.filter_by(folder_id=folder_id)
count = query.count()
log.info(f"Count of chats for folder '{folder_id}': {count}")
return count
def delete_tag_by_id_and_user_id_and_tag_name(
self, id: str, user_id: str, tag_name: str
) -> bool:
@ -938,6 +1143,20 @@ class ChatTable:
except Exception:
return False
def move_chats_by_user_id_and_folder_id(
self, user_id: str, folder_id: str, new_folder_id: Optional[str]
) -> bool:
try:
with get_db() as db:
db.query(Chat).filter_by(user_id=user_id, folder_id=folder_id).update(
{"folder_id": new_folder_id}
)
db.commit()
return True
except Exception:
return False
def delete_shared_chats_by_user_id(self, user_id: str) -> bool:
try:
with get_db() as db:

90
backend/open_webui/models/feedbacks.py
View file

@ -4,7 +4,7 @@ import uuid
from typing import Optional
from open_webui.internal.db import Base, get_db
from open_webui.models.chats import Chats
from open_webui.models.users import User
from open_webui.env import SRC_LOG_LEVELS
from pydantic import BaseModel, ConfigDict
@ -21,7 +21,7 @@ log.setLevel(SRC_LOG_LEVELS["MODELS"])
class Feedback(Base):
__tablename__ = "feedback"
id = Column(Text, primary_key=True)
id = Column(Text, primary_key=True, unique=True)
user_id = Column(Text)
version = Column(BigInteger, default=0)
type = Column(Text)
@ -92,6 +92,28 @@ class FeedbackForm(BaseModel):
model_config = ConfigDict(extra="allow")
class UserResponse(BaseModel):
id: str
name: str
email: str
role: str = "pending"
last_active_at: int # timestamp in epoch
updated_at: int # timestamp in epoch
created_at: int # timestamp in epoch
model_config = ConfigDict(from_attributes=True)
class FeedbackUserResponse(FeedbackResponse):
user: Optional[UserResponse] = None
class FeedbackListResponse(BaseModel):
items: list[FeedbackUserResponse]
total: int
class FeedbackTable:
def insert_new_feedback(
self, user_id: str, form_data: FeedbackForm
@ -143,6 +165,70 @@ class FeedbackTable:
except Exception:
return None
def get_feedback_items(
self, filter: dict = {}, skip: int = 0, limit: int = 30
) -> FeedbackListResponse:
with get_db() as db:
query = db.query(Feedback, User).join(User, Feedback.user_id == User.id)
if filter:
order_by = filter.get("order_by")
direction = filter.get("direction")
if order_by == "username":
if direction == "asc":
query = query.order_by(User.name.asc())
else:
query = query.order_by(User.name.desc())
elif order_by == "model_id":
# it's stored in feedback.data['model_id']
if direction == "asc":
query = query.order_by(
Feedback.data["model_id"].as_string().asc()
)
else:
query = query.order_by(
Feedback.data["model_id"].as_string().desc()
)
elif order_by == "rating":
# it's stored in feedback.data['rating']
if direction == "asc":
query = query.order_by(
Feedback.data["rating"].as_string().asc()
)
else:
query = query.order_by(
Feedback.data["rating"].as_string().desc()
)
elif order_by == "updated_at":
if direction == "asc":
query = query.order_by(Feedback.updated_at.asc())
else:
query = query.order_by(Feedback.updated_at.desc())
else:
query = query.order_by(Feedback.created_at.desc())
# Count BEFORE pagination
total = query.count()
if skip:
query = query.offset(skip)
if limit:
query = query.limit(limit)
items = query.all()
feedbacks = []
for feedback, user in items:
feedback_model = FeedbackModel.model_validate(feedback)
user_model = UserResponse.model_validate(user)
feedbacks.append(
FeedbackUserResponse(**feedback_model.model_dump(), user=user_model)
)
return FeedbackListResponse(items=feedbacks, total=total)
def get_all_feedbacks(self) -> list[FeedbackModel]:
with get_db() as db:
return [

58
backend/open_webui/models/files.py
View file

@ -17,7 +17,7 @@ log.setLevel(SRC_LOG_LEVELS["MODELS"])
class File(Base):
__tablename__ = "file"
id = Column(String, primary_key=True)
id = Column(String, primary_key=True, unique=True)
user_id = Column(String)
hash = Column(Text, nullable=True)
@ -82,6 +82,7 @@ class FileModelResponse(BaseModel):
class FileMetadataResponse(BaseModel):
id: str
hash: Optional[str] = None
meta: dict
created_at: int # timestamp in epoch
updated_at: int # timestamp in epoch
@ -97,6 +98,12 @@ class FileForm(BaseModel):
access_control: Optional[dict] = None
class FileUpdateForm(BaseModel):
hash: Optional[str] = None
data: Optional[dict] = None
meta: Optional[dict] = None
class FilesTable:
def insert_new_file(self, user_id: str, form_data: FileForm) -> Optional[FileModel]:
with get_db() as db:
@ -130,12 +137,24 @@ class FilesTable:
except Exception:
return None
def get_file_by_id_and_user_id(self, id: str, user_id: str) -> Optional[FileModel]:
with get_db() as db:
try:
file = db.query(File).filter_by(id=id, user_id=user_id).first()
if file:
return FileModel.model_validate(file)
else:
return None
except Exception:
return None
def get_file_metadata_by_id(self, id: str) -> Optional[FileMetadataResponse]:
with get_db() as db:
try:
file = db.get(File, id)
return FileMetadataResponse(
id=file.id,
hash=file.hash,
meta=file.meta,
created_at=file.created_at,
updated_at=file.updated_at,
@ -147,6 +166,15 @@ class FilesTable:
with get_db() as db:
return [FileModel.model_validate(file) for file in db.query(File).all()]
def check_access_by_user_id(self, id, user_id, permission="write") -> bool:
file = self.get_file_by_id(id)
if not file:
return False
if file.user_id == user_id:
return True
# Implement additional access control logic here as needed
return False
def get_files_by_ids(self, ids: list[str]) -> list[FileModel]:
with get_db() as db:
return [
@ -162,11 +190,14 @@ class FilesTable:
return [
FileMetadataResponse(
id=file.id,
hash=file.hash,
meta=file.meta,
created_at=file.created_at,
updated_at=file.updated_at,
)
for file in db.query(File)
for file in db.query(
File.id, File.hash, File.meta, File.created_at, File.updated_at
)
.filter(File.id.in_(ids))
.order_by(File.updated_at.desc())
.all()
@ -179,6 +210,29 @@ class FilesTable:
for file in db.query(File).filter_by(user_id=user_id).all()
]
def update_file_by_id(
self, id: str, form_data: FileUpdateForm
) -> Optional[FileModel]:
with get_db() as db:
try:
file = db.query(File).filter_by(id=id).first()
if form_data.hash is not None:
file.hash = form_data.hash
if form_data.data is not None:
file.data = {**(file.data if file.data else {}), **form_data.data}
if form_data.meta is not None:
file.meta = {**(file.meta if file.meta else {}), **form_data.meta}
file.updated_at = int(time.time())
db.commit()
return FileModel.model_validate(file)
except Exception as e:
log.exception(f"Error updating file completely by id: {e}")
return None
def update_file_hash_by_id(self, id: str, hash: str) -> Optional[FileModel]:
with get_db() as db:
try:

144
backend/open_webui/models/folders.py
View file

@ -2,14 +2,14 @@ import logging
import time
import uuid
from typing import Optional
import re
from pydantic import BaseModel, ConfigDict
from sqlalchemy import BigInteger, Column, Text, JSON, Boolean, func
from open_webui.internal.db import Base, get_db
from open_webui.models.chats import Chats
from open_webui.env import SRC_LOG_LEVELS
from pydantic import BaseModel, ConfigDict
from sqlalchemy import BigInteger, Column, Text, JSON, Boolean
from open_webui.utils.access_control import get_permissions
log = logging.getLogger(__name__)
@ -23,12 +23,13 @@ log.setLevel(SRC_LOG_LEVELS["MODELS"])
class Folder(Base):
__tablename__ = "folder"
id = Column(Text, primary_key=True)
id = Column(Text, primary_key=True, unique=True)
parent_id = Column(Text, nullable=True)
user_id = Column(Text)
name = Column(Text)
items = Column(JSON, nullable=True)
meta = Column(JSON, nullable=True)
data = Column(JSON, nullable=True)
is_expanded = Column(Boolean, default=False)
created_at = Column(BigInteger)
updated_at = Column(BigInteger)
@ -41,6 +42,7 @@ class FolderModel(BaseModel):
name: str
items: Optional[dict] = None
meta: Optional[dict] = None
data: Optional[dict] = None
is_expanded: bool = False
created_at: int
updated_at: int
@ -48,6 +50,20 @@ class FolderModel(BaseModel):
model_config = ConfigDict(from_attributes=True)
class FolderMetadataResponse(BaseModel):
icon: Optional[str] = None
class FolderNameIdResponse(BaseModel):
id: str
name: str
meta: Optional[FolderMetadataResponse] = None
parent_id: Optional[str] = None
is_expanded: bool = False
created_at: int
updated_at: int
####################
# Forms
####################
@ -55,12 +71,21 @@ class FolderModel(BaseModel):
class FolderForm(BaseModel):
name: str
data: Optional[dict] = None
meta: Optional[dict] = None
model_config = ConfigDict(extra="allow")
class FolderUpdateForm(BaseModel):
name: Optional[str] = None
data: Optional[dict] = None
meta: Optional[dict] = None
model_config = ConfigDict(extra="allow")
class FolderTable:
def insert_new_folder(
self, user_id: str, name: str, parent_id: Optional[str] = None
self, user_id: str, form_data: FolderForm, parent_id: Optional[str] = None
) -> Optional[FolderModel]:
with get_db() as db:
id = str(uuid.uuid4())
@ -68,7 +93,7 @@ class FolderTable:
**{
"id": id,
"user_id": user_id,
"name": name,
**(form_data.model_dump(exclude_unset=True) or {}),
"parent_id": parent_id,
"created_at": int(time.time()),
"updated_at": int(time.time()),
@ -103,7 +128,7 @@ class FolderTable:
def get_children_folders_by_id_and_user_id(
self, id: str, user_id: str
) -> Optional[FolderModel]:
) -> Optional[list[FolderModel]]:
try:
with get_db() as db:
folders = []
@ -187,8 +212,8 @@ class FolderTable:
log.error(f"update_folder: {e}")
return
def update_folder_name_by_id_and_user_id(
self, id: str, user_id: str, name: str
def update_folder_by_id_and_user_id(
self, id: str, user_id: str, form_data: FolderUpdateForm
) -> Optional[FolderModel]:
try:
with get_db() as db:
@ -197,18 +222,35 @@ class FolderTable:
if not folder:
return None
form_data = form_data.model_dump(exclude_unset=True)
existing_folder = (
db.query(Folder)
.filter_by(name=name, parent_id=folder.parent_id, user_id=user_id)
.filter_by(
name=form_data.get("name"),
parent_id=folder.parent_id,
user_id=user_id,
)
.first()
)
if existing_folder:
if existing_folder and existing_folder.id != id:
return None
folder.name = name
folder.updated_at = int(time.time())
folder.name = form_data.get("name", folder.name)
if "data" in form_data:
folder.data = {
**(folder.data or {}),
**form_data["data"],
}
if "meta" in form_data:
folder.meta = {
**(folder.meta or {}),
**form_data["meta"],
}
folder.updated_at = int(time.time())
db.commit()
return FolderModel.model_validate(folder)
@ -236,18 +278,15 @@ class FolderTable:
log.error(f"update_folder: {e}")
return
def delete_folder_by_id_and_user_id(
self, id: str, user_id: str, delete_chats=True
) -> bool:
def delete_folder_by_id_and_user_id(self, id: str, user_id: str) -> list[str]:
try:
folder_ids = []
with get_db() as db:
folder = db.query(Folder).filter_by(id=id, user_id=user_id).first()
if not folder:
return False
return folder_ids
if delete_chats:
# Delete all chats in the folder
Chats.delete_chats_by_user_id_and_folder_id(user_id, folder.id)
folder_ids.append(folder.id)
# Delete all children folders
def delete_children(folder):
@ -255,12 +294,9 @@ class FolderTable:
folder.id, user_id
)
for folder_child in folder_children:
if delete_chats:
Chats.delete_chats_by_user_id_and_folder_id(
user_id, folder_child.id
)
delete_children(folder_child)
folder_ids.append(folder_child.id)
folder = db.query(Folder).filter_by(id=folder_child.id).first()
db.delete(folder)
@ -269,10 +305,62 @@ class FolderTable:
delete_children(folder)
db.delete(folder)
db.commit()
return True
return folder_ids
except Exception as e:
log.error(f"delete_folder: {e}")
return False
return []
def normalize_folder_name(self, name: str) -> str:
# Replace _ and space with a single space, lower case, collapse multiple spaces
name = re.sub(r"[\s_]+", " ", name)
return name.strip().lower()
def search_folders_by_names(
self, user_id: str, queries: list[str]
) -> list[FolderModel]:
"""
Search for folders for a user where the name matches any of the queries, treating _ and space as equivalent, case-insensitive.
"""
normalized_queries = [self.normalize_folder_name(q) for q in queries]
if not normalized_queries:
return []
results = {}
with get_db() as db:
folders = db.query(Folder).filter_by(user_id=user_id).all()
for folder in folders:
if self.normalize_folder_name(folder.name) in normalized_queries:
results[folder.id] = FolderModel.model_validate(folder)
# get children folders
children = self.get_children_folders_by_id_and_user_id(
folder.id, user_id
)
for child in children:
results[child.id] = child
# Return the results as a list
if not results:
return []
else:
results = list(results.values())
return results
def search_folders_by_name_contains(
self, user_id: str, query: str
) -> list[FolderModel]:
"""
Partial match: normalized name contains (as substring) the normalized query.
"""
normalized_query = self.normalize_folder_name(query)
results = []
with get_db() as db:
folders = db.query(Folder).filter_by(user_id=user_id).all()
for folder in folders:
norm_name = self.normalize_folder_name(folder.name)
if normalized_query in norm_name:
results.append(FolderModel.model_validate(folder))
return results
Folders = FolderTable()

99
backend/open_webui/models/functions.py
View file

@ -3,10 +3,10 @@ import time
from typing import Optional
from open_webui.internal.db import Base, JSONField, get_db
from open_webui.models.users import Users
from open_webui.models.users import Users, UserModel
from open_webui.env import SRC_LOG_LEVELS
from pydantic import BaseModel, ConfigDict
from sqlalchemy import BigInteger, Boolean, Column, String, Text
from sqlalchemy import BigInteger, Boolean, Column, String, Text, Index
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["MODELS"])
@ -19,7 +19,7 @@ log.setLevel(SRC_LOG_LEVELS["MODELS"])
class Function(Base):
__tablename__ = "function"
id = Column(String, primary_key=True)
id = Column(String, primary_key=True, unique=True)
user_id = Column(String)
name = Column(Text)
type = Column(Text)
@ -31,10 +31,13 @@ class Function(Base):
updated_at = Column(BigInteger)
created_at = Column(BigInteger)
__table_args__ = (Index("is_global_idx", "is_global"),)
class FunctionMeta(BaseModel):
description: Optional[str] = None
manifest: Optional[dict] = {}
model_config = ConfigDict(extra="allow")
class FunctionModel(BaseModel):
@ -52,11 +55,31 @@ class FunctionModel(BaseModel):
model_config = ConfigDict(from_attributes=True)
class FunctionWithValvesModel(BaseModel):
id: str
user_id: str
name: str
type: str
content: str
meta: FunctionMeta
valves: Optional[dict] = None
is_active: bool = False
is_global: bool = False
updated_at: int # timestamp in epoch
created_at: int # timestamp in epoch
model_config = ConfigDict(from_attributes=True)
####################
# Forms
####################
class FunctionUserResponse(FunctionModel):
user: Optional[UserModel] = None
class FunctionResponse(BaseModel):
id: str
user_id: str
@ -109,8 +132,8 @@ class FunctionsTable:
return None
def sync_functions(
self, user_id: str, functions: list[FunctionModel]
) -> list[FunctionModel]:
self, user_id: str, functions: list[FunctionWithValvesModel]
) -> list[FunctionWithValvesModel]:
# Synchronize functions for a user by updating existing ones, inserting new ones, and removing those that are no longer present.
try:
with get_db() as db:
@ -164,19 +187,48 @@ class FunctionsTable:
except Exception:
return None
def get_functions(self, active_only=False) -> list[FunctionModel]:
def get_functions(
self, active_only=False, include_valves=False
) -> list[FunctionModel | FunctionWithValvesModel]:
with get_db() as db:
if active_only:
functions = db.query(Function).filter_by(is_active=True).all()
else:
functions = db.query(Function).all()
if include_valves:
return [
FunctionModel.model_validate(function)
for function in db.query(Function).filter_by(is_active=True).all()
FunctionWithValvesModel.model_validate(function)
for function in functions
]
else:
return [
FunctionModel.model_validate(function)
for function in db.query(Function).all()
FunctionModel.model_validate(function) for function in functions
]
def get_function_list(self) -> list[FunctionUserResponse]:
with get_db() as db:
functions = db.query(Function).order_by(Function.updated_at.desc()).all()
user_ids = list(set(func.user_id for func in functions))
users = Users.get_users_by_user_ids(user_ids) if user_ids else []
users_dict = {user.id: user for user in users}
return [
FunctionUserResponse.model_validate(
{
**FunctionModel.model_validate(func).model_dump(),
"user": (
users_dict.get(func.user_id).model_dump()
if func.user_id in users_dict
else None
),
}
)
for func in functions
]
def get_functions_by_type(
self, type: str, active_only=False
) -> list[FunctionModel]:
@ -235,6 +287,29 @@ class FunctionsTable:
except Exception:
return None
def update_function_metadata_by_id(
self, id: str, metadata: dict
) -> Optional[FunctionModel]:
with get_db() as db:
try:
function = db.get(Function, id)
if function:
if function.meta:
function.meta = {**function.meta, **metadata}
else:
function.meta = metadata
function.updated_at = int(time.time())
db.commit()
db.refresh(function)
return self.get_function_by_id(id)
else:
return None
except Exception as e:
log.exception(f"Error updating function metadata by id {id}: {e}")
return None
def get_user_valves_by_id_and_user_id(
self, id: str, user_id: str
) -> Optional[dict]:
@ -250,9 +325,7 @@ class FunctionsTable:
return user_settings["functions"]["valves"].get(id, {})
except Exception as e:
log.exception(
f"Error getting user values by id {id} and user id {user_id}: {e}"
)
log.exception(f"Error getting user values by id {id} and user id {user_id}")
return None
def update_user_valves_by_id_and_user_id(

268
backend/open_webui/models/groups.py
View file

@ -11,7 +11,7 @@ from open_webui.models.files import FileMetadataResponse
from pydantic import BaseModel, ConfigDict
from sqlalchemy import BigInteger, Column, String, Text, JSON, func
from sqlalchemy import BigInteger, Column, String, Text, JSON, func, ForeignKey
log = logging.getLogger(__name__)
@ -35,7 +35,6 @@ class Group(Base):
meta = Column(JSON, nullable=True)
permissions = Column(JSON, nullable=True)
user_ids = Column(JSON, nullable=True)
created_at = Column(BigInteger)
updated_at = Column(BigInteger)
@ -53,12 +52,33 @@ class GroupModel(BaseModel):
meta: Optional[dict] = None
permissions: Optional[dict] = None
user_ids: list[str] = []
created_at: int # timestamp in epoch
updated_at: int # timestamp in epoch
class GroupMember(Base):
__tablename__ = "group_member"
id = Column(Text, unique=True, primary_key=True)
group_id = Column(
Text,
ForeignKey("group.id", ondelete="CASCADE"),
nullable=False,
)
user_id = Column(Text, nullable=False)
created_at = Column(BigInteger, nullable=True)
updated_at = Column(BigInteger, nullable=True)
class GroupMemberModel(BaseModel):
id: str
group_id: str
user_id: str
created_at: Optional[int] = None # timestamp in epoch
updated_at: Optional[int] = None # timestamp in epoch
####################
# Forms
####################
@ -72,7 +92,7 @@ class GroupResponse(BaseModel):
permissions: Optional[dict] = None
data: Optional[dict] = None
meta: Optional[dict] = None
user_ids: list[str] = []
member_count: Optional[int] = None
created_at: int # timestamp in epoch
updated_at: int # timestamp in epoch
@ -81,10 +101,15 @@ class GroupForm(BaseModel):
name: str
description: str
permissions: Optional[dict] = None
data: Optional[dict] = None
class UserIdsForm(BaseModel):
user_ids: Optional[list[str]] = None
class GroupUpdateForm(GroupForm):
user_ids: Optional[list[str]] = None
pass
class GroupTable:
@ -127,12 +152,8 @@ class GroupTable:
return [
GroupModel.model_validate(group)
for group in db.query(Group)
.filter(
func.json_array_length(Group.user_ids) > 0
) # Ensure array exists
.filter(
Group.user_ids.cast(String).like(f'%"{user_id}"%')
) # String-based check
.join(GroupMember, GroupMember.group_id == Group.id)
.filter(GroupMember.user_id == user_id)
.order_by(Group.updated_at.desc())
.all()
]
@ -145,12 +166,63 @@ class GroupTable:
except Exception:
return None
def get_group_user_ids_by_id(self, id: str) -> Optional[str]:
group = self.get_group_by_id(id)
if group:
return group.user_ids
else:
return None
def get_group_user_ids_by_id(self, id: str) -> Optional[list[str]]:
with get_db() as db:
members = (
db.query(GroupMember.user_id).filter(GroupMember.group_id == id).all()
)
if not members:
return None
return [m[0] for m in members]
def get_group_user_ids_by_ids(self, group_ids: list[str]) -> dict[str, list[str]]:
with get_db() as db:
members = (
db.query(GroupMember.group_id, GroupMember.user_id)
.filter(GroupMember.group_id.in_(group_ids))
.all()
)
group_user_ids: dict[str, list[str]] = {
group_id: [] for group_id in group_ids
}
for group_id, user_id in members:
group_user_ids[group_id].append(user_id)
return group_user_ids
def set_group_user_ids_by_id(self, group_id: str, user_ids: list[str]) -> None:
with get_db() as db:
# Delete existing members
db.query(GroupMember).filter(GroupMember.group_id == group_id).delete()
# Insert new members
now = int(time.time())
new_members = [
GroupMember(
id=str(uuid.uuid4()),
group_id=group_id,
user_id=user_id,
created_at=now,
updated_at=now,
)
for user_id in user_ids
]
db.add_all(new_members)
db.commit()
def get_group_member_count_by_id(self, id: str) -> int:
with get_db() as db:
count = (
db.query(func.count(GroupMember.user_id))
.filter(GroupMember.group_id == id)
.scalar()
)
return count if count else 0
def update_group_by_id(
self, id: str, form_data: GroupUpdateForm, overwrite: bool = False
@ -191,20 +263,29 @@ class GroupTable:
def remove_user_from_all_groups(self, user_id: str) -> bool:
with get_db() as db:
try:
groups = self.get_groups_by_member_id(user_id)
# Find all groups the user belongs to
groups = (
db.query(Group)
.join(GroupMember, GroupMember.group_id == Group.id)
.filter(GroupMember.user_id == user_id)
.all()
)
# Remove the user from each group
for group in groups:
group.user_ids.remove(user_id)
db.query(Group).filter_by(id=group.id).update(
{
"user_ids": group.user_ids,
"updated_at": int(time.time()),
}
)
db.commit()
db.query(GroupMember).filter(
GroupMember.group_id == group.id, GroupMember.user_id == user_id
).delete()
db.query(Group).filter_by(id=group.id).update(
{"updated_at": int(time.time())}
)
db.commit()
return True
except Exception:
db.rollback()
return False
def create_groups_by_group_names(
@ -242,38 +323,129 @@ class GroupTable:
def sync_groups_by_group_names(self, user_id: str, group_names: list[str]) -> bool:
with get_db() as db:
try:
groups = db.query(Group).filter(Group.name.in_(group_names)).all()
group_ids = [group.id for group in groups]
now = int(time.time())
# Remove user from groups not in the new list
existing_groups = self.get_groups_by_member_id(user_id)
# 1. Groups that SHOULD contain the user
target_groups = (
db.query(Group).filter(Group.name.in_(group_names)).all()
)
target_group_ids = {g.id for g in target_groups}
for group in existing_groups:
if group.id not in group_ids:
group.user_ids.remove(user_id)
db.query(Group).filter_by(id=group.id).update(
{
"user_ids": group.user_ids,
"updated_at": int(time.time()),
}
# 2. Groups the user is CURRENTLY in
existing_group_ids = {
g.id
for g in db.query(Group)
.join(GroupMember, GroupMember.group_id == Group.id)
.filter(GroupMember.user_id == user_id)
.all()
}
# 3. Determine adds + removals
groups_to_add = target_group_ids - existing_group_ids
groups_to_remove = existing_group_ids - target_group_ids
# 4. Remove in one bulk delete
if groups_to_remove:
db.query(GroupMember).filter(
GroupMember.user_id == user_id,
GroupMember.group_id.in_(groups_to_remove),
).delete(synchronize_session=False)
db.query(Group).filter(Group.id.in_(groups_to_remove)).update(
{"updated_at": now}, synchronize_session=False
)
# 5. Bulk insert missing memberships
for group_id in groups_to_add:
db.add(
GroupMember(
id=str(uuid.uuid4()),
group_id=group_id,
user_id=user_id,
created_at=now,
updated_at=now,
)
)
# Add user to new groups
for group in groups:
if user_id not in group.user_ids:
group.user_ids.append(user_id)
db.query(Group).filter_by(id=group.id).update(
{
"user_ids": group.user_ids,
"updated_at": int(time.time()),
}
)
if groups_to_add:
db.query(Group).filter(Group.id.in_(groups_to_add)).update(
{"updated_at": now}, synchronize_session=False
)
db.commit()
return True
except Exception as e:
log.exception(e)
db.rollback()
return False
def add_users_to_group(
self, id: str, user_ids: Optional[list[str]] = None
) -> Optional[GroupModel]:
try:
with get_db() as db:
group = db.query(Group).filter_by(id=id).first()
if not group:
return None
now = int(time.time())
for user_id in user_ids or []:
try:
db.add(
GroupMember(
id=str(uuid.uuid4()),
group_id=id,
user_id=user_id,
created_at=now,
updated_at=now,
)
)
db.flush() # Detect unique constraint violation early
except Exception:
db.rollback() # Clear failed INSERT
db.begin() # Start a new transaction
continue # Duplicate → ignore
group.updated_at = now
db.commit()
db.refresh(group)
return GroupModel.model_validate(group)
except Exception as e:
log.exception(e)
return None
def remove_users_from_group(
self, id: str, user_ids: Optional[list[str]] = None
) -> Optional[GroupModel]:
try:
with get_db() as db:
group = db.query(Group).filter_by(id=id).first()
if not group:
return None
if not user_ids:
return GroupModel.model_validate(group)
# Remove each user from group_member
for user_id in user_ids:
db.query(GroupMember).filter(
GroupMember.group_id == id, GroupMember.user_id == user_id
).delete()
# Update group timestamp
group.updated_at = int(time.time())
db.commit()
db.refresh(group)
return GroupModel.model_validate(group)
except Exception as e:
log.exception(e)
return None
Groups = GroupTable()

30
backend/open_webui/models/knowledge.py
View file

@ -8,6 +8,7 @@ from open_webui.internal.db import Base, get_db
from open_webui.env import SRC_LOG_LEVELS
from open_webui.models.files import FileMetadataResponse
from open_webui.models.groups import Groups
from open_webui.models.users import Users, UserResponse
@ -128,11 +129,18 @@ class KnowledgeTable:
def get_knowledge_bases(self) -> list[KnowledgeUserModel]:
with get_db() as db:
knowledge_bases = []
for knowledge in (
all_knowledge = (
db.query(Knowledge).order_by(Knowledge.updated_at.desc()).all()
):
user = Users.get_user_by_id(knowledge.user_id)
)
user_ids = list(set(knowledge.user_id for knowledge in all_knowledge))
users = Users.get_users_by_user_ids(user_ids) if user_ids else []
users_dict = {user.id: user for user in users}
knowledge_bases = []
for knowledge in all_knowledge:
user = users_dict.get(knowledge.user_id)
knowledge_bases.append(
KnowledgeUserModel.model_validate(
{
@ -143,15 +151,27 @@ class KnowledgeTable:
)
return knowledge_bases
def check_access_by_user_id(self, id, user_id, permission="write") -> bool:
knowledge = self.get_knowledge_by_id(id)
if not knowledge:
return False
if knowledge.user_id == user_id:
return True
user_group_ids = {group.id for group in Groups.get_groups_by_member_id(user_id)}
return has_access(user_id, permission, knowledge.access_control, user_group_ids)
def get_knowledge_bases_by_user_id(
self, user_id: str, permission: str = "write"
) -> list[KnowledgeUserModel]:
knowledge_bases = self.get_knowledge_bases()
user_group_ids = {group.id for group in Groups.get_groups_by_member_id(user_id)}
return [
knowledge_base
for knowledge_base in knowledge_bases
if knowledge_base.user_id == user_id
or has_access(user_id, permission, knowledge_base.access_control)
or has_access(
user_id, permission, knowledge_base.access_control, user_group_ids
)
]
def get_knowledge_by_id(self, id: str) -> Optional[KnowledgeModel]:

19
backend/open_webui/models/memories.py
View file

@ -14,7 +14,7 @@ from sqlalchemy import BigInteger, Column, String, Text
class Memory(Base):
__tablename__ = "memory"
id = Column(String, primary_key=True)
id = Column(String, primary_key=True, unique=True)
user_id = Column(String)
content = Column(Text)
updated_at = Column(BigInteger)
@ -71,9 +71,13 @@ class MemoriesTable:
) -> Optional[MemoryModel]:
with get_db() as db:
try:
db.query(Memory).filter_by(id=id, user_id=user_id).update(
{"content": content, "updated_at": int(time.time())}
)
memory = db.get(Memory, id)
if not memory or memory.user_id != user_id:
return None
memory.content = content
memory.updated_at = int(time.time())
db.commit()
return self.get_memory_by_id(id)
except Exception:
@ -127,7 +131,12 @@ class MemoriesTable:
def delete_memory_by_id_and_user_id(self, id: str, user_id: str) -> bool:
with get_db() as db:
try:
db.query(Memory).filter_by(id=id, user_id=user_id).delete()
memory = db.get(Memory, id)
if not memory or memory.user_id != user_id:
return None
# Delete the memory
db.delete(memory)
db.commit()
return True

123
backend/open_webui/models/messages.py
View file

@ -5,6 +5,7 @@ from typing import Optional
from open_webui.internal.db import Base, get_db
from open_webui.models.tags import TagModel, Tag, Tags
from open_webui.models.users import Users, UserNameResponse
from pydantic import BaseModel, ConfigDict
@ -19,7 +20,7 @@ from sqlalchemy.sql import exists
class MessageReaction(Base):
__tablename__ = "message_reaction"
id = Column(Text, primary_key=True)
id = Column(Text, primary_key=True, unique=True)
user_id = Column(Text)
message_id = Column(Text)
name = Column(Text)
@ -43,6 +44,7 @@ class Message(Base):
user_id = Column(Text)
channel_id = Column(Text, nullable=True)
reply_to_id = Column(Text, nullable=True)
parent_id = Column(Text, nullable=True)
content = Column(Text)
@ -60,6 +62,7 @@ class MessageModel(BaseModel):
user_id: str
channel_id: Optional[str] = None
reply_to_id: Optional[str] = None
parent_id: Optional[str] = None
content: str
@ -77,6 +80,7 @@ class MessageModel(BaseModel):
class MessageForm(BaseModel):
content: str
reply_to_id: Optional[str] = None
parent_id: Optional[str] = None
data: Optional[dict] = None
meta: Optional[dict] = None
@ -88,7 +92,15 @@ class Reactions(BaseModel):
count: int
class MessageResponse(MessageModel):
class MessageUserResponse(MessageModel):
user: Optional[UserNameResponse] = None
class MessageReplyToResponse(MessageUserResponse):
reply_to_message: Optional[MessageUserResponse] = None
class MessageResponse(MessageReplyToResponse):
latest_reply_at: Optional[int]
reply_count: int
reactions: list[Reactions]
@ -107,6 +119,7 @@ class MessageTable:
"id": id,
"user_id": user_id,
"channel_id": channel_id,
"reply_to_id": form_data.reply_to_id,
"parent_id": form_data.parent_id,
"content": form_data.content,
"data": form_data.data,
@ -128,19 +141,32 @@ class MessageTable:
if not message:
return None
reactions = self.get_reactions_by_message_id(id)
replies = self.get_replies_by_message_id(id)
reply_to_message = (
self.get_message_by_id(message.reply_to_id)
if message.reply_to_id
else None
)
return MessageResponse(
**{
reactions = self.get_reactions_by_message_id(id)
thread_replies = self.get_thread_replies_by_message_id(id)
user = Users.get_user_by_id(message.user_id)
return MessageResponse.model_validate(
{
**MessageModel.model_validate(message).model_dump(),
"latest_reply_at": replies[0].created_at if replies else None,
"reply_count": len(replies),
"user": user.model_dump() if user else None,
"reply_to_message": (
reply_to_message.model_dump() if reply_to_message else None
),
"latest_reply_at": (
thread_replies[0].created_at if thread_replies else None
),
"reply_count": len(thread_replies),
"reactions": reactions,
}
)
def get_replies_by_message_id(self, id: str) -> list[MessageModel]:
def get_thread_replies_by_message_id(self, id: str) -> list[MessageReplyToResponse]:
with get_db() as db:
all_messages = (
db.query(Message)
@ -148,7 +174,27 @@ class MessageTable:
.order_by(Message.created_at.desc())
.all()
)
return [MessageModel.model_validate(message) for message in all_messages]
messages = []
for message in all_messages:
reply_to_message = (
self.get_message_by_id(message.reply_to_id)
if message.reply_to_id
else None
)
messages.append(
MessageReplyToResponse.model_validate(
{
**MessageModel.model_validate(message).model_dump(),
"reply_to_message": (
reply_to_message.model_dump()
if reply_to_message
else None
),
}
)
)
return messages
def get_reply_user_ids_by_message_id(self, id: str) -> list[str]:
with get_db() as db:
@ -159,7 +205,7 @@ class MessageTable:
def get_messages_by_channel_id(
self, channel_id: str, skip: int = 0, limit: int = 50
) -> list[MessageModel]:
) -> list[MessageReplyToResponse]:
with get_db() as db:
all_messages = (
db.query(Message)
@ -169,11 +215,31 @@ class MessageTable:
.limit(limit)
.all()
)
return [MessageModel.model_validate(message) for message in all_messages]
messages = []
for message in all_messages:
reply_to_message = (
self.get_message_by_id(message.reply_to_id)
if message.reply_to_id
else None
)
messages.append(
MessageReplyToResponse.model_validate(
{
**MessageModel.model_validate(message).model_dump(),
"reply_to_message": (
reply_to_message.model_dump()
if reply_to_message
else None
),
}
)
)
return messages
def get_messages_by_parent_id(
self, channel_id: str, parent_id: str, skip: int = 0, limit: int = 50
) -> list[MessageModel]:
) -> list[MessageReplyToResponse]:
with get_db() as db:
message = db.get(Message, parent_id)
@ -193,7 +259,26 @@ class MessageTable:
if len(all_messages) < limit:
all_messages.append(message)
return [MessageModel.model_validate(message) for message in all_messages]
messages = []
for message in all_messages:
reply_to_message = (
self.get_message_by_id(message.reply_to_id)
if message.reply_to_id
else None
)
messages.append(
MessageReplyToResponse.model_validate(
{
**MessageModel.model_validate(message).model_dump(),
"reply_to_message": (
reply_to_message.model_dump()
if reply_to_message
else None
),
}
)
)
return messages
def update_message_by_id(
self, id: str, form_data: MessageForm
@ -201,8 +286,14 @@ class MessageTable:
with get_db() as db:
message = db.get(Message, id)
message.content = form_data.content
message.data = form_data.data
message.meta = form_data.meta
message.data = {
**(message.data if message.data else {}),
**(form_data.data if form_data.data else {}),
}
message.meta = {
**(message.meta if message.meta else {}),
**(form_data.meta if form_data.meta else {}),
}
message.updated_at = int(time.time_ns())
db.commit()
db.refresh(message)

159
backend/open_webui/models/models.py
View file

@ -5,12 +5,13 @@ from typing import Optional
from open_webui.internal.db import Base, JSONField, get_db
from open_webui.env import SRC_LOG_LEVELS
from open_webui.models.users import Users, UserResponse
from open_webui.models.groups import Groups
from open_webui.models.users import User, UserModel, Users, UserResponse
from pydantic import BaseModel, ConfigDict
from sqlalchemy import or_, and_, func
from sqlalchemy import String, cast, or_, and_, func
from sqlalchemy.dialects import postgresql, sqlite
from sqlalchemy import BigInteger, Column, Text, JSON, Boolean
@ -132,6 +133,11 @@ class ModelResponse(ModelModel):
pass
class ModelListResponse(BaseModel):
items: list[ModelUserResponse]
total: int
class ModelForm(BaseModel):
id: str
base_model_id: Optional[str] = None
@ -175,9 +181,16 @@ class ModelsTable:
def get_models(self) -> list[ModelUserResponse]:
with get_db() as db:
all_models = db.query(Model).filter(Model.base_model_id != None).all()
user_ids = list(set(model.user_id for model in all_models))
users = Users.get_users_by_user_ids(user_ids) if user_ids else []
users_dict = {user.id: user for user in users}
models = []
for model in db.query(Model).filter(Model.base_model_id != None).all():
user = Users.get_user_by_id(model.user_id)
for model in all_models:
user = users_dict.get(model.user_id)
models.append(
ModelUserResponse.model_validate(
{
@ -199,13 +212,97 @@ class ModelsTable:
self, user_id: str, permission: str = "write"
) -> list[ModelUserResponse]:
models = self.get_models()
user_group_ids = {group.id for group in Groups.get_groups_by_member_id(user_id)}
return [
model
for model in models
if model.user_id == user_id
or has_access(user_id, permission, model.access_control)
or has_access(user_id, permission, model.access_control, user_group_ids)
]
def search_models(
self, user_id: str, filter: dict = {}, skip: int = 0, limit: int = 30
) -> ModelListResponse:
with get_db() as db:
# Join GroupMember so we can order by group_id when requested
query = db.query(Model, User).outerjoin(User, User.id == Model.user_id)
query = query.filter(Model.base_model_id != None)
if filter:
query_key = filter.get("query")
if query_key:
query = query.filter(
or_(
Model.name.ilike(f"%{query_key}%"),
Model.base_model_id.ilike(f"%{query_key}%"),
)
)
if filter.get("user_id"):
query = query.filter(Model.user_id == filter.get("user_id"))
view_option = filter.get("view_option")
if view_option == "created":
query = query.filter(Model.user_id == user_id)
elif view_option == "shared":
query = query.filter(Model.user_id != user_id)
tag = filter.get("tag")
if tag:
# TODO: This is a simple implementation and should be improved for performance
like_pattern = f'%"{tag.lower()}"%' # `"tag"` inside JSON array
meta_text = func.lower(cast(Model.meta, String))
query = query.filter(meta_text.like(like_pattern))
order_by = filter.get("order_by")
direction = filter.get("direction")
if order_by == "name":
if direction == "asc":
query = query.order_by(Model.name.asc())
else:
query = query.order_by(Model.name.desc())
elif order_by == "created_at":
if direction == "asc":
query = query.order_by(Model.created_at.asc())
else:
query = query.order_by(Model.created_at.desc())
elif order_by == "updated_at":
if direction == "asc":
query = query.order_by(Model.updated_at.asc())
else:
query = query.order_by(Model.updated_at.desc())
else:
query = query.order_by(Model.created_at.desc())
# Count BEFORE pagination
total = query.count()
if skip:
query = query.offset(skip)
if limit:
query = query.limit(limit)
items = query.all()
models = []
for model, user in items:
models.append(
ModelUserResponse(
**ModelModel.model_validate(model).model_dump(),
user=(
UserResponse(**UserModel.model_validate(user).model_dump())
if user
else None
),
)
)
return ModelListResponse(items=models, total=total)
def get_model_by_id(self, id: str) -> Optional[ModelModel]:
try:
with get_db() as db:
@ -235,11 +332,9 @@ class ModelsTable:
try:
with get_db() as db:
# update only the fields that are present in the model
result = (
db.query(Model)
.filter_by(id=id)
.update(model.model_dump(exclude={"id"}))
)
data = model.model_dump(exclude={"id"})
result = db.query(Model).filter_by(id=id).update(data)
db.commit()
model = db.get(Model, id)
@ -269,5 +364,49 @@ class ModelsTable:
except Exception:
return False
def sync_models(self, user_id: str, models: list[ModelModel]) -> list[ModelModel]:
try:
with get_db() as db:
# Get existing models
existing_models = db.query(Model).all()
existing_ids = {model.id for model in existing_models}
# Prepare a set of new model IDs
new_model_ids = {model.id for model in models}
# Update or insert models
for model in models:
if model.id in existing_ids:
db.query(Model).filter_by(id=model.id).update(
{
**model.model_dump(),
"user_id": user_id,
"updated_at": int(time.time()),
}
)
else:
new_model = Model(
**{
**model.model_dump(),
"user_id": user_id,
"updated_at": int(time.time()),
}
)
db.add(new_model)
# Remove models that are no longer present
for model in existing_models:
if model.id not in new_model_ids:
db.delete(model)
db.commit()
return [
ModelModel.model_validate(model) for model in db.query(Model).all()
]
except Exception as e:
log.exception(f"Error syncing models for user {user_id}: {e}")
return []
Models = ModelsTable()

113
backend/open_webui/models/notes.py
View file

@ -2,8 +2,10 @@ import json
import time
import uuid
from typing import Optional
from functools import lru_cache
from open_webui.internal.db import Base, get_db
from open_webui.models.groups import Groups
from open_webui.utils.access_control import has_access
from open_webui.models.users import Users, UserResponse
@ -62,6 +64,13 @@ class NoteForm(BaseModel):
access_control: Optional[dict] = None
class NoteUpdateForm(BaseModel):
title: Optional[str] = None
data: Optional[dict] = None
meta: Optional[dict] = None
access_control: Optional[dict] = None
class NoteUserResponse(NoteModel):
user: Optional[UserResponse] = None
@ -89,37 +98,111 @@ class NoteTable:
db.commit()
return note
def get_notes(self) -> list[NoteModel]:
def get_notes(
self, skip: Optional[int] = None, limit: Optional[int] = None
) -> list[NoteModel]:
with get_db() as db:
notes = db.query(Note).order_by(Note.updated_at.desc()).all()
query = db.query(Note).order_by(Note.updated_at.desc())
if skip is not None:
query = query.offset(skip)
if limit is not None:
query = query.limit(limit)
notes = query.all()
return [NoteModel.model_validate(note) for note in notes]
def get_notes_by_user_id(
self, user_id: str, permission: str = "write"
self,
user_id: str,
skip: Optional[int] = None,
limit: Optional[int] = None,
) -> list[NoteModel]:
notes = self.get_notes()
return [
note
for note in notes
if note.user_id == user_id
or has_access(user_id, permission, note.access_control)
]
with get_db() as db:
query = db.query(Note).filter(Note.user_id == user_id)
query = query.order_by(Note.updated_at.desc())
if skip is not None:
query = query.offset(skip)
if limit is not None:
query = query.limit(limit)
notes = query.all()
return [NoteModel.model_validate(note) for note in notes]
def get_notes_by_permission(
self,
user_id: str,
permission: str = "write",
skip: Optional[int] = None,
limit: Optional[int] = None,
) -> list[NoteModel]:
with get_db() as db:
user_groups = Groups.get_groups_by_member_id(user_id)
user_group_ids = {group.id for group in user_groups}
# Order newest-first. We stream to keep memory usage low.
query = (
db.query(Note)
.order_by(Note.updated_at.desc())
.execution_options(stream_results=True)
.yield_per(256)
)
results: list[NoteModel] = []
n_skipped = 0
for note in query:
# Fast-pass #1: owner
if note.user_id == user_id:
permitted = True
# Fast-pass #2: public/open
elif note.access_control is None:
# Technically this should mean public access for both read and write, but we'll only do read for now
# We might want to change this behavior later
permitted = permission == "read"
else:
permitted = has_access(
user_id, permission, note.access_control, user_group_ids
)
if not permitted:
continue
# Apply skip AFTER permission filtering so it counts only accessible notes
if skip and n_skipped < skip:
n_skipped += 1
continue
results.append(NoteModel.model_validate(note))
if limit is not None and len(results) >= limit:
break
return results
def get_note_by_id(self, id: str) -> Optional[NoteModel]:
with get_db() as db:
note = db.query(Note).filter(Note.id == id).first()
return NoteModel.model_validate(note) if note else None
def update_note_by_id(self, id: str, form_data: NoteForm) -> Optional[NoteModel]:
def update_note_by_id(
self, id: str, form_data: NoteUpdateForm
) -> Optional[NoteModel]:
with get_db() as db:
note = db.query(Note).filter(Note.id == id).first()
if not note:
return None
note.title = form_data.title
note.data = form_data.data
note.meta = form_data.meta
note.access_control = form_data.access_control
form_data = form_data.model_dump(exclude_unset=True)
if "title" in form_data:
note.title = form_data["title"]
if "data" in form_data:
note.data = {**note.data, **form_data["data"]}
if "meta" in form_data:
note.meta = {**note.meta, **form_data["meta"]}
if "access_control" in form_data:
note.access_control = form_data["access_control"]
note.updated_at = int(time.time_ns())
db.commit()

277
backend/open_webui/models/oauth_sessions.py Normal file
View file

@ -0,0 +1,277 @@
import time
import logging
import uuid
from typing import Optional, List
import base64
import hashlib
import json
from cryptography.fernet import Fernet
from open_webui.internal.db import Base, get_db
from open_webui.env import SRC_LOG_LEVELS, OAUTH_SESSION_TOKEN_ENCRYPTION_KEY
from pydantic import BaseModel, ConfigDict
from sqlalchemy import BigInteger, Column, String, Text, Index
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["MODELS"])
####################
# DB MODEL
####################
class OAuthSession(Base):
__tablename__ = "oauth_session"
id = Column(Text, primary_key=True)
user_id = Column(Text, nullable=False)
provider = Column(Text, nullable=False)
token = Column(
Text, nullable=False
) # JSON with access_token, id_token, refresh_token
expires_at = Column(BigInteger, nullable=False)
created_at = Column(BigInteger, nullable=False)
updated_at = Column(BigInteger, nullable=False)
# Add indexes for better performance
__table_args__ = (
Index("idx_oauth_session_user_id", "user_id"),
Index("idx_oauth_session_expires_at", "expires_at"),
Index("idx_oauth_session_user_provider", "user_id", "provider"),
)
class OAuthSessionModel(BaseModel):
id: str
user_id: str
provider: str
token: dict
expires_at: int # timestamp in epoch
created_at: int # timestamp in epoch
updated_at: int # timestamp in epoch
model_config = ConfigDict(from_attributes=True)
####################
# Forms
####################
class OAuthSessionResponse(BaseModel):
id: str
user_id: str
provider: str
expires_at: int
class OAuthSessionTable:
def __init__(self):
self.encryption_key = OAUTH_SESSION_TOKEN_ENCRYPTION_KEY
if not self.encryption_key:
raise Exception("OAUTH_SESSION_TOKEN_ENCRYPTION_KEY is not set")
# check if encryption key is in the right format for Fernet (32 url-safe base64-encoded bytes)
if len(self.encryption_key) != 44:
key_bytes = hashlib.sha256(self.encryption_key.encode()).digest()
self.encryption_key = base64.urlsafe_b64encode(key_bytes)
else:
self.encryption_key = self.encryption_key.encode()
try:
self.fernet = Fernet(self.encryption_key)
except Exception as e:
log.error(f"Error initializing Fernet with provided key: {e}")
raise
def _encrypt_token(self, token) -> str:
"""Encrypt OAuth tokens for storage"""
try:
token_json = json.dumps(token)
encrypted = self.fernet.encrypt(token_json.encode()).decode()
return encrypted
except Exception as e:
log.error(f"Error encrypting tokens: {e}")
raise
def _decrypt_token(self, token: str):
"""Decrypt OAuth tokens from storage"""
try:
decrypted = self.fernet.decrypt(token.encode()).decode()
return json.loads(decrypted)
except Exception as e:
log.error(f"Error decrypting tokens: {e}")
raise
def create_session(
self,
user_id: str,
provider: str,
token: dict,
) -> Optional[OAuthSessionModel]:
"""Create a new OAuth session"""
try:
with get_db() as db:
current_time = int(time.time())
id = str(uuid.uuid4())
result = OAuthSession(
**{
"id": id,
"user_id": user_id,
"provider": provider,
"token": self._encrypt_token(token),
"expires_at": token.get("expires_at"),
"created_at": current_time,
"updated_at": current_time,
}
)
db.add(result)
db.commit()
db.refresh(result)
if result:
result.token = token # Return decrypted token
return OAuthSessionModel.model_validate(result)
else:
return None
except Exception as e:
log.error(f"Error creating OAuth session: {e}")
return None
def get_session_by_id(self, session_id: str) -> Optional[OAuthSessionModel]:
"""Get OAuth session by ID"""
try:
with get_db() as db:
session = db.query(OAuthSession).filter_by(id=session_id).first()
if session:
session.token = self._decrypt_token(session.token)
return OAuthSessionModel.model_validate(session)
return None
except Exception as e:
log.error(f"Error getting OAuth session by ID: {e}")
return None
def get_session_by_id_and_user_id(
self, session_id: str, user_id: str
) -> Optional[OAuthSessionModel]:
"""Get OAuth session by ID and user ID"""
try:
with get_db() as db:
session = (
db.query(OAuthSession)
.filter_by(id=session_id, user_id=user_id)
.first()
)
if session:
session.token = self._decrypt_token(session.token)
return OAuthSessionModel.model_validate(session)
return None
except Exception as e:
log.error(f"Error getting OAuth session by ID: {e}")
return None
def get_session_by_provider_and_user_id(
self, provider: str, user_id: str
) -> Optional[OAuthSessionModel]:
"""Get OAuth session by provider and user ID"""
try:
with get_db() as db:
session = (
db.query(OAuthSession)
.filter_by(provider=provider, user_id=user_id)
.first()
)
if session:
session.token = self._decrypt_token(session.token)
return OAuthSessionModel.model_validate(session)
return None
except Exception as e:
log.error(f"Error getting OAuth session by provider and user ID: {e}")
return None
def get_sessions_by_user_id(self, user_id: str) -> List[OAuthSessionModel]:
"""Get all OAuth sessions for a user"""
try:
with get_db() as db:
sessions = db.query(OAuthSession).filter_by(user_id=user_id).all()
results = []
for session in sessions:
session.token = self._decrypt_token(session.token)
results.append(OAuthSessionModel.model_validate(session))
return results
except Exception as e:
log.error(f"Error getting OAuth sessions by user ID: {e}")
return []
def update_session_by_id(
self, session_id: str, token: dict
) -> Optional[OAuthSessionModel]:
"""Update OAuth session tokens"""
try:
with get_db() as db:
current_time = int(time.time())
db.query(OAuthSession).filter_by(id=session_id).update(
{
"token": self._encrypt_token(token),
"expires_at": token.get("expires_at"),
"updated_at": current_time,
}
)
db.commit()
session = db.query(OAuthSession).filter_by(id=session_id).first()
if session:
session.token = self._decrypt_token(session.token)
return OAuthSessionModel.model_validate(session)
return None
except Exception as e:
log.error(f"Error updating OAuth session tokens: {e}")
return None
def delete_session_by_id(self, session_id: str) -> bool:
"""Delete an OAuth session"""
try:
with get_db() as db:
result = db.query(OAuthSession).filter_by(id=session_id).delete()
db.commit()
return result > 0
except Exception as e:
log.error(f"Error deleting OAuth session: {e}")
return False
def delete_sessions_by_user_id(self, user_id: str) -> bool:
"""Delete all OAuth sessions for a user"""
try:
with get_db() as db:
result = db.query(OAuthSession).filter_by(user_id=user_id).delete()
db.commit()
return True
except Exception as e:
log.error(f"Error deleting OAuth sessions by user ID: {e}")
return False
def delete_sessions_by_provider(self, provider: str) -> bool:
"""Delete all OAuth sessions for a provider"""
try:
with get_db() as db:
db.query(OAuthSession).filter_by(provider=provider).delete()
db.commit()
return True
except Exception as e:
log.error(f"Error deleting OAuth sessions by provider {provider}: {e}")
return False
OAuthSessions = OAuthSessionTable()

16
backend/open_webui/models/prompts.py
View file

@ -2,6 +2,7 @@ import time
from typing import Optional
from open_webui.internal.db import Base, get_db
from open_webui.models.groups import Groups
from open_webui.models.users import Users, UserResponse
from pydantic import BaseModel, ConfigDict
@ -103,10 +104,16 @@ class PromptsTable:
def get_prompts(self) -> list[PromptUserResponse]:
with get_db() as db:
prompts = []
all_prompts = db.query(Prompt).order_by(Prompt.timestamp.desc()).all()
for prompt in db.query(Prompt).order_by(Prompt.timestamp.desc()).all():
user = Users.get_user_by_id(prompt.user_id)
user_ids = list(set(prompt.user_id for prompt in all_prompts))
users = Users.get_users_by_user_ids(user_ids) if user_ids else []
users_dict = {user.id: user for user in users}
prompts = []
for prompt in all_prompts:
user = users_dict.get(prompt.user_id)
prompts.append(
PromptUserResponse.model_validate(
{
@ -122,12 +129,13 @@ class PromptsTable:
self, user_id: str, permission: str = "write"
) -> list[PromptUserResponse]:
prompts = self.get_prompts()
user_group_ids = {group.id for group in Groups.get_groups_by_member_id(user_id)}
return [
prompt
for prompt in prompts
if prompt.user_id == user_id
or has_access(user_id, permission, prompt.access_control)
or has_access(user_id, permission, prompt.access_control, user_group_ids)
]
def update_prompt_by_command(

7
backend/open_webui/models/tags.py
View file

@ -8,7 +8,7 @@ from open_webui.internal.db import Base, get_db
from open_webui.env import SRC_LOG_LEVELS
from pydantic import BaseModel, ConfigDict
from sqlalchemy import BigInteger, Column, String, JSON, PrimaryKeyConstraint
from sqlalchemy import BigInteger, Column, String, JSON, PrimaryKeyConstraint, Index
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["MODELS"])
@ -24,6 +24,11 @@ class Tag(Base):
user_id = Column(String)
meta = Column(JSON, nullable=True)
__table_args__ = (
PrimaryKeyConstraint("id", "user_id", name="pk_id_user_id"),
Index("user_id_idx", "user_id"),
)
# Unique constraint ensuring (id, user_id) is unique, not just the `id` column
__table_args__ = (PrimaryKeyConstraint("id", "user_id", name="pk_id_user_id"),)

20
backend/open_webui/models/tools.py
View file

@ -4,6 +4,8 @@ from typing import Optional
from open_webui.internal.db import Base, JSONField, get_db
from open_webui.models.users import Users, UserResponse
from open_webui.models.groups import Groups
from open_webui.env import SRC_LOG_LEVELS
from pydantic import BaseModel, ConfigDict
from sqlalchemy import BigInteger, Column, String, Text, JSON
@ -93,6 +95,8 @@ class ToolResponse(BaseModel):
class ToolUserResponse(ToolResponse):
user: Optional[UserResponse] = None
model_config = ConfigDict(extra="allow")
class ToolForm(BaseModel):
id: str
@ -144,9 +148,16 @@ class ToolsTable:
def get_tools(self) -> list[ToolUserModel]:
with get_db() as db:
all_tools = db.query(Tool).order_by(Tool.updated_at.desc()).all()
user_ids = list(set(tool.user_id for tool in all_tools))
users = Users.get_users_by_user_ids(user_ids) if user_ids else []
users_dict = {user.id: user for user in users}
tools = []
for tool in db.query(Tool).order_by(Tool.updated_at.desc()).all():
user = Users.get_user_by_id(tool.user_id)
for tool in all_tools:
user = users_dict.get(tool.user_id)
tools.append(
ToolUserModel.model_validate(
{
@ -161,12 +172,13 @@ class ToolsTable:
self, user_id: str, permission: str = "write"
) -> list[ToolUserModel]:
tools = self.get_tools()
user_group_ids = {group.id for group in Groups.get_groups_by_member_id(user_id)}
return [
tool
for tool in tools
if tool.user_id == user_id
or has_access(user_id, permission, tool.access_control)
or has_access(user_id, permission, tool.access_control, user_group_ids)
]
def get_tool_valves_by_id(self, id: str) -> Optional[dict]:
@ -175,7 +187,7 @@ class ToolsTable:
tool = db.get(Tool, id)
return tool.valves if tool.valves else {}
except Exception as e:
log.exception(f"Error getting tool valves by id {id}: {e}")
log.exception(f"Error getting tool valves by id {id}")
return None
def update_tool_valves_by_id(self, id: str, valves: dict) -> Optional[ToolValves]:

176
backend/open_webui/models/users.py
View file

@ -4,14 +4,17 @@ from typing import Optional
from open_webui.internal.db import Base, JSONField, get_db
from open_webui.env import DATABASE_USER_ACTIVE_STATUS_UPDATE_INTERVAL
from open_webui.models.chats import Chats
from open_webui.models.groups import Groups
from open_webui.models.groups import Groups, GroupMember
from open_webui.utils.misc import throttle
from pydantic import BaseModel, ConfigDict
from sqlalchemy import BigInteger, Column, String, Text
from sqlalchemy import or_
from sqlalchemy import BigInteger, Column, String, Text, Date, exists, select
from sqlalchemy import or_, case
import datetime
####################
# User DB Schema
@ -23,20 +26,28 @@ class User(Base):
id = Column(String, primary_key=True)
name = Column(String)
email = Column(String)
username = Column(String(50), nullable=True)
role = Column(String)
profile_image_url = Column(Text)
last_active_at = Column(BigInteger)
updated_at = Column(BigInteger)
created_at = Column(BigInteger)
bio = Column(Text, nullable=True)
gender = Column(Text, nullable=True)
date_of_birth = Column(Date, nullable=True)
info = Column(JSONField, nullable=True)
settings = Column(JSONField, nullable=True)
api_key = Column(String, nullable=True, unique=True)
settings = Column(JSONField, nullable=True)
info = Column(JSONField, nullable=True)
oauth_sub = Column(Text, unique=True)
last_active_at = Column(BigInteger)
updated_at = Column(BigInteger)
created_at = Column(BigInteger)
class UserSettings(BaseModel):
ui: Optional[dict] = {}
@ -47,20 +58,27 @@ class UserSettings(BaseModel):
class UserModel(BaseModel):
id: str
name: str
email: str
username: Optional[str] = None
role: str = "pending"
profile_image_url: str
bio: Optional[str] = None
gender: Optional[str] = None
date_of_birth: Optional[datetime.date] = None
info: Optional[dict] = None
settings: Optional[UserSettings] = None
api_key: Optional[str] = None
oauth_sub: Optional[str] = None
last_active_at: int # timestamp in epoch
updated_at: int # timestamp in epoch
created_at: int # timestamp in epoch
api_key: Optional[str] = None
settings: Optional[UserSettings] = None
info: Optional[dict] = None
oauth_sub: Optional[str] = None
model_config = ConfigDict(from_attributes=True)
@ -69,8 +87,51 @@ class UserModel(BaseModel):
####################
class UpdateProfileForm(BaseModel):
profile_image_url: str
name: str
bio: Optional[str] = None
gender: Optional[str] = None
date_of_birth: Optional[datetime.date] = None
class UserGroupIdsModel(UserModel):
group_ids: list[str] = []
class UserModelResponse(UserModel):
model_config = ConfigDict(extra="allow")
class UserListResponse(BaseModel):
users: list[UserModel]
users: list[UserModelResponse]
total: int
class UserGroupIdsListResponse(BaseModel):
users: list[UserGroupIdsModel]
total: int
class UserInfoResponse(BaseModel):
id: str
name: str
email: str
role: str
class UserIdNameResponse(BaseModel):
id: str
name: str
class UserInfoListResponse(BaseModel):
users: list[UserInfoResponse]
total: int
class UserIdNameListResponse(BaseModel):
users: list[UserIdNameResponse]
total: int
@ -172,8 +233,9 @@ class UsersTable:
filter: Optional[dict] = None,
skip: Optional[int] = None,
limit: Optional[int] = None,
) -> UserListResponse:
) -> dict:
with get_db() as db:
# Join GroupMember so we can order by group_id when requested
query = db.query(User)
if filter:
@ -186,14 +248,65 @@ class UsersTable:
)
)
user_ids = filter.get("user_ids")
group_ids = filter.get("group_ids")
if isinstance(user_ids, list) and isinstance(group_ids, list):
# If both are empty lists, return no users
if not user_ids and not group_ids:
return {"users": [], "total": 0}
if user_ids:
query = query.filter(User.id.in_(user_ids))
if group_ids:
query = query.filter(
exists(
select(GroupMember.id).where(
GroupMember.user_id == User.id,
GroupMember.group_id.in_(group_ids),
)
)
)
roles = filter.get("roles")
if roles:
include_roles = [role for role in roles if not role.startswith("!")]
exclude_roles = [role[1:] for role in roles if role.startswith("!")]
if include_roles:
query = query.filter(User.role.in_(include_roles))
if exclude_roles:
query = query.filter(~User.role.in_(exclude_roles))
order_by = filter.get("order_by")
direction = filter.get("direction")
if order_by == "name":
if order_by and order_by.startswith("group_id:"):
group_id = order_by.split(":", 1)[1]
# Subquery that checks if the user belongs to the group
membership_exists = exists(
select(GroupMember.id).where(
GroupMember.user_id == User.id,
GroupMember.group_id == group_id,
)
)
# CASE: user in group → 1, user not in group → 0
group_sort = case((membership_exists, 1), else_=0)
if direction == "asc":
query = query.order_by(group_sort.asc(), User.name.asc())
else:
query = query.order_by(group_sort.desc(), User.name.asc())
elif order_by == "name":
if direction == "asc":
query = query.order_by(User.name.asc())
else:
query = query.order_by(User.name.desc())
elif order_by == "email":
if direction == "asc":
query = query.order_by(User.email.asc())
@ -226,15 +339,19 @@ class UsersTable:
else:
query = query.order_by(User.created_at.desc())
if skip:
# Count BEFORE pagination
total = query.count()
# correct pagination logic
if skip is not None:
query = query.offset(skip)
if limit:
if limit is not None:
query = query.limit(limit)
users = query.all()
return {
"users": [UserModel.model_validate(user) for user in users],
"total": db.query(User).count(),
"total": total,
}
def get_users_by_user_ids(self, user_ids: list[str]) -> list[UserModel]:
@ -246,6 +363,10 @@ class UsersTable:
with get_db() as db:
return db.query(User).count()
def has_users(self) -> bool:
with get_db() as db:
return db.query(db.query(User).exists()).scalar()
def get_first_user(self) -> UserModel:
try:
with get_db() as db:
@ -270,6 +391,15 @@ class UsersTable:
except Exception:
return None
def get_num_users_active_today(self) -> Optional[int]:
with get_db() as db:
current_timestamp = int(datetime.datetime.now().timestamp())
today_midnight_timestamp = current_timestamp - (current_timestamp % 86400)
query = db.query(User).filter(
User.last_active_at > today_midnight_timestamp
)
return query.count()
def update_user_role_by_id(self, id: str, role: str) -> Optional[UserModel]:
try:
with get_db() as db:
@ -295,6 +425,7 @@ class UsersTable:
except Exception:
return None
@throttle(DATABASE_USER_ACTIVE_STATUS_UPDATE_INTERVAL)
def update_user_last_active_by_id(self, id: str) -> Optional[UserModel]:
try:
with get_db() as db:
@ -330,7 +461,8 @@ class UsersTable:
user = db.query(User).filter_by(id=id).first()
return UserModel.model_validate(user)
# return UserModel(**user.dict())
except Exception:
except Exception as e:
print(e)
return None
def update_user_settings_by_id(self, id: str, updated: dict) -> Optional[UserModel]:

165
backend/open_webui/retrieval/loaders/datalab_marker.py
View file

@ -15,24 +15,28 @@ class DatalabMarkerLoader:
self,
file_path: str,
api_key: str,
langs: Optional[str] = None,
api_base_url: str,
additional_config: Optional[str] = None,
use_llm: bool = False,
skip_cache: bool = False,
force_ocr: bool = False,
paginate: bool = False,
strip_existing_ocr: bool = False,
disable_image_extraction: bool = False,
format_lines: bool = False,
output_format: str = None,
):
self.file_path = file_path
self.api_key = api_key
self.langs = langs
self.api_base_url = api_base_url
self.additional_config = additional_config
self.use_llm = use_llm
self.skip_cache = skip_cache
self.force_ocr = force_ocr
self.paginate = paginate
self.strip_existing_ocr = strip_existing_ocr
self.disable_image_extraction = disable_image_extraction
self.format_lines = format_lines
self.output_format = output_format
def _get_mime_type(self, filename: str) -> str:
@ -60,7 +64,7 @@ class DatalabMarkerLoader:
return mime_map.get(ext, "application/octet-stream")
def check_marker_request_status(self, request_id: str) -> dict:
url = f"https://www.datalab.to/api/v1/marker/{request_id}"
url = f"{self.api_base_url}/{request_id}"
headers = {"X-Api-Key": self.api_key}
try:
response = requests.get(url, headers=headers)
@ -81,22 +85,24 @@ class DatalabMarkerLoader:
)
def load(self) -> List[Document]:
url = "https://www.datalab.to/api/v1/marker"
filename = os.path.basename(self.file_path)
mime_type = self._get_mime_type(filename)
headers = {"X-Api-Key": self.api_key}
form_data = {
"langs": self.langs,
"use_llm": str(self.use_llm).lower(),
"skip_cache": str(self.skip_cache).lower(),
"force_ocr": str(self.force_ocr).lower(),
"paginate": str(self.paginate).lower(),
"strip_existing_ocr": str(self.strip_existing_ocr).lower(),
"disable_image_extraction": str(self.disable_image_extraction).lower(),
"format_lines": str(self.format_lines).lower(),
"output_format": self.output_format,
}
if self.additional_config and self.additional_config.strip():
form_data["additional_config"] = self.additional_config
log.info(
f"Datalab Marker POST request parameters: {{'filename': '{filename}', 'mime_type': '{mime_type}', **{form_data}}}"
)
@ -105,7 +111,10 @@ class DatalabMarkerLoader:
with open(self.file_path, "rb") as f:
files = {"file": (filename, f, mime_type)}
response = requests.post(
url, data=form_data, files=files, headers=headers
f"{self.api_base_url}",
data=form_data,
files=files,
headers=headers,
)
response.raise_for_status()
result = response.json()
@ -133,74 +142,92 @@ class DatalabMarkerLoader:
check_url = result.get("request_check_url")
request_id = result.get("request_id")
if not check_url:
raise HTTPException(
status.HTTP_502_BAD_GATEWAY, detail="No request_check_url returned."
)
for _ in range(300): # Up to 10 minutes
time.sleep(2)
try:
poll_response = requests.get(check_url, headers=headers)
poll_response.raise_for_status()
poll_result = poll_response.json()
except (requests.HTTPError, ValueError) as e:
raw_body = poll_response.text
log.error(f"Polling error: {e}, response body: {raw_body}")
raise HTTPException(
status.HTTP_502_BAD_GATEWAY, detail=f"Polling failed: {e}"
)
status_val = poll_result.get("status")
success_val = poll_result.get("success")
if status_val == "complete":
summary = {
k: poll_result.get(k)
for k in (
"status",
"output_format",
"success",
"error",
"page_count",
"total_cost",
# Check if this is a direct response (self-hosted) or polling response (DataLab)
if check_url:
# DataLab polling pattern
for _ in range(300): # Up to 10 minutes
time.sleep(2)
try:
poll_response = requests.get(check_url, headers=headers)
poll_response.raise_for_status()
poll_result = poll_response.json()
except (requests.HTTPError, ValueError) as e:
raw_body = poll_response.text
log.error(f"Polling error: {e}, response body: {raw_body}")
raise HTTPException(
status.HTTP_502_BAD_GATEWAY, detail=f"Polling failed: {e}"
)
}
log.info(
f"Marker processing completed successfully: {json.dumps(summary, indent=2)}"
)
break
if status_val == "failed" or success_val is False:
log.error(
f"Marker poll failed full response: {json.dumps(poll_result, indent=2)}"
)
error_msg = (
poll_result.get("error")
or "Marker returned failure without error message"
status_val = poll_result.get("status")
success_val = poll_result.get("success")
if status_val == "complete":
summary = {
k: poll_result.get(k)
for k in (
"status",
"output_format",
"success",
"error",
"page_count",
"total_cost",
)
}
log.info(
f"Marker processing completed successfully: {json.dumps(summary, indent=2)}"
)
break
if status_val == "failed" or success_val is False:
log.error(
f"Marker poll failed full response: {json.dumps(poll_result, indent=2)}"
)
error_msg = (
poll_result.get("error")
or "Marker returned failure without error message"
)
raise HTTPException(
status.HTTP_400_BAD_REQUEST,
detail=f"Marker processing failed: {error_msg}",
)
else:
raise HTTPException(
status.HTTP_504_GATEWAY_TIMEOUT,
detail="Marker processing timed out",
)
if not poll_result.get("success", False):
error_msg = poll_result.get("error") or "Unknown processing error"
raise HTTPException(
status.HTTP_400_BAD_REQUEST,
detail=f"Marker processing failed: {error_msg}",
detail=f"Final processing failed: {error_msg}",
)
# DataLab format - content in format-specific fields
content_key = self.output_format.lower()
raw_content = poll_result.get(content_key)
final_result = poll_result
else:
raise HTTPException(
status.HTTP_504_GATEWAY_TIMEOUT, detail="Marker processing timed out"
)
# Self-hosted direct response - content in "output" field
if "output" in result:
log.info("Self-hosted Marker returned direct response without polling")
raw_content = result.get("output")
final_result = result
else:
available_fields = (
list(result.keys())
if isinstance(result, dict)
else "non-dict response"
)
raise HTTPException(
status.HTTP_502_BAD_GATEWAY,
detail=f"Custom Marker endpoint returned success but no 'output' field found. Available fields: {available_fields}. Expected either 'request_check_url' for polling or 'output' field for direct response.",
)
if not poll_result.get("success", False):
error_msg = poll_result.get("error") or "Unknown processing error"
raise HTTPException(
status.HTTP_400_BAD_REQUEST,
detail=f"Final processing failed: {error_msg}",
)
content_key = self.output_format.lower()
raw_content = poll_result.get(content_key)
if content_key == "json":
if self.output_format.lower() == "json":
full_text = json.dumps(raw_content, indent=2)
elif content_key in {"markdown", "html"}:
elif self.output_format.lower() in {"markdown", "html"}:
full_text = str(raw_content).strip()
else:
raise HTTPException(
@ -211,14 +238,14 @@ class DatalabMarkerLoader:
if not full_text:
raise HTTPException(
status.HTTP_400_BAD_REQUEST,
detail="Datalab Marker returned empty content",
detail="Marker returned empty content",
)
marker_output_dir = os.path.join("/app/backend/data/uploads", "marker_output")
os.makedirs(marker_output_dir, exist_ok=True)
file_ext_map = {"markdown": "md", "json": "json", "html": "html"}
file_ext = file_ext_map.get(content_key, "txt")
file_ext = file_ext_map.get(self.output_format.lower(), "txt")
output_filename = f"{os.path.splitext(filename)[0]}.{file_ext}"
output_path = os.path.join(marker_output_dir, output_filename)
@ -231,13 +258,13 @@ class DatalabMarkerLoader:
metadata = {
"source": filename,
"output_format": poll_result.get("output_format", self.output_format),
"page_count": poll_result.get("page_count", 0),
"output_format": final_result.get("output_format", self.output_format),
"page_count": final_result.get("page_count", 0),
"processed_with_llm": self.use_llm,
"request_id": request_id or "",
}
images = poll_result.get("images", {})
images = final_result.get("images", {})
if images:
metadata["image_count"] = len(images)
metadata["images"] = json.dumps(list(images.keys()))

10
backend/open_webui/retrieval/loaders/external_document.py
View file

@ -1,9 +1,11 @@
import requests
import logging, os
from typing import Iterator, List, Union
from urllib.parse import quote
from langchain_core.document_loaders import BaseLoader
from langchain_core.documents import Document
from open_webui.utils.headers import include_user_info_headers
from open_webui.env import SRC_LOG_LEVELS
log = logging.getLogger(__name__)
@ -17,6 +19,7 @@ class ExternalDocumentLoader(BaseLoader):
url: str,
api_key: str,
mime_type=None,
user=None,
**kwargs,
) -> None:
self.url = url
@ -25,6 +28,8 @@ class ExternalDocumentLoader(BaseLoader):
self.file_path = file_path
self.mime_type = mime_type
self.user = user
def load(self) -> List[Document]:
with open(self.file_path, "rb") as f:
data = f.read()
@ -37,10 +42,13 @@ class ExternalDocumentLoader(BaseLoader):
headers["Authorization"] = f"Bearer {self.api_key}"
try:
headers["X-Filename"] = os.path.basename(self.file_path)
headers["X-Filename"] = quote(os.path.basename(self.file_path))
except:
pass
if self.user is not None:
headers = include_user_info_headers(headers, self.user)
url = self.url
if url.endswith("/"):
url = url[:-1]

118
backend/open_webui/retrieval/loaders/main.py
View file

@ -4,6 +4,7 @@ import ftfy
import sys
import json
from azure.identity import DefaultAzureCredential
from langchain_community.document_loaders import (
AzureAIDocumentIntelligenceLoader,
BSHTMLLoader,
@ -14,7 +15,7 @@ from langchain_community.document_loaders import (
TextLoader,
UnstructuredEPubLoader,
UnstructuredExcelLoader,
UnstructuredMarkdownLoader,
UnstructuredODTLoader,
UnstructuredPowerPointLoader,
UnstructuredRSTLoader,
UnstructuredXMLLoader,
@ -26,6 +27,7 @@ from open_webui.retrieval.loaders.external_document import ExternalDocumentLoade
from open_webui.retrieval.loaders.mistral import MistralLoader
from open_webui.retrieval.loaders.datalab_marker import DatalabMarkerLoader
from open_webui.retrieval.loaders.mineru import MinerULoader
from open_webui.env import SRC_LOG_LEVELS, GLOBAL_LOG_LEVEL
@ -130,8 +132,9 @@ class TikaLoader:
class DoclingLoader:
def __init__(self, url, file_path=None, mime_type=None, params=None):
def __init__(self, url, api_key=None, file_path=None, mime_type=None, params=None):
self.url = url.rstrip("/")
self.api_key = api_key
self.file_path = file_path
self.mime_type = mime_type
@ -139,6 +142,10 @@ class DoclingLoader:
def load(self) -> list[Document]:
with open(self.file_path, "rb") as f:
headers = {}
if self.api_key:
headers["Authorization"] = f"Bearer {self.api_key}"
files = {
"files": (
self.file_path,
@ -147,43 +154,15 @@ class DoclingLoader:
)
}
params = {"image_export_mode": "placeholder", "table_mode": "accurate"}
if self.params:
if self.params.get("do_picture_description"):
params["do_picture_description"] = self.params.get(
"do_picture_description"
)
picture_description_mode = self.params.get(
"picture_description_mode", ""
).lower()
if picture_description_mode == "local" and self.params.get(
"picture_description_local", {}
):
params["picture_description_local"] = json.dumps(
self.params.get("picture_description_local", {})
)
elif picture_description_mode == "api" and self.params.get(
"picture_description_api", {}
):
params["picture_description_api"] = json.dumps(
self.params.get("picture_description_api", {})
)
if self.params.get("ocr_engine") and self.params.get("ocr_lang"):
params["ocr_engine"] = self.params.get("ocr_engine")
params["ocr_lang"] = [
lang.strip()
for lang in self.params.get("ocr_lang").split(",")
if lang.strip()
]
endpoint = f"{self.url}/v1alpha/convert/file"
r = requests.post(endpoint, files=files, data=params)
r = requests.post(
f"{self.url}/v1/convert/file",
files=files,
data={
"image_export_mode": "placeholder",
**self.params,
},
headers=headers,
)
if r.ok:
result = r.json()
document_data = result.get("document", {})
@ -192,7 +171,6 @@ class DoclingLoader:
metadata = {"Content-Type": self.mime_type} if self.mime_type else {}
log.debug("Docling extracted text: %s", text)
return [Document(page_content=text, metadata=metadata)]
else:
error_msg = f"Error calling Docling API: {r.reason}"
@ -209,6 +187,7 @@ class DoclingLoader:
class Loader:
def __init__(self, engine: str = "", **kwargs):
self.engine = engine
self.user = kwargs.get("user", None)
self.kwargs = kwargs
def load(
@ -226,7 +205,10 @@ class Loader:
def _is_text_file(self, file_ext: str, file_content_type: str) -> bool:
return file_ext in known_source_ext or (
file_content_type and file_content_type.find("text/") >= 0
file_content_type
and file_content_type.find("text/") >= 0
# Avoid text/html files being detected as text
and not file_content_type.find("html") >= 0
)
def _get_loader(self, filename: str, file_content_type: str, file_path: str):
@ -242,6 +224,7 @@ class Loader:
url=self.kwargs.get("EXTERNAL_DOCUMENT_LOADER_URL"),
api_key=self.kwargs.get("EXTERNAL_DOCUMENT_LOADER_API_KEY"),
mime_type=file_content_type,
user=self.user,
)
elif self.engine == "tika" and self.kwargs.get("TIKA_SERVER_URL"):
if self._is_text_file(file_ext, file_content_type):
@ -250,7 +233,6 @@ class Loader:
loader = TikaLoader(
url=self.kwargs.get("TIKA_SERVER_URL"),
file_path=file_path,
mime_type=file_content_type,
extract_images=self.kwargs.get("PDF_EXTRACT_IMAGES"),
)
elif (
@ -278,10 +260,15 @@ class Loader:
"tiff",
]
):
api_base_url = self.kwargs.get("DATALAB_MARKER_API_BASE_URL", "")
if not api_base_url or api_base_url.strip() == "":
api_base_url = "https://www.datalab.to/api/v1/marker" # https://github.com/open-webui/open-webui/pull/16867#issuecomment-3218424349
loader = DatalabMarkerLoader(
file_path=file_path,
api_key=self.kwargs["DATALAB_MARKER_API_KEY"],
langs=self.kwargs.get("DATALAB_MARKER_LANGS"),
api_base_url=api_base_url,
additional_config=self.kwargs.get("DATALAB_MARKER_ADDITIONAL_CONFIG"),
use_llm=self.kwargs.get("DATALAB_MARKER_USE_LLM", False),
skip_cache=self.kwargs.get("DATALAB_MARKER_SKIP_CACHE", False),
force_ocr=self.kwargs.get("DATALAB_MARKER_FORCE_OCR", False),
@ -292,6 +279,7 @@ class Loader:
disable_image_extraction=self.kwargs.get(
"DATALAB_MARKER_DISABLE_IMAGE_EXTRACTION", False
),
format_lines=self.kwargs.get("DATALAB_MARKER_FORMAT_LINES", False),
output_format=self.kwargs.get(
"DATALAB_MARKER_OUTPUT_FORMAT", "markdown"
),
@ -311,6 +299,7 @@ class Loader:
loader = DoclingLoader(
url=self.kwargs.get("DOCLING_SERVER_URL"),
api_key=self.kwargs.get("DOCLING_API_KEY", None),
file_path=file_path,
mime_type=file_content_type,
params=params,
@ -318,23 +307,37 @@ class Loader:
elif (
self.engine == "document_intelligence"
and self.kwargs.get("DOCUMENT_INTELLIGENCE_ENDPOINT") != ""
and self.kwargs.get("DOCUMENT_INTELLIGENCE_KEY") != ""
and (
file_ext in ["pdf", "xls", "xlsx", "docx", "ppt", "pptx"]
file_ext in ["pdf", "docx", "ppt", "pptx"]
or file_content_type
in [
"application/vnd.ms-excel",
"application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
"application/vnd.openxmlformats-officedocument.wordprocessingml.document",
"application/vnd.ms-powerpoint",
"application/vnd.openxmlformats-officedocument.presentationml.presentation",
]
)
):
loader = AzureAIDocumentIntelligenceLoader(
if self.kwargs.get("DOCUMENT_INTELLIGENCE_KEY") != "":
loader = AzureAIDocumentIntelligenceLoader(
file_path=file_path,
api_endpoint=self.kwargs.get("DOCUMENT_INTELLIGENCE_ENDPOINT"),
api_key=self.kwargs.get("DOCUMENT_INTELLIGENCE_KEY"),
)
else:
loader = AzureAIDocumentIntelligenceLoader(
file_path=file_path,
api_endpoint=self.kwargs.get("DOCUMENT_INTELLIGENCE_ENDPOINT"),
azure_credential=DefaultAzureCredential(),
)
elif self.engine == "mineru" and file_ext in [
"pdf"
]: # MinerU currently only supports PDF
loader = MinerULoader(
file_path=file_path,
api_endpoint=self.kwargs.get("DOCUMENT_INTELLIGENCE_ENDPOINT"),
api_key=self.kwargs.get("DOCUMENT_INTELLIGENCE_KEY"),
api_mode=self.kwargs.get("MINERU_API_MODE", "local"),
api_url=self.kwargs.get("MINERU_API_URL", "http://localhost:8000"),
api_key=self.kwargs.get("MINERU_API_KEY", ""),
params=self.kwargs.get("MINERU_PARAMS", {}),
)
elif (
self.engine == "mistral_ocr"
@ -343,16 +346,9 @@ class Loader:
in ["pdf"] # Mistral OCR currently only supports PDF and images
):
loader = MistralLoader(
api_key=self.kwargs.get("MISTRAL_OCR_API_KEY"), file_path=file_path
)
elif (
self.engine == "external"
and self.kwargs.get("MISTRAL_OCR_API_KEY") != ""
and file_ext
in ["pdf"] # Mistral OCR currently only supports PDF and images
):
loader = MistralLoader(
api_key=self.kwargs.get("MISTRAL_OCR_API_KEY"), file_path=file_path
base_url=self.kwargs.get("MISTRAL_OCR_API_BASE_URL"),
api_key=self.kwargs.get("MISTRAL_OCR_API_KEY"),
file_path=file_path,
)
else:
if file_ext == "pdf":
@ -389,6 +385,8 @@ class Loader:
loader = UnstructuredPowerPointLoader(file_path)
elif file_ext == "msg":
loader = OutlookMessageLoader(file_path)
elif file_ext == "odt":
loader = UnstructuredODTLoader(file_path)
elif self._is_text_file(file_ext, file_content_type):
loader = TextLoader(file_path, autodetect_encoding=True)
else:

522
backend/open_webui/retrieval/loaders/mineru.py Normal file
View file

@ -0,0 +1,522 @@
import os
import time
import requests
import logging
import tempfile
import zipfile
from typing import List, Optional
from langchain_core.documents import Document
from fastapi import HTTPException, status
log = logging.getLogger(__name__)
class MinerULoader:
"""
MinerU document parser loader supporting both Cloud API and Local API modes.
Cloud API: Uses MinerU managed service with async task-based processing
Local API: Uses self-hosted MinerU API with synchronous processing
"""
def __init__(
self,
file_path: str,
api_mode: str = "local",
api_url: str = "http://localhost:8000",
api_key: str = "",
params: dict = None,
):
self.file_path = file_path
self.api_mode = api_mode.lower()
self.api_url = api_url.rstrip("/")
self.api_key = api_key
# Parse params dict with defaults
self.params = params or {}
self.enable_ocr = params.get("enable_ocr", False)
self.enable_formula = params.get("enable_formula", True)
self.enable_table = params.get("enable_table", True)
self.language = params.get("language", "en")
self.model_version = params.get("model_version", "pipeline")
self.page_ranges = self.params.pop("page_ranges", "")
# Validate API mode
if self.api_mode not in ["local", "cloud"]:
raise ValueError(
f"Invalid API mode: {self.api_mode}. Must be 'local' or 'cloud'"
)
# Validate Cloud API requirements
if self.api_mode == "cloud" and not self.api_key:
raise ValueError("API key is required for Cloud API mode")
def load(self) -> List[Document]:
"""
Main entry point for loading and parsing the document.
Routes to Cloud or Local API based on api_mode.
"""
try:
if self.api_mode == "cloud":
return self._load_cloud_api()
else:
return self._load_local_api()
except Exception as e:
log.error(f"Error loading document with MinerU: {e}")
raise
def _load_local_api(self) -> List[Document]:
"""
Load document using Local API (synchronous).
Posts file to /file_parse endpoint and gets immediate response.
"""
log.info(f"Using MinerU Local API at {self.api_url}")
filename = os.path.basename(self.file_path)
# Build form data for Local API
form_data = {
**self.params,
"return_md": "true",
}
# Page ranges (Local API uses start_page_id and end_page_id)
if self.page_ranges:
# For simplicity, if page_ranges is specified, log a warning
# Full page range parsing would require parsing the string
log.warning(
f"Page ranges '{self.page_ranges}' specified but Local API uses different format. "
"Consider using start_page_id/end_page_id parameters if needed."
)
try:
with open(self.file_path, "rb") as f:
files = {"files": (filename, f, "application/octet-stream")}
log.info(f"Sending file to MinerU Local API: {filename}")
log.debug(f"Local API parameters: {form_data}")
response = requests.post(
f"{self.api_url}/file_parse",
data=form_data,
files=files,
timeout=300, # 5 minute timeout for large documents
)
response.raise_for_status()
except FileNotFoundError:
raise HTTPException(
status.HTTP_404_NOT_FOUND, detail=f"File not found: {self.file_path}"
)
except requests.Timeout:
raise HTTPException(
status.HTTP_504_GATEWAY_TIMEOUT,
detail="MinerU Local API request timed out",
)
except requests.HTTPError as e:
error_detail = f"MinerU Local API request failed: {e}"
if e.response is not None:
try:
error_data = e.response.json()
error_detail += f" - {error_data}"
except:
error_detail += f" - {e.response.text}"
raise HTTPException(status.HTTP_400_BAD_REQUEST, detail=error_detail)
except Exception as e:
raise HTTPException(
status.HTTP_500_INTERNAL_SERVER_ERROR,
detail=f"Error calling MinerU Local API: {str(e)}",
)
# Parse response
try:
result = response.json()
except ValueError as e:
raise HTTPException(
status.HTTP_502_BAD_GATEWAY,
detail=f"Invalid JSON response from MinerU Local API: {e}",
)
# Extract markdown content from response
if "results" not in result:
raise HTTPException(
status.HTTP_502_BAD_GATEWAY,
detail="MinerU Local API response missing 'results' field",
)
results = result["results"]
if not results:
raise HTTPException(
status.HTTP_400_BAD_REQUEST,
detail="MinerU returned empty results",
)
# Get the first (and typically only) result
file_result = list(results.values())[0]
markdown_content = file_result.get("md_content", "")
if not markdown_content:
raise HTTPException(
status.HTTP_400_BAD_REQUEST,
detail="MinerU returned empty markdown content",
)
log.info(f"Successfully parsed document with MinerU Local API: {filename}")
# Create metadata
metadata = {
"source": filename,
"api_mode": "local",
"backend": result.get("backend", "unknown"),
"version": result.get("version", "unknown"),
}
return [Document(page_content=markdown_content, metadata=metadata)]
def _load_cloud_api(self) -> List[Document]:
"""
Load document using Cloud API (asynchronous).
Uses batch upload endpoint to avoid need for public file URLs.
"""
log.info(f"Using MinerU Cloud API at {self.api_url}")
filename = os.path.basename(self.file_path)
# Step 1: Request presigned upload URL
batch_id, upload_url = self._request_upload_url(filename)
# Step 2: Upload file to presigned URL
self._upload_to_presigned_url(upload_url)
# Step 3: Poll for results
result = self._poll_batch_status(batch_id, filename)
# Step 4: Download and extract markdown from ZIP
markdown_content = self._download_and_extract_zip(
result["full_zip_url"], filename
)
log.info(f"Successfully parsed document with MinerU Cloud API: {filename}")
# Create metadata
metadata = {
"source": filename,
"api_mode": "cloud",
"batch_id": batch_id,
}
return [Document(page_content=markdown_content, metadata=metadata)]
def _request_upload_url(self, filename: str) -> tuple:
"""
Request presigned upload URL from Cloud API.
Returns (batch_id, upload_url).
"""
headers = {
"Authorization": f"Bearer {self.api_key}",
"Content-Type": "application/json",
}
# Build request body
request_body = {
**self.params,
"files": [
{
"name": filename,
"is_ocr": self.enable_ocr,
}
],
}
# Add page ranges if specified
if self.page_ranges:
request_body["files"][0]["page_ranges"] = self.page_ranges
log.info(f"Requesting upload URL for: {filename}")
log.debug(f"Cloud API request body: {request_body}")
try:
response = requests.post(
f"{self.api_url}/file-urls/batch",
headers=headers,
json=request_body,
timeout=30,
)
response.raise_for_status()
except requests.HTTPError as e:
error_detail = f"Failed to request upload URL: {e}"
if e.response is not None:
try:
error_data = e.response.json()
error_detail += f" - {error_data.get('msg', error_data)}"
except:
error_detail += f" - {e.response.text}"
raise HTTPException(status.HTTP_400_BAD_REQUEST, detail=error_detail)
except Exception as e:
raise HTTPException(
status.HTTP_500_INTERNAL_SERVER_ERROR,
detail=f"Error requesting upload URL: {str(e)}",
)
try:
result = response.json()
except ValueError as e:
raise HTTPException(
status.HTTP_502_BAD_GATEWAY,
detail=f"Invalid JSON response: {e}",
)
# Check for API error response
if result.get("code") != 0:
raise HTTPException(
status.HTTP_400_BAD_REQUEST,
detail=f"MinerU Cloud API error: {result.get('msg', 'Unknown error')}",
)
data = result.get("data", {})
batch_id = data.get("batch_id")
file_urls = data.get("file_urls", [])
if not batch_id or not file_urls:
raise HTTPException(
status.HTTP_502_BAD_GATEWAY,
detail="MinerU Cloud API response missing batch_id or file_urls",
)
upload_url = file_urls[0]
log.info(f"Received upload URL for batch: {batch_id}")
return batch_id, upload_url
def _upload_to_presigned_url(self, upload_url: str) -> None:
"""
Upload file to presigned URL (no authentication needed).
"""
log.info(f"Uploading file to presigned URL")
try:
with open(self.file_path, "rb") as f:
response = requests.put(
upload_url,
data=f,
timeout=300, # 5 minute timeout for large files
)
response.raise_for_status()
except FileNotFoundError:
raise HTTPException(
status.HTTP_404_NOT_FOUND, detail=f"File not found: {self.file_path}"
)
except requests.Timeout:
raise HTTPException(
status.HTTP_504_GATEWAY_TIMEOUT,
detail="File upload to presigned URL timed out",
)
except requests.HTTPError as e:
raise HTTPException(
status.HTTP_400_BAD_REQUEST,
detail=f"Failed to upload file to presigned URL: {e}",
)
except Exception as e:
raise HTTPException(
status.HTTP_500_INTERNAL_SERVER_ERROR,
detail=f"Error uploading file: {str(e)}",
)
log.info("File uploaded successfully")
def _poll_batch_status(self, batch_id: str, filename: str) -> dict:
"""
Poll batch status until completion.
Returns the result dict for the file.
"""
headers = {
"Authorization": f"Bearer {self.api_key}",
}
max_iterations = 300 # 10 minutes max (2 seconds per iteration)
poll_interval = 2 # seconds
log.info(f"Polling batch status: {batch_id}")
for iteration in range(max_iterations):
try:
response = requests.get(
f"{self.api_url}/extract-results/batch/{batch_id}",
headers=headers,
timeout=30,
)
response.raise_for_status()
except requests.HTTPError as e:
error_detail = f"Failed to poll batch status: {e}"
if e.response is not None:
try:
error_data = e.response.json()
error_detail += f" - {error_data.get('msg', error_data)}"
except:
error_detail += f" - {e.response.text}"
raise HTTPException(status.HTTP_400_BAD_REQUEST, detail=error_detail)
except Exception as e:
raise HTTPException(
status.HTTP_500_INTERNAL_SERVER_ERROR,
detail=f"Error polling batch status: {str(e)}",
)
try:
result = response.json()
except ValueError as e:
raise HTTPException(
status.HTTP_502_BAD_GATEWAY,
detail=f"Invalid JSON response while polling: {e}",
)
# Check for API error response
if result.get("code") != 0:
raise HTTPException(
status.HTTP_400_BAD_REQUEST,
detail=f"MinerU Cloud API error: {result.get('msg', 'Unknown error')}",
)
data = result.get("data", {})
extract_result = data.get("extract_result", [])
# Find our file in the batch results
file_result = None
for item in extract_result:
if item.get("file_name") == filename:
file_result = item
break
if not file_result:
raise HTTPException(
status.HTTP_502_BAD_GATEWAY,
detail=f"File {filename} not found in batch results",
)
state = file_result.get("state")
if state == "done":
log.info(f"Processing complete for {filename}")
return file_result
elif state == "failed":
error_msg = file_result.get("err_msg", "Unknown error")
raise HTTPException(
status.HTTP_400_BAD_REQUEST,
detail=f"MinerU processing failed: {error_msg}",
)
elif state in ["waiting-file", "pending", "running", "converting"]:
# Still processing
if iteration % 10 == 0: # Log every 20 seconds
log.info(
f"Processing status: {state} (iteration {iteration + 1}/{max_iterations})"
)
time.sleep(poll_interval)
else:
log.warning(f"Unknown state: {state}")
time.sleep(poll_interval)
# Timeout
raise HTTPException(
status.HTTP_504_GATEWAY_TIMEOUT,
detail="MinerU processing timed out after 10 minutes",
)
def _download_and_extract_zip(self, zip_url: str, filename: str) -> str:
"""
Download ZIP file from CDN and extract markdown content.
Returns the markdown content as a string.
"""
log.info(f"Downloading results from: {zip_url}")
try:
response = requests.get(zip_url, timeout=60)
response.raise_for_status()
except requests.HTTPError as e:
raise HTTPException(
status.HTTP_400_BAD_REQUEST,
detail=f"Failed to download results ZIP: {e}",
)
except Exception as e:
raise HTTPException(
status.HTTP_500_INTERNAL_SERVER_ERROR,
detail=f"Error downloading results: {str(e)}",
)
# Save ZIP to temporary file and extract
try:
with tempfile.NamedTemporaryFile(delete=False, suffix=".zip") as tmp_zip:
tmp_zip.write(response.content)
tmp_zip_path = tmp_zip.name
with tempfile.TemporaryDirectory() as tmp_dir:
# Extract ZIP
with zipfile.ZipFile(tmp_zip_path, "r") as zip_ref:
zip_ref.extractall(tmp_dir)
# Find markdown file - search recursively for any .md file
markdown_content = None
found_md_path = None
# First, list all files in the ZIP for debugging
all_files = []
for root, dirs, files in os.walk(tmp_dir):
for file in files:
full_path = os.path.join(root, file)
all_files.append(full_path)
# Look for any .md file
if file.endswith(".md"):
found_md_path = full_path
log.info(f"Found markdown file at: {full_path}")
try:
with open(full_path, "r", encoding="utf-8") as f:
markdown_content = f.read()
if (
markdown_content
): # Use the first non-empty markdown file
break
except Exception as e:
log.warning(f"Failed to read {full_path}: {e}")
if markdown_content:
break
if markdown_content is None:
log.error(f"Available files in ZIP: {all_files}")
# Try to provide more helpful error message
md_files = [f for f in all_files if f.endswith(".md")]
if md_files:
error_msg = (
f"Found .md files but couldn't read them: {md_files}"
)
else:
error_msg = (
f"No .md files found in ZIP. Available files: {all_files}"
)
raise HTTPException(
status.HTTP_502_BAD_GATEWAY,
detail=error_msg,
)
# Clean up temporary ZIP file
os.unlink(tmp_zip_path)
except zipfile.BadZipFile as e:
raise HTTPException(
status.HTTP_502_BAD_GATEWAY,
detail=f"Invalid ZIP file received: {e}",
)
except Exception as e:
raise HTTPException(
status.HTTP_500_INTERNAL_SERVER_ERROR,
detail=f"Error extracting ZIP: {str(e)}",
)
if not markdown_content:
raise HTTPException(
status.HTTP_400_BAD_REQUEST,
detail="Extracted markdown content is empty",
)
log.info(
f"Successfully extracted markdown content ({len(markdown_content)} characters)"
)
return markdown_content

23
backend/open_webui/retrieval/loaders/mistral.py
View file

@ -30,10 +30,9 @@ class MistralLoader:
- Enhanced error handling with retryable error classification
"""
BASE_API_URL = "https://api.mistral.ai/v1"
def __init__(
self,
base_url: str,
api_key: str,
file_path: str,
timeout: int = 300, # 5 minutes default
@ -55,6 +54,9 @@ class MistralLoader:
if not os.path.exists(file_path):
raise FileNotFoundError(f"File not found at {file_path}")
self.base_url = (
base_url.rstrip("/") if base_url else "https://api.mistral.ai/v1"
)
self.api_key = api_key
self.file_path = file_path
self.timeout = timeout
@ -240,7 +242,7 @@ class MistralLoader:
in a context manager to minimize memory usage duration.
"""
log.info("Uploading file to Mistral API")
url = f"{self.BASE_API_URL}/files"
url = f"{self.base_url}/files"
def upload_request():
# MEMORY OPTIMIZATION: Use context manager to minimize file handle lifetime
@ -275,7 +277,7 @@ class MistralLoader:
async def _upload_file_async(self, session: aiohttp.ClientSession) -> str:
"""Async file upload with streaming for better memory efficiency."""
url = f"{self.BASE_API_URL}/files"
url = f"{self.base_url}/files"
async def upload_request():
# Create multipart writer for streaming upload
@ -321,7 +323,7 @@ class MistralLoader:
def _get_signed_url(self, file_id: str) -> str:
"""Retrieves a temporary signed URL for the uploaded file (sync version)."""
log.info(f"Getting signed URL for file ID: {file_id}")
url = f"{self.BASE_API_URL}/files/{file_id}/url"
url = f"{self.base_url}/files/{file_id}/url"
params = {"expiry": 1}
signed_url_headers = {**self.headers, "Accept": "application/json"}
@ -346,7 +348,7 @@ class MistralLoader:
self, session: aiohttp.ClientSession, file_id: str
) -> str:
"""Async signed URL retrieval."""
url = f"{self.BASE_API_URL}/files/{file_id}/url"
url = f"{self.base_url}/files/{file_id}/url"
params = {"expiry": 1}
headers = {**self.headers, "Accept": "application/json"}
@ -373,7 +375,7 @@ class MistralLoader:
def _process_ocr(self, signed_url: str) -> Dict[str, Any]:
"""Sends the signed URL to the OCR endpoint for processing (sync version)."""
log.info("Processing OCR via Mistral API")
url = f"{self.BASE_API_URL}/ocr"
url = f"{self.base_url}/ocr"
ocr_headers = {
**self.headers,
"Content-Type": "application/json",
@ -407,7 +409,7 @@ class MistralLoader:
self, session: aiohttp.ClientSession, signed_url: str
) -> Dict[str, Any]:
"""Async OCR processing with timing metrics."""
url = f"{self.BASE_API_URL}/ocr"
url = f"{self.base_url}/ocr"
headers = {
**self.headers,
@ -446,7 +448,7 @@ class MistralLoader:
def _delete_file(self, file_id: str) -> None:
"""Deletes the file from Mistral storage (sync version)."""
log.info(f"Deleting uploaded file ID: {file_id}")
url = f"{self.BASE_API_URL}/files/{file_id}"
url = f"{self.base_url}/files/{file_id}"
try:
response = requests.delete(
@ -467,7 +469,7 @@ class MistralLoader:
async def delete_request():
self._debug_log(f"Deleting file ID: {file_id}")
async with session.delete(
url=f"{self.BASE_API_URL}/files/{file_id}",
url=f"{self.base_url}/files/{file_id}",
headers=self.headers,
timeout=aiohttp.ClientTimeout(
total=self.cleanup_timeout
@ -507,6 +509,7 @@ class MistralLoader:
timeout=timeout,
headers={"User-Agent": "OpenWebUI-MistralLoader/2.0"},
raise_for_status=False, # We handle status codes manually
trust_env=True,
) as session:
yield session

20
backend/open_webui/retrieval/loaders/youtube.py
View file

@ -83,6 +83,7 @@ class YoutubeLoader:
TranscriptsDisabled,
YouTubeTranscriptApi,
)
from youtube_transcript_api.proxies import GenericProxyConfig
except ImportError:
raise ImportError(
'Could not import "youtube_transcript_api" Python package. '
@ -90,18 +91,16 @@ class YoutubeLoader:
)
if self.proxy_url:
youtube_proxies = {
"http": self.proxy_url,
"https": self.proxy_url,
}
youtube_proxies = GenericProxyConfig(
http_url=self.proxy_url, https_url=self.proxy_url
)
log.debug(f"Using proxy URL: {self.proxy_url[:14]}...")
else:
youtube_proxies = None
transcript_api = YouTubeTranscriptApi(proxy_config=youtube_proxies)
try:
transcript_list = YouTubeTranscriptApi.list_transcripts(
self.video_id, proxies=youtube_proxies
)
transcript_list = transcript_api.list(self.video_id)
except Exception as e:
log.exception("Loading YouTube transcript failed")
return []
@ -158,3 +157,10 @@ class YoutubeLoader:
f"No transcript found for any of the specified languages: {languages_tried}. Verify if the video has transcripts, add more languages if needed."
)
raise NoTranscriptFound(self.video_id, self.language, list(transcript_list))
async def aload(self) -> Generator[Document, None, None]:
"""Asynchronously load YouTube transcripts into `Document` objects."""
import asyncio
loop = asyncio.get_event_loop()
return await loop.run_in_executor(None, self.load)

22
backend/open_webui/retrieval/models/external.py
View file

@ -1,9 +1,12 @@
import logging
import requests
from typing import Optional, List, Tuple
from urllib.parse import quote
from open_webui.env import SRC_LOG_LEVELS
from open_webui.env import ENABLE_FORWARD_USER_INFO_HEADERS, SRC_LOG_LEVELS
from open_webui.retrieval.models.base_reranker import BaseReranker
from open_webui.utils.headers import include_user_info_headers
log = logging.getLogger(__name__)
@ -21,7 +24,9 @@ class ExternalReranker(BaseReranker):
self.url = url
self.model = model
def predict(self, sentences: List[Tuple[str, str]]) -> Optional[List[float]]:
def predict(
self, sentences: List[Tuple[str, str]], user=None
) -> Optional[List[float]]:
query = sentences[0][0]
docs = [i[1] for i in sentences]
@ -36,12 +41,17 @@ class ExternalReranker(BaseReranker):
log.info(f"ExternalReranker:predict:model {self.model}")
log.info(f"ExternalReranker:predict:query {query}")
headers = {
"Content-Type": "application/json",
"Authorization": f"Bearer {self.api_key}",
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user:
headers = include_user_info_headers(headers, user)
r = requests.post(
f"{self.url}",
headers={
"Content-Type": "application/json",
"Authorization": f"Bearer {self.api_key}",
},
headers=headers,
json=payload,
)

1159
backend/open_webui/retrieval/utils.py
View file

File diff suppressed because it is too large Load diff

6
backend/open_webui/retrieval/vector/dbs/chroma.py
View file

@ -11,6 +11,8 @@ from open_webui.retrieval.vector.main import (
SearchResult,
GetResult,
)
from open_webui.retrieval.vector.utils import process_metadata
from open_webui.config import (
CHROMA_DATA_PATH,
CHROMA_HTTP_HOST,
@ -144,7 +146,7 @@ class ChromaClient(VectorDBBase):
ids = [item["id"] for item in items]
documents = [item["text"] for item in items]
embeddings = [item["vector"] for item in items]
metadatas = [item["metadata"] for item in items]
metadatas = [process_metadata(item["metadata"]) for item in items]
for batch in create_batches(
api=self.client,
@ -164,7 +166,7 @@ class ChromaClient(VectorDBBase):
ids = [item["id"] for item in items]
documents = [item["text"] for item in items]
embeddings = [item["vector"] for item in items]
metadatas = [item["metadata"] for item in items]
metadatas = [process_metadata(item["metadata"]) for item in items]
collection.upsert(
ids=ids, documents=documents, embeddings=embeddings, metadatas=metadatas

6
backend/open_webui/retrieval/vector/dbs/elasticsearch.py
View file

@ -2,6 +2,8 @@ from elasticsearch import Elasticsearch, BadRequestError
from typing import Optional
import ssl
from elasticsearch.helpers import bulk, scan
from open_webui.retrieval.vector.utils import process_metadata
from open_webui.retrieval.vector.main import (
VectorDBBase,
VectorItem,
@ -243,7 +245,7 @@ class ElasticsearchClient(VectorDBBase):
"collection": collection_name,
"vector": item["vector"],
"text": item["text"],
"metadata": item["metadata"],
"metadata": process_metadata(item["metadata"]),
},
}
for item in batch
@ -264,7 +266,7 @@ class ElasticsearchClient(VectorDBBase):
"collection": collection_name,
"vector": item["vector"],
"text": item["text"],
"metadata": item["metadata"],
"metadata": process_metadata(item["metadata"]),
},
"doc_as_upsert": True,
}

96
backend/open_webui/retrieval/vector/dbs/milvus.py
View file

@ -1,8 +1,12 @@
from pymilvus import MilvusClient as Client
from pymilvus import FieldSchema, DataType
from pymilvus import connections, Collection
import json
import logging
from typing import Optional
from open_webui.retrieval.vector.utils import process_metadata
from open_webui.retrieval.vector.main import (
VectorDBBase,
VectorItem,
@ -18,6 +22,8 @@ from open_webui.config import (
MILVUS_HNSW_M,
MILVUS_HNSW_EFCONSTRUCTION,
MILVUS_IVF_FLAT_NLIST,
MILVUS_DISKANN_MAX_DEGREE,
MILVUS_DISKANN_SEARCH_LIST_SIZE,
)
from open_webui.env import SRC_LOG_LEVELS
@ -127,12 +133,18 @@ class MilvusClient(VectorDBBase):
elif index_type == "IVF_FLAT":
index_creation_params = {"nlist": MILVUS_IVF_FLAT_NLIST}
log.info(f"IVF_FLAT params: {index_creation_params}")
elif index_type == "DISKANN":
index_creation_params = {
"max_degree": MILVUS_DISKANN_MAX_DEGREE,
"search_list_size": MILVUS_DISKANN_SEARCH_LIST_SIZE,
}
log.info(f"DISKANN params: {index_creation_params}")
elif index_type in ["FLAT", "AUTOINDEX"]:
log.info(f"Using {index_type} index with no specific build-time params.")
else:
log.warning(
f"Unsupported MILVUS_INDEX_TYPE: '{index_type}'. "
f"Supported types: HNSW, IVF_FLAT, FLAT, AUTOINDEX. "
f"Supported types: HNSW, IVF_FLAT, DISKANN, FLAT, AUTOINDEX. "
f"Milvus will use its default for the collection if this type is not directly supported for index creation."
)
# For unsupported types, pass the type directly to Milvus; it might handle it or use a default.
@ -185,7 +197,9 @@ class MilvusClient(VectorDBBase):
)
return self._result_to_search_result(result)
def query(self, collection_name: str, filter: dict, limit: Optional[int] = None):
def query(self, collection_name: str, filter: dict, limit: int = -1):
connections.connect(uri=MILVUS_URI, token=MILVUS_TOKEN, db_name=MILVUS_DB)
# Construct the filter string for querying
collection_name = collection_name.replace("-", "_")
if not self.has_collection(collection_name):
@ -199,72 +213,36 @@ class MilvusClient(VectorDBBase):
for key, value in filter.items()
]
)
max_limit = 16383 # The maximum number of records per request
all_results = []
if limit is None:
# Milvus default limit for query if not specified is 16384, but docs mention iteration.
# Let's set a practical high number if "all" is intended, or handle true pagination.
# For now, if limit is None, we'll fetch in batches up to a very large number.
# This part could be refined based on expected use cases for "get all".
# For this function signature, None implies "as many as possible" up to Milvus limits.
limit = (
16384 * 10
) # A large number to signify fetching many, will be capped by actual data or max_limit per call.
log.info(
f"Limit not specified for query, fetching up to {limit} results in batches."
)
# Initialize offset and remaining to handle pagination
offset = 0
remaining = limit
collection = Collection(f"{self.collection_prefix}_{collection_name}")
collection.load()
all_results = []
try:
log.info(
f"Querying collection {self.collection_prefix}_{collection_name} with filter: '{filter_string}', limit: {limit}"
)
# Loop until there are no more items to fetch or the desired limit is reached
while remaining > 0:
current_fetch = min(
max_limit, remaining if isinstance(remaining, int) else max_limit
)
log.debug(
f"Querying with offset: {offset}, current_fetch: {current_fetch}"
)
results = self.client.query(
collection_name=f"{self.collection_prefix}_{collection_name}",
filter=filter_string,
output_fields=[
"id",
"data",
"metadata",
], # Explicitly list needed fields. Vector not usually needed in query.
limit=current_fetch,
offset=offset,
)
iterator = collection.query_iterator(
filter=filter_string,
output_fields=[
"id",
"data",
"metadata",
],
limit=limit, # Pass the limit directly; -1 means no limit.
)
if not results:
log.debug("No more results from query.")
break
all_results.extend(results)
results_count = len(results)
log.debug(f"Fetched {results_count} results in this batch.")
if isinstance(remaining, int):
remaining -= results_count
offset += results_count
# Break the loop if the results returned are less than the requested fetch count (means end of data)
if results_count < current_fetch:
log.debug(
"Fetched less than requested, assuming end of results for this query."
)
while True:
result = iterator.next()
if not result:
iterator.close()
break
all_results += result
log.info(f"Total results from query: {len(all_results)}")
return self._result_to_get_result([all_results])
except Exception as e:
log.exception(
f"Error querying collection {self.collection_prefix}_{collection_name} with filter '{filter_string}' and limit {limit}: {e}"
@ -279,7 +257,7 @@ class MilvusClient(VectorDBBase):
)
# Using query with a trivial filter to get all items.
# This will use the paginated query logic.
return self.query(collection_name=collection_name, filter={}, limit=None)
return self.query(collection_name=collection_name, filter={}, limit=-1)
def insert(self, collection_name: str, items: list[VectorItem]):
# Insert the items into the collection, if the collection does not exist, it will be created.
@ -311,7 +289,7 @@ class MilvusClient(VectorDBBase):
"id": item["id"],
"vector": item["vector"],
"data": {"text": item["text"]},
"metadata": item["metadata"],
"metadata": process_metadata(item["metadata"]),
}
for item in items
],
@ -347,7 +325,7 @@ class MilvusClient(VectorDBBase):
"id": item["id"],
"vector": item["vector"],
"data": {"text": item["text"]},
"metadata": item["metadata"],
"metadata": process_metadata(item["metadata"]),
}
for item in items
],

282
backend/open_webui/retrieval/vector/dbs/milvus_multitenancy.py Normal file
View file

@ -0,0 +1,282 @@
import logging
from typing import Optional, Tuple, List, Dict, Any
from open_webui.config import (
MILVUS_URI,
MILVUS_TOKEN,
MILVUS_DB,
MILVUS_COLLECTION_PREFIX,
MILVUS_INDEX_TYPE,
MILVUS_METRIC_TYPE,
MILVUS_HNSW_M,
MILVUS_HNSW_EFCONSTRUCTION,
MILVUS_IVF_FLAT_NLIST,
)
from open_webui.env import SRC_LOG_LEVELS
from open_webui.retrieval.vector.main import (
GetResult,
SearchResult,
VectorDBBase,
VectorItem,
)
from pymilvus import (
connections,
utility,
Collection,
CollectionSchema,
FieldSchema,
DataType,
)
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["RAG"])
RESOURCE_ID_FIELD = "resource_id"
class MilvusClient(VectorDBBase):
def __init__(self):
# Milvus collection names can only contain numbers, letters, and underscores.
self.collection_prefix = MILVUS_COLLECTION_PREFIX.replace("-", "_")
connections.connect(
alias="default",
uri=MILVUS_URI,
token=MILVUS_TOKEN,
db_name=MILVUS_DB,
)
# Main collection types for multi-tenancy
self.MEMORY_COLLECTION = f"{self.collection_prefix}_memories"
self.KNOWLEDGE_COLLECTION = f"{self.collection_prefix}_knowledge"
self.FILE_COLLECTION = f"{self.collection_prefix}_files"
self.WEB_SEARCH_COLLECTION = f"{self.collection_prefix}_web_search"
self.HASH_BASED_COLLECTION = f"{self.collection_prefix}_hash_based"
self.shared_collections = [
self.MEMORY_COLLECTION,
self.KNOWLEDGE_COLLECTION,
self.FILE_COLLECTION,
self.WEB_SEARCH_COLLECTION,
self.HASH_BASED_COLLECTION,
]
def _get_collection_and_resource_id(self, collection_name: str) -> Tuple[str, str]:
"""
Maps the traditional collection name to multi-tenant collection and resource ID.
WARNING: This mapping relies on current Open WebUI naming conventions for
collection names. If Open WebUI changes how it generates collection names
(e.g., "user-memory-" prefix, "file-" prefix, web search patterns, or hash
formats), this mapping will break and route data to incorrect collections.
POTENTIALLY CAUSING HUGE DATA CORRUPTION, DATA CONSISTENCY ISSUES AND INCORRECT
DATA MAPPING INSIDE THE DATABASE.
"""
resource_id = collection_name
if collection_name.startswith("user-memory-"):
return self.MEMORY_COLLECTION, resource_id
elif collection_name.startswith("file-"):
return self.FILE_COLLECTION, resource_id
elif collection_name.startswith("web-search-"):
return self.WEB_SEARCH_COLLECTION, resource_id
elif len(collection_name) == 63 and all(
c in "0123456789abcdef" for c in collection_name
):
return self.HASH_BASED_COLLECTION, resource_id
else:
return self.KNOWLEDGE_COLLECTION, resource_id
def _create_shared_collection(self, mt_collection_name: str, dimension: int):
fields = [
FieldSchema(
name="id",
dtype=DataType.VARCHAR,
is_primary=True,
auto_id=False,
max_length=36,
),
FieldSchema(name="vector", dtype=DataType.FLOAT_VECTOR, dim=dimension),
FieldSchema(name="text", dtype=DataType.VARCHAR, max_length=65535),
FieldSchema(name="metadata", dtype=DataType.JSON),
FieldSchema(name=RESOURCE_ID_FIELD, dtype=DataType.VARCHAR, max_length=255),
]
schema = CollectionSchema(fields, "Shared collection for multi-tenancy")
collection = Collection(mt_collection_name, schema)
index_params = {
"metric_type": MILVUS_METRIC_TYPE,
"index_type": MILVUS_INDEX_TYPE,
"params": {},
}
if MILVUS_INDEX_TYPE == "HNSW":
index_params["params"] = {
"M": MILVUS_HNSW_M,
"efConstruction": MILVUS_HNSW_EFCONSTRUCTION,
}
elif MILVUS_INDEX_TYPE == "IVF_FLAT":
index_params["params"] = {"nlist": MILVUS_IVF_FLAT_NLIST}
collection.create_index("vector", index_params)
collection.create_index(RESOURCE_ID_FIELD)
log.info(f"Created shared collection: {mt_collection_name}")
return collection
def _ensure_collection(self, mt_collection_name: str, dimension: int):
if not utility.has_collection(mt_collection_name):
self._create_shared_collection(mt_collection_name, dimension)
def has_collection(self, collection_name: str) -> bool:
mt_collection, resource_id = self._get_collection_and_resource_id(
collection_name
)
if not utility.has_collection(mt_collection):
return False
collection = Collection(mt_collection)
collection.load()
res = collection.query(expr=f"{RESOURCE_ID_FIELD} == '{resource_id}'", limit=1)
return len(res) > 0
def upsert(self, collection_name: str, items: List[VectorItem]):
if not items:
return
mt_collection, resource_id = self._get_collection_and_resource_id(
collection_name
)
dimension = len(items[0]["vector"])
self._ensure_collection(mt_collection, dimension)
collection = Collection(mt_collection)
entities = [
{
"id": item["id"],
"vector": item["vector"],
"text": item["text"],
"metadata": item["metadata"],
RESOURCE_ID_FIELD: resource_id,
}
for item in items
]
collection.insert(entities)
collection.flush()
def search(
self, collection_name: str, vectors: List[List[float]], limit: int
) -> Optional[SearchResult]:
if not vectors:
return None
mt_collection, resource_id = self._get_collection_and_resource_id(
collection_name
)
if not utility.has_collection(mt_collection):
return None
collection = Collection(mt_collection)
collection.load()
search_params = {"metric_type": MILVUS_METRIC_TYPE, "params": {}}
results = collection.search(
data=vectors,
anns_field="vector",
param=search_params,
limit=limit,
expr=f"{RESOURCE_ID_FIELD} == '{resource_id}'",
output_fields=["id", "text", "metadata"],
)
ids, documents, metadatas, distances = [], [], [], []
for hits in results:
batch_ids, batch_docs, batch_metadatas, batch_dists = [], [], [], []
for hit in hits:
batch_ids.append(hit.entity.get("id"))
batch_docs.append(hit.entity.get("text"))
batch_metadatas.append(hit.entity.get("metadata"))
batch_dists.append(hit.distance)
ids.append(batch_ids)
documents.append(batch_docs)
metadatas.append(batch_metadatas)
distances.append(batch_dists)
return SearchResult(
ids=ids, documents=documents, metadatas=metadatas, distances=distances
)
def delete(
self,
collection_name: str,
ids: Optional[List[str]] = None,
filter: Optional[Dict[str, Any]] = None,
):
mt_collection, resource_id = self._get_collection_and_resource_id(
collection_name
)
if not utility.has_collection(mt_collection):
return
collection = Collection(mt_collection)
# Build expression
expr = [f"{RESOURCE_ID_FIELD} == '{resource_id}'"]
if ids:
# Milvus expects a string list for 'in' operator
id_list_str = ", ".join([f"'{id_val}'" for id_val in ids])
expr.append(f"id in [{id_list_str}]")
if filter:
for key, value in filter.items():
expr.append(f"metadata['{key}'] == '{value}'")
collection.delete(" and ".join(expr))
def reset(self):
for collection_name in self.shared_collections:
if utility.has_collection(collection_name):
utility.drop_collection(collection_name)
def delete_collection(self, collection_name: str):
mt_collection, resource_id = self._get_collection_and_resource_id(
collection_name
)
if not utility.has_collection(mt_collection):
return
collection = Collection(mt_collection)
collection.delete(f"{RESOURCE_ID_FIELD} == '{resource_id}'")
def query(
self, collection_name: str, filter: Dict[str, Any], limit: Optional[int] = None
) -> Optional[GetResult]:
mt_collection, resource_id = self._get_collection_and_resource_id(
collection_name
)
if not utility.has_collection(mt_collection):
return None
collection = Collection(mt_collection)
collection.load()
expr = [f"{RESOURCE_ID_FIELD} == '{resource_id}'"]
if filter:
for key, value in filter.items():
if isinstance(value, str):
expr.append(f"metadata['{key}'] == '{value}'")
else:
expr.append(f"metadata['{key}'] == {value}")
results = collection.query(
expr=" and ".join(expr),
output_fields=["id", "text", "metadata"],
limit=limit,
)
ids = [res["id"] for res in results]
documents = [res["text"] for res in results]
metadatas = [res["metadata"] for res in results]
return GetResult(ids=[ids], documents=[documents], metadatas=[metadatas])
def get(self, collection_name: str) -> Optional[GetResult]:
return self.query(collection_name, filter={}, limit=None)
def insert(self, collection_name: str, items: List[VectorItem]):
return self.upsert(collection_name, items)

14
backend/open_webui/retrieval/vector/dbs/opensearch.py
View file

@ -2,6 +2,7 @@ from opensearchpy import OpenSearch
from opensearchpy.helpers import bulk
from typing import Optional
from open_webui.retrieval.vector.utils import process_metadata
from open_webui.retrieval.vector.main import (
VectorDBBase,
VectorItem,
@ -157,10 +158,10 @@ class OpenSearchClient(VectorDBBase):
for field, value in filter.items():
query_body["query"]["bool"]["filter"].append(
{"match": {"metadata." + str(field): value}}
{"term": {"metadata." + str(field) + ".keyword": value}}
)
size = limit if limit else 10
size = limit if limit else 10000
try:
result = self.client.search(
@ -200,12 +201,13 @@ class OpenSearchClient(VectorDBBase):
"_source": {
"vector": item["vector"],
"text": item["text"],
"metadata": item["metadata"],
"metadata": process_metadata(item["metadata"]),
},
}
for item in batch
]
bulk(self.client, actions)
self.client.indices.refresh(self._get_index_name(collection_name))
def upsert(self, collection_name: str, items: list[VectorItem]):
self._create_index_if_not_exists(
@ -221,13 +223,14 @@ class OpenSearchClient(VectorDBBase):
"doc": {
"vector": item["vector"],
"text": item["text"],
"metadata": item["metadata"],
"metadata": process_metadata(item["metadata"]),
},
"doc_as_upsert": True,
}
for item in batch
]
bulk(self.client, actions)
self.client.indices.refresh(self._get_index_name(collection_name))
def delete(
self,
@ -251,11 +254,12 @@ class OpenSearchClient(VectorDBBase):
}
for field, value in filter.items():
query_body["query"]["bool"]["filter"].append(
{"match": {"metadata." + str(field): value}}
{"term": {"metadata." + str(field) + ".keyword": value}}
)
self.client.delete_by_query(
index=self._get_index_name(collection_name), body=query_body
)
self.client.indices.refresh(self._get_index_name(collection_name))
def reset(self):
indices = self.client.indices.get(index=f"{self.index_prefix}_*")

943
backend/open_webui/retrieval/vector/dbs/oracle23ai.py Normal file
View file

@ -0,0 +1,943 @@
"""
Oracle 23ai Vector Database Client - Fixed Version
# .env
VECTOR_DB = "oracle23ai"
## DBCS or oracle 23ai free
ORACLE_DB_USE_WALLET = false
ORACLE_DB_USER = "DEMOUSER"
ORACLE_DB_PASSWORD = "Welcome123456"
ORACLE_DB_DSN = "localhost:1521/FREEPDB1"
## ADW or ATP
# ORACLE_DB_USE_WALLET = true
# ORACLE_DB_USER = "DEMOUSER"
# ORACLE_DB_PASSWORD = "Welcome123456"
# ORACLE_DB_DSN = "medium"
# ORACLE_DB_DSN = "(description= (retry_count=3)(retry_delay=3)(address=(protocol=tcps)(port=1522)(host=xx.oraclecloud.com))(connect_data=(service_name=yy.adb.oraclecloud.com))(security=(ssl_server_dn_match=no)))"
# ORACLE_WALLET_DIR = "/home/opc/adb_wallet"
# ORACLE_WALLET_PASSWORD = "Welcome1"
ORACLE_VECTOR_LENGTH = 768
ORACLE_DB_POOL_MIN = 2
ORACLE_DB_POOL_MAX = 10
ORACLE_DB_POOL_INCREMENT = 1
"""
from typing import Optional, List, Dict, Any, Union
from decimal import Decimal
import logging
import os
import threading
import time
import json
import array
import oracledb
from open_webui.retrieval.vector.main import (
VectorDBBase,
VectorItem,
SearchResult,
GetResult,
)
from open_webui.config import (
ORACLE_DB_USE_WALLET,
ORACLE_DB_USER,
ORACLE_DB_PASSWORD,
ORACLE_DB_DSN,
ORACLE_WALLET_DIR,
ORACLE_WALLET_PASSWORD,
ORACLE_VECTOR_LENGTH,
ORACLE_DB_POOL_MIN,
ORACLE_DB_POOL_MAX,
ORACLE_DB_POOL_INCREMENT,
)
from open_webui.env import SRC_LOG_LEVELS
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["RAG"])
class Oracle23aiClient(VectorDBBase):
"""
Oracle Vector Database Client for vector similarity search using Oracle Database 23ai.
This client provides an interface to store, retrieve, and search vector embeddings
in an Oracle database. It uses connection pooling for efficient database access
and supports vector similarity search operations.
Attributes:
pool: Connection pool for Oracle database connections
"""
def __init__(self) -> None:
"""
Initialize the Oracle23aiClient with a connection pool.
Creates a connection pool with configurable min/max connections, initializes
the database schema if needed, and sets up necessary tables and indexes.
Raises:
ValueError: If required configuration parameters are missing
Exception: If database initialization fails
"""
self.pool = None
try:
# Create the appropriate connection pool based on DB type
if ORACLE_DB_USE_WALLET:
self._create_adb_pool()
else: # DBCS
self._create_dbcs_pool()
dsn = ORACLE_DB_DSN
log.info(f"Creating Connection Pool [{ORACLE_DB_USER}:**@{dsn}]")
with self.get_connection() as connection:
log.info(f"Connection version: {connection.version}")
self._initialize_database(connection)
log.info("Oracle Vector Search initialization complete.")
except Exception as e:
log.exception(f"Error during Oracle Vector Search initialization: {e}")
raise
def _create_adb_pool(self) -> None:
"""
Create connection pool for Oracle Autonomous Database.
Uses wallet-based authentication.
"""
self.pool = oracledb.create_pool(
user=ORACLE_DB_USER,
password=ORACLE_DB_PASSWORD,
dsn=ORACLE_DB_DSN,
min=ORACLE_DB_POOL_MIN,
max=ORACLE_DB_POOL_MAX,
increment=ORACLE_DB_POOL_INCREMENT,
config_dir=ORACLE_WALLET_DIR,
wallet_location=ORACLE_WALLET_DIR,
wallet_password=ORACLE_WALLET_PASSWORD,
)
log.info("Created ADB connection pool with wallet authentication.")
def _create_dbcs_pool(self) -> None:
"""
Create connection pool for Oracle Database Cloud Service.
Uses basic authentication without wallet.
"""
self.pool = oracledb.create_pool(
user=ORACLE_DB_USER,
password=ORACLE_DB_PASSWORD,
dsn=ORACLE_DB_DSN,
min=ORACLE_DB_POOL_MIN,
max=ORACLE_DB_POOL_MAX,
increment=ORACLE_DB_POOL_INCREMENT,
)
log.info("Created DB connection pool with basic authentication.")
def get_connection(self):
"""
Acquire a connection from the connection pool with retry logic.
Returns:
connection: A database connection with output type handler configured
"""
max_retries = 3
for attempt in range(max_retries):
try:
connection = self.pool.acquire()
connection.outputtypehandler = self._output_type_handler
return connection
except oracledb.DatabaseError as e:
(error_obj,) = e.args
log.exception(
f"Connection attempt {attempt + 1} failed: {error_obj.message}"
)
if attempt < max_retries - 1:
wait_time = 2**attempt
log.info(f"Retrying in {wait_time} seconds...")
time.sleep(wait_time)
else:
raise
def start_health_monitor(self, interval_seconds: int = 60):
"""
Start a background thread to periodically check the health of the connection pool.
Args:
interval_seconds (int): Number of seconds between health checks
"""
def _monitor():
while True:
try:
log.info("[HealthCheck] Running periodic DB health check...")
self.ensure_connection()
log.info("[HealthCheck] Connection is healthy.")
except Exception as e:
log.exception(f"[HealthCheck] Connection health check failed: {e}")
time.sleep(interval_seconds)
thread = threading.Thread(target=_monitor, daemon=True)
thread.start()
log.info(f"Started DB health monitor every {interval_seconds} seconds.")
def _reconnect_pool(self):
"""
Attempt to reinitialize the connection pool if it's been closed or broken.
"""
try:
log.info("Attempting to reinitialize the Oracle connection pool...")
# Close existing pool if it exists
if self.pool:
try:
self.pool.close()
except Exception as close_error:
log.warning(f"Error closing existing pool: {close_error}")
# Re-create the appropriate connection pool based on DB type
if ORACLE_DB_USE_WALLET:
self._create_adb_pool()
else: # DBCS
self._create_dbcs_pool()
log.info("Connection pool reinitialized.")
except Exception as e:
log.exception(f"Failed to reinitialize the connection pool: {e}")
raise
def ensure_connection(self):
"""
Ensure the database connection is alive, reconnecting pool if needed.
"""
try:
with self.get_connection() as connection:
with connection.cursor() as cursor:
cursor.execute("SELECT 1 FROM dual")
except Exception as e:
log.exception(
f"Connection check failed: {e}, attempting to reconnect pool..."
)
self._reconnect_pool()
def _output_type_handler(self, cursor, metadata):
"""
Handle Oracle vector type conversion.
Args:
cursor: Oracle database cursor
metadata: Metadata for the column
Returns:
A variable with appropriate conversion for vector types
"""
if metadata.type_code is oracledb.DB_TYPE_VECTOR:
return cursor.var(
metadata.type_code, arraysize=cursor.arraysize, outconverter=list
)
def _initialize_database(self, connection) -> None:
"""
Initialize database schema, tables and indexes.
Creates the document_chunk table and necessary indexes if they don't exist.
Args:
connection: Oracle database connection
Raises:
Exception: If schema initialization fails
"""
with connection.cursor() as cursor:
try:
log.info("Creating Table document_chunk")
cursor.execute(
"""
BEGIN
EXECUTE IMMEDIATE '
CREATE TABLE IF NOT EXISTS document_chunk (
id VARCHAR2(255) PRIMARY KEY,
collection_name VARCHAR2(255) NOT NULL,
text CLOB,
vmetadata JSON,
vector vector(*, float32)
)
';
EXCEPTION
WHEN OTHERS THEN
IF SQLCODE != -955 THEN
RAISE;
END IF;
END;
"""
)
log.info("Creating Index document_chunk_collection_name_idx")
cursor.execute(
"""
BEGIN
EXECUTE IMMEDIATE '
CREATE INDEX IF NOT EXISTS document_chunk_collection_name_idx
ON document_chunk (collection_name)
';
EXCEPTION
WHEN OTHERS THEN
IF SQLCODE != -955 THEN
RAISE;
END IF;
END;
"""
)
log.info("Creating VECTOR INDEX document_chunk_vector_ivf_idx")
cursor.execute(
"""
BEGIN
EXECUTE IMMEDIATE '
CREATE VECTOR INDEX IF NOT EXISTS document_chunk_vector_ivf_idx
ON document_chunk(vector)
ORGANIZATION NEIGHBOR PARTITIONS
DISTANCE COSINE
WITH TARGET ACCURACY 95
PARAMETERS (TYPE IVF, NEIGHBOR PARTITIONS 100)
';
EXCEPTION
WHEN OTHERS THEN
IF SQLCODE != -955 THEN
RAISE;
END IF;
END;
"""
)
connection.commit()
log.info("Database initialization completed successfully.")
except Exception as e:
connection.rollback()
log.exception(f"Error during database initialization: {e}")
raise
def check_vector_length(self) -> None:
"""
Check vector length compatibility (placeholder).
This method would check if the configured vector length matches the database schema.
Currently implemented as a placeholder.
"""
pass
def _vector_to_blob(self, vector: List[float]) -> bytes:
"""
Convert a vector to Oracle BLOB format.
Args:
vector (List[float]): The vector to convert
Returns:
bytes: The vector in Oracle BLOB format
"""
return array.array("f", vector)
def adjust_vector_length(self, vector: List[float]) -> List[float]:
"""
Adjust vector to the expected length if needed.
Args:
vector (List[float]): The vector to adjust
Returns:
List[float]: The adjusted vector
"""
return vector
def _decimal_handler(self, obj):
"""
Handle Decimal objects for JSON serialization.
Args:
obj: Object to serialize
Returns:
float: Converted decimal value
Raises:
TypeError: If object is not JSON serializable
"""
if isinstance(obj, Decimal):
return float(obj)
raise TypeError(f"{obj} is not JSON serializable")
def _metadata_to_json(self, metadata: Dict) -> str:
"""
Convert metadata dictionary to JSON string.
Args:
metadata (Dict): Metadata dictionary
Returns:
str: JSON representation of metadata
"""
return json.dumps(metadata, default=self._decimal_handler) if metadata else "{}"
def _json_to_metadata(self, json_str: str) -> Dict:
"""
Convert JSON string to metadata dictionary.
Args:
json_str (str): JSON string
Returns:
Dict: Metadata dictionary
"""
return json.loads(json_str) if json_str else {}
def insert(self, collection_name: str, items: List[VectorItem]) -> None:
"""
Insert vector items into the database.
Args:
collection_name (str): Name of the collection
items (List[VectorItem]): List of vector items to insert
Raises:
Exception: If insertion fails
Example:
>>> client = Oracle23aiClient()
>>> items = [
... {"id": "1", "text": "Sample text", "vector": [0.1, 0.2, ...], "metadata": {"source": "doc1"}},
... {"id": "2", "text": "Another text", "vector": [0.3, 0.4, ...], "metadata": {"source": "doc2"}}
... ]
>>> client.insert("my_collection", items)
"""
log.info(f"Inserting {len(items)} items into collection '{collection_name}'.")
with self.get_connection() as connection:
try:
with connection.cursor() as cursor:
for item in items:
vector_blob = self._vector_to_blob(item["vector"])
metadata_json = self._metadata_to_json(item["metadata"])
cursor.execute(
"""
INSERT INTO document_chunk
(id, collection_name, text, vmetadata, vector)
VALUES (:id, :collection_name, :text, :metadata, :vector)
""",
{
"id": item["id"],
"collection_name": collection_name,
"text": item["text"],
"metadata": metadata_json,
"vector": vector_blob,
},
)
connection.commit()
log.info(
f"Successfully inserted {len(items)} items into collection '{collection_name}'."
)
except Exception as e:
connection.rollback()
log.exception(f"Error during insert: {e}")
raise
def upsert(self, collection_name: str, items: List[VectorItem]) -> None:
"""
Update or insert vector items into the database.
If an item with the same ID exists, it will be updated;
otherwise, it will be inserted.
Args:
collection_name (str): Name of the collection
items (List[VectorItem]): List of vector items to upsert
Raises:
Exception: If upsert operation fails
Example:
>>> client = Oracle23aiClient()
>>> items = [
... {"id": "1", "text": "Updated text", "vector": [0.1, 0.2, ...], "metadata": {"source": "doc1"}},
... {"id": "3", "text": "New item", "vector": [0.5, 0.6, ...], "metadata": {"source": "doc3"}}
... ]
>>> client.upsert("my_collection", items)
"""
log.info(f"Upserting {len(items)} items into collection '{collection_name}'.")
with self.get_connection() as connection:
try:
with connection.cursor() as cursor:
for item in items:
vector_blob = self._vector_to_blob(item["vector"])
metadata_json = self._metadata_to_json(item["metadata"])
cursor.execute(
"""
MERGE INTO document_chunk d
USING (SELECT :merge_id as id FROM dual) s
ON (d.id = s.id)
WHEN MATCHED THEN
UPDATE SET
collection_name = :upd_collection_name,
text = :upd_text,
vmetadata = :upd_metadata,
vector = :upd_vector
WHEN NOT MATCHED THEN
INSERT (id, collection_name, text, vmetadata, vector)
VALUES (:ins_id, :ins_collection_name, :ins_text, :ins_metadata, :ins_vector)
""",
{
"merge_id": item["id"],
"upd_collection_name": collection_name,
"upd_text": item["text"],
"upd_metadata": metadata_json,
"upd_vector": vector_blob,
"ins_id": item["id"],
"ins_collection_name": collection_name,
"ins_text": item["text"],
"ins_metadata": metadata_json,
"ins_vector": vector_blob,
},
)
connection.commit()
log.info(
f"Successfully upserted {len(items)} items into collection '{collection_name}'."
)
except Exception as e:
connection.rollback()
log.exception(f"Error during upsert: {e}")
raise
def search(
self, collection_name: str, vectors: List[List[Union[float, int]]], limit: int
) -> Optional[SearchResult]:
"""
Search for similar vectors in the database.
Performs vector similarity search using cosine distance.
Args:
collection_name (str): Name of the collection to search
vectors (List[List[Union[float, int]]]): Query vectors to find similar items for
limit (int): Maximum number of results to return per query
Returns:
Optional[SearchResult]: Search results containing ids, distances, documents, and metadata
Example:
>>> client = Oracle23aiClient()
>>> query_vector = [0.1, 0.2, 0.3, ...] # Must match VECTOR_LENGTH
>>> results = client.search("my_collection", [query_vector], limit=5)
>>> if results:
... log.info(f"Found {len(results.ids[0])} matches")
... for i, (id, dist) in enumerate(zip(results.ids[0], results.distances[0])):
... log.info(f"Match {i+1}: id={id}, distance={dist}")
"""
log.info(
f"Searching items from collection '{collection_name}' with limit {limit}."
)
try:
if not vectors:
log.warning("No vectors provided for search.")
return None
num_queries = len(vectors)
ids = [[] for _ in range(num_queries)]
distances = [[] for _ in range(num_queries)]
documents = [[] for _ in range(num_queries)]
metadatas = [[] for _ in range(num_queries)]
with self.get_connection() as connection:
with connection.cursor() as cursor:
for qid, vector in enumerate(vectors):
vector_blob = self._vector_to_blob(vector)
cursor.execute(
"""
SELECT dc.id, dc.text,
JSON_SERIALIZE(dc.vmetadata RETURNING VARCHAR2(4096)) as vmetadata,
VECTOR_DISTANCE(dc.vector, :query_vector, COSINE) as distance
FROM document_chunk dc
WHERE dc.collection_name = :collection_name
ORDER BY VECTOR_DISTANCE(dc.vector, :query_vector, COSINE)
FETCH APPROX FIRST :limit ROWS ONLY
""",
{
"query_vector": vector_blob,
"collection_name": collection_name,
"limit": limit,
},
)
results = cursor.fetchall()
for row in results:
ids[qid].append(row[0])
documents[qid].append(
row[1].read()
if isinstance(row[1], oracledb.LOB)
else str(row[1])
)
# 🔧 FIXED: Parse JSON metadata properly
metadata_str = (
row[2].read()
if isinstance(row[2], oracledb.LOB)
else row[2]
)
metadatas[qid].append(self._json_to_metadata(metadata_str))
distances[qid].append(float(row[3]))
log.info(
f"Search completed. Found {sum(len(ids[i]) for i in range(num_queries))} total results."
)
return SearchResult(
ids=ids, distances=distances, documents=documents, metadatas=metadatas
)
except Exception as e:
log.exception(f"Error during search: {e}")
return None
def query(
self, collection_name: str, filter: Dict, limit: Optional[int] = None
) -> Optional[GetResult]:
"""
Query items based on metadata filters.
Retrieves items that match specified metadata criteria.
Args:
collection_name (str): Name of the collection to query
filter (Dict[str, Any]): Metadata filters to apply
limit (Optional[int]): Maximum number of results to return
Returns:
Optional[GetResult]: Query results containing ids, documents, and metadata
Example:
>>> client = Oracle23aiClient()
>>> filter = {"source": "doc1", "category": "finance"}
>>> results = client.query("my_collection", filter, limit=20)
>>> if results:
... print(f"Found {len(results.ids[0])} matching documents")
"""
log.info(f"Querying items from collection '{collection_name}' with filters.")
try:
limit = limit or 100
query = """
SELECT id, text, JSON_SERIALIZE(vmetadata RETURNING VARCHAR2(4096)) as vmetadata
FROM document_chunk
WHERE collection_name = :collection_name
"""
params = {"collection_name": collection_name}
for i, (key, value) in enumerate(filter.items()):
param_name = f"value_{i}"
query += f" AND JSON_VALUE(vmetadata, '$.{key}' RETURNING VARCHAR2(4096)) = :{param_name}"
params[param_name] = str(value)
query += " FETCH FIRST :limit ROWS ONLY"
params["limit"] = limit
with self.get_connection() as connection:
with connection.cursor() as cursor:
cursor.execute(query, params)
results = cursor.fetchall()
if not results:
log.info("No results found for query.")
return None
ids = [[row[0] for row in results]]
documents = [
[
row[1].read() if isinstance(row[1], oracledb.LOB) else str(row[1])
for row in results
]
]
# 🔧 FIXED: Parse JSON metadata properly
metadatas = [
[
self._json_to_metadata(
row[2].read() if isinstance(row[2], oracledb.LOB) else row[2]
)
for row in results
]
]
log.info(f"Query completed. Found {len(results)} results.")
return GetResult(ids=ids, documents=documents, metadatas=metadatas)
except Exception as e:
log.exception(f"Error during query: {e}")
return None
def get(self, collection_name: str) -> Optional[GetResult]:
"""
Get all items in a collection.
Retrieves items from a specified collection up to the limit.
Args:
collection_name (str): Name of the collection to retrieve
limit (Optional[int]): Maximum number of items to retrieve
Returns:
Optional[GetResult]: Result containing ids, documents, and metadata
Example:
>>> client = Oracle23aiClient()
>>> results = client.get("my_collection", limit=50)
>>> if results:
... print(f"Retrieved {len(results.ids[0])} documents from collection")
"""
log.info(
f"Getting items from collection '{collection_name}' with limit {limit}."
)
try:
limit = 1000 # Hardcoded limit for get operation
with self.get_connection() as connection:
with connection.cursor() as cursor:
cursor.execute(
"""
SELECT /*+ MONITOR */ id, text, JSON_SERIALIZE(vmetadata RETURNING VARCHAR2(4096)) as vmetadata
FROM document_chunk
WHERE collection_name = :collection_name
FETCH FIRST :limit ROWS ONLY
""",
{"collection_name": collection_name, "limit": limit},
)
results = cursor.fetchall()
if not results:
log.info("No results found.")
return None
ids = [[row[0] for row in results]]
documents = [
[
row[1].read() if isinstance(row[1], oracledb.LOB) else str(row[1])
for row in results
]
]
# 🔧 FIXED: Parse JSON metadata properly
metadatas = [
[
self._json_to_metadata(
row[2].read() if isinstance(row[2], oracledb.LOB) else row[2]
)
for row in results
]
]
return GetResult(ids=ids, documents=documents, metadatas=metadatas)
except Exception as e:
log.exception(f"Error during get: {e}")
return None
def delete(
self,
collection_name: str,
ids: Optional[List[str]] = None,
filter: Optional[Dict[str, Any]] = None,
) -> None:
"""
Delete items from the database.
Deletes items from a collection based on IDs or metadata filters.
Args:
collection_name (str): Name of the collection to delete from
ids (Optional[List[str]]): Specific item IDs to delete
filter (Optional[Dict[str, Any]]): Metadata filters for deletion
Raises:
Exception: If deletion fails
Example:
>>> client = Oracle23aiClient()
>>> # Delete specific items by ID
>>> client.delete("my_collection", ids=["1", "3", "5"])
>>> # Or delete by metadata filter
>>> client.delete("my_collection", filter={"source": "deprecated_source"})
"""
log.info(f"Deleting items from collection '{collection_name}'.")
try:
query = (
"DELETE FROM document_chunk WHERE collection_name = :collection_name"
)
params = {"collection_name": collection_name}
if ids:
# 🔧 FIXED: Use proper parameterized query to prevent SQL injection
placeholders = ",".join([f":id_{i}" for i in range(len(ids))])
query += f" AND id IN ({placeholders})"
for i, id_val in enumerate(ids):
params[f"id_{i}"] = id_val
if filter:
for i, (key, value) in enumerate(filter.items()):
param_name = f"value_{i}"
query += f" AND JSON_VALUE(vmetadata, '$.{key}' RETURNING VARCHAR2(4096)) = :{param_name}"
params[param_name] = str(value)
with self.get_connection() as connection:
with connection.cursor() as cursor:
cursor.execute(query, params)
deleted = cursor.rowcount
connection.commit()
log.info(f"Deleted {deleted} items from collection '{collection_name}'.")
except Exception as e:
log.exception(f"Error during delete: {e}")
raise
def reset(self) -> None:
"""
Reset the database by deleting all items.
Deletes all items from the document_chunk table.
Raises:
Exception: If reset fails
Example:
>>> client = Oracle23aiClient()
>>> client.reset() # Warning: Removes all data!
"""
log.info("Resetting database - deleting all items.")
try:
with self.get_connection() as connection:
with connection.cursor() as cursor:
cursor.execute("DELETE FROM document_chunk")
deleted = cursor.rowcount
connection.commit()
log.info(
f"Reset complete. Deleted {deleted} items from 'document_chunk' table."
)
except Exception as e:
log.exception(f"Error during reset: {e}")
raise
def close(self) -> None:
"""
Close the database connection pool.
Properly closes the connection pool and releases all resources.
Example:
>>> client = Oracle23aiClient()
>>> # After finishing all operations
>>> client.close()
"""
try:
if hasattr(self, "pool") and self.pool:
self.pool.close()
log.info("Oracle Vector Search connection pool closed.")
except Exception as e:
log.exception(f"Error closing connection pool: {e}")
def has_collection(self, collection_name: str) -> bool:
"""
Check if a collection exists.
Args:
collection_name (str): Name of the collection to check
Returns:
bool: True if the collection exists, False otherwise
Example:
>>> client = Oracle23aiClient()
>>> if client.has_collection("my_collection"):
... print("Collection exists!")
... else:
... print("Collection does not exist.")
"""
try:
with self.get_connection() as connection:
with connection.cursor() as cursor:
cursor.execute(
"""
SELECT COUNT(*)
FROM document_chunk
WHERE collection_name = :collection_name
FETCH FIRST 1 ROWS ONLY
""",
{"collection_name": collection_name},
)
count = cursor.fetchone()[0]
return count > 0
except Exception as e:
log.exception(f"Error checking collection existence: {e}")
return False
def delete_collection(self, collection_name: str) -> None:
"""
Delete an entire collection.
Removes all items belonging to the specified collection.
Args:
collection_name (str): Name of the collection to delete
Example:
>>> client = Oracle23aiClient()
>>> client.delete_collection("obsolete_collection")
"""
log.info(f"Deleting collection '{collection_name}'.")
try:
with self.get_connection() as connection:
with connection.cursor() as cursor:
cursor.execute(
"""
DELETE FROM document_chunk
WHERE collection_name = :collection_name
""",
{"collection_name": collection_name},
)
deleted = cursor.rowcount
connection.commit()
log.info(
f"Collection '{collection_name}' deleted. Removed {deleted} items."
)
except Exception as e:
log.exception(f"Error deleting collection '{collection_name}': {e}")
raise

206
backend/open_webui/retrieval/vector/dbs/pgvector.py
View file

@ -1,4 +1,4 @@
from typing import Optional, List, Dict, Any
from typing import Optional, List, Dict, Any, Tuple
import logging
import json
from sqlalchemy import (
@ -18,14 +18,16 @@ from sqlalchemy import (
values,
)
from sqlalchemy.sql import true
from sqlalchemy.pool import NullPool
from sqlalchemy.pool import NullPool, QueuePool
from sqlalchemy.orm import declarative_base, scoped_session, sessionmaker
from sqlalchemy.dialects.postgresql import JSONB, array
from pgvector.sqlalchemy import Vector
from pgvector.sqlalchemy import Vector, HALFVEC
from sqlalchemy.ext.mutable import MutableDict
from sqlalchemy.exc import NoSuchTableError
from open_webui.retrieval.vector.utils import process_metadata
from open_webui.retrieval.vector.main import (
VectorDBBase,
VectorItem,
@ -35,13 +37,27 @@ from open_webui.retrieval.vector.main import (
from open_webui.config import (
PGVECTOR_DB_URL,
PGVECTOR_INITIALIZE_MAX_VECTOR_LENGTH,
PGVECTOR_CREATE_EXTENSION,
PGVECTOR_PGCRYPTO,
PGVECTOR_PGCRYPTO_KEY,
PGVECTOR_POOL_SIZE,
PGVECTOR_POOL_MAX_OVERFLOW,
PGVECTOR_POOL_TIMEOUT,
PGVECTOR_POOL_RECYCLE,
PGVECTOR_INDEX_METHOD,
PGVECTOR_HNSW_M,
PGVECTOR_HNSW_EF_CONSTRUCTION,
PGVECTOR_IVFFLAT_LISTS,
PGVECTOR_USE_HALFVEC,
)
from open_webui.env import SRC_LOG_LEVELS
VECTOR_LENGTH = PGVECTOR_INITIALIZE_MAX_VECTOR_LENGTH
USE_HALFVEC = PGVECTOR_USE_HALFVEC
VECTOR_TYPE_FACTORY = HALFVEC if USE_HALFVEC else Vector
VECTOR_OPCLASS = "halfvec_cosine_ops" if USE_HALFVEC else "vector_cosine_ops"
Base = declarative_base()
log = logging.getLogger(__name__)
@ -60,7 +76,7 @@ class DocumentChunk(Base):
__tablename__ = "document_chunk"
id = Column(Text, primary_key=True)
vector = Column(Vector(dim=VECTOR_LENGTH), nullable=True)
vector = Column(VECTOR_TYPE_FACTORY(dim=VECTOR_LENGTH), nullable=True)
collection_name = Column(Text, nullable=False)
if PGVECTOR_PGCRYPTO:
@ -80,9 +96,24 @@ class PgvectorClient(VectorDBBase):
self.session = Session
else:
engine = create_engine(
PGVECTOR_DB_URL, pool_pre_ping=True, poolclass=NullPool
)
if isinstance(PGVECTOR_POOL_SIZE, int):
if PGVECTOR_POOL_SIZE > 0:
engine = create_engine(
PGVECTOR_DB_URL,
pool_size=PGVECTOR_POOL_SIZE,
max_overflow=PGVECTOR_POOL_MAX_OVERFLOW,
pool_timeout=PGVECTOR_POOL_TIMEOUT,
pool_recycle=PGVECTOR_POOL_RECYCLE,
pool_pre_ping=True,
poolclass=QueuePool,
)
else:
engine = create_engine(
PGVECTOR_DB_URL, pool_pre_ping=True, poolclass=NullPool
)
else:
engine = create_engine(PGVECTOR_DB_URL, pool_pre_ping=True)
SessionLocal = sessionmaker(
autocommit=False, autoflush=False, bind=engine, expire_on_commit=False
)
@ -90,11 +121,36 @@ class PgvectorClient(VectorDBBase):
try:
# Ensure the pgvector extension is available
self.session.execute(text("CREATE EXTENSION IF NOT EXISTS vector;"))
# Use a conditional check to avoid permission issues on Azure PostgreSQL
if PGVECTOR_CREATE_EXTENSION:
self.session.execute(
text(
"""
DO $$
BEGIN
IF NOT EXISTS (SELECT 1 FROM pg_extension WHERE extname = 'vector') THEN
CREATE EXTENSION IF NOT EXISTS vector;
END IF;
END $$;
"""
)
)
if PGVECTOR_PGCRYPTO:
# Ensure the pgcrypto extension is available for encryption
self.session.execute(text("CREATE EXTENSION IF NOT EXISTS pgcrypto;"))
# Use a conditional check to avoid permission issues on Azure PostgreSQL
self.session.execute(
text(
"""
DO $$
BEGIN
IF NOT EXISTS (SELECT 1 FROM pg_extension WHERE extname = 'pgcrypto') THEN
CREATE EXTENSION IF NOT EXISTS pgcrypto;
END IF;
END $$;
"""
)
)
if not PGVECTOR_PGCRYPTO_KEY:
raise ValueError(
@ -110,13 +166,9 @@ class PgvectorClient(VectorDBBase):
connection = self.session.connection()
Base.metadata.create_all(bind=connection)
# Create an index on the vector column if it doesn't exist
self.session.execute(
text(
"CREATE INDEX IF NOT EXISTS idx_document_chunk_vector "
"ON document_chunk USING ivfflat (vector vector_cosine_ops) WITH (lists = 100);"
)
)
index_method, index_options = self._vector_index_configuration()
self._ensure_vector_index(index_method, index_options)
self.session.execute(
text(
"CREATE INDEX IF NOT EXISTS idx_document_chunk_collection_name "
@ -130,6 +182,78 @@ class PgvectorClient(VectorDBBase):
log.exception(f"Error during initialization: {e}")
raise
@staticmethod
def _extract_index_method(index_def: Optional[str]) -> Optional[str]:
if not index_def:
return None
try:
after_using = index_def.lower().split("using ", 1)[1]
return after_using.split()[0]
except (IndexError, AttributeError):
return None
def _vector_index_configuration(self) -> Tuple[str, str]:
if PGVECTOR_INDEX_METHOD:
index_method = PGVECTOR_INDEX_METHOD
log.info(
"Using vector index method '%s' from PGVECTOR_INDEX_METHOD.",
index_method,
)
elif USE_HALFVEC:
index_method = "hnsw"
log.info(
"VECTOR_LENGTH=%s exceeds 2000; using halfvec column type with hnsw index.",
VECTOR_LENGTH,
)
else:
index_method = "ivfflat"
if index_method == "hnsw":
index_options = f"WITH (m = {PGVECTOR_HNSW_M}, ef_construction = {PGVECTOR_HNSW_EF_CONSTRUCTION})"
else:
index_options = f"WITH (lists = {PGVECTOR_IVFFLAT_LISTS})"
return index_method, index_options
def _ensure_vector_index(self, index_method: str, index_options: str) -> None:
index_name = "idx_document_chunk_vector"
existing_index_def = self.session.execute(
text(
"""
SELECT indexdef
FROM pg_indexes
WHERE schemaname = current_schema()
AND tablename = 'document_chunk'
AND indexname = :index_name
"""
),
{"index_name": index_name},
).scalar()
existing_method = self._extract_index_method(existing_index_def)
if existing_method and existing_method != index_method:
raise RuntimeError(
f"Existing pgvector index '{index_name}' uses method '{existing_method}' but configuration now "
f"requires '{index_method}'. Automatic rebuild is disabled to prevent long-running maintenance. "
"Drop the index manually (optionally after tuning maintenance_work_mem/max_parallel_maintenance_workers) "
"and recreate it with the new method before restarting Open WebUI."
)
if not existing_index_def:
index_sql = (
f"CREATE INDEX IF NOT EXISTS {index_name} "
f"ON document_chunk USING {index_method} (vector {VECTOR_OPCLASS})"
)
if index_options:
index_sql = f"{index_sql} {index_options}"
self.session.execute(text(index_sql))
log.info(
"Ensured vector index '%s' using %s%s.",
index_name,
index_method,
f" {index_options}" if index_options else "",
)
def check_vector_length(self) -> None:
"""
Check if the VECTOR_LENGTH matches the existing vector column dimension in the database.
@ -149,16 +273,19 @@ class PgvectorClient(VectorDBBase):
if "vector" in document_chunk_table.columns:
vector_column = document_chunk_table.columns["vector"]
vector_type = vector_column.type
if isinstance(vector_type, Vector):
db_vector_length = vector_type.dim
if db_vector_length != VECTOR_LENGTH:
raise Exception(
f"VECTOR_LENGTH {VECTOR_LENGTH} does not match existing vector column dimension {db_vector_length}. "
"Cannot change vector size after initialization without migrating the data."
)
else:
expected_type = HALFVEC if USE_HALFVEC else Vector
if not isinstance(vector_type, expected_type):
raise Exception(
"The 'vector' column exists but is not of type 'Vector'."
"The 'vector' column type does not match the expected type "
f"('{expected_type.__name__}') for VECTOR_LENGTH {VECTOR_LENGTH}."
)
db_vector_length = getattr(vector_type, "dim", None)
if db_vector_length is not None and db_vector_length != VECTOR_LENGTH:
raise Exception(
f"VECTOR_LENGTH {VECTOR_LENGTH} does not match existing vector column dimension {db_vector_length}. "
"Cannot change vector size after initialization without migrating the data."
)
else:
raise Exception(
@ -182,6 +309,8 @@ class PgvectorClient(VectorDBBase):
for item in items:
vector = self.adjust_vector_length(item["vector"])
# Use raw SQL for BYTEA/pgcrypto
# Ensure metadata is converted to its JSON text representation
json_metadata = json.dumps(item["metadata"])
self.session.execute(
text(
"""
@ -190,7 +319,7 @@ class PgvectorClient(VectorDBBase):
VALUES (
:id, :vector, :collection_name,
pgp_sym_encrypt(:text, :key),
pgp_sym_encrypt(:metadata::text, :key)
pgp_sym_encrypt(:metadata_text, :key)
)
ON CONFLICT (id) DO NOTHING
"""
@ -200,7 +329,7 @@ class PgvectorClient(VectorDBBase):
"vector": vector,
"collection_name": collection_name,
"text": item["text"],
"metadata": json.dumps(item["metadata"]),
"metadata_text": json_metadata,
"key": PGVECTOR_PGCRYPTO_KEY,
},
)
@ -216,7 +345,7 @@ class PgvectorClient(VectorDBBase):
vector=vector,
collection_name=collection_name,
text=item["text"],
vmetadata=item["metadata"],
vmetadata=process_metadata(item["metadata"]),
)
new_items.append(new_chunk)
self.session.bulk_save_objects(new_items)
@ -234,6 +363,7 @@ class PgvectorClient(VectorDBBase):
if PGVECTOR_PGCRYPTO:
for item in items:
vector = self.adjust_vector_length(item["vector"])
json_metadata = json.dumps(item["metadata"])
self.session.execute(
text(
"""
@ -242,7 +372,7 @@ class PgvectorClient(VectorDBBase):
VALUES (
:id, :vector, :collection_name,
pgp_sym_encrypt(:text, :key),
pgp_sym_encrypt(:metadata::text, :key)
pgp_sym_encrypt(:metadata_text, :key)
)
ON CONFLICT (id) DO UPDATE SET
vector = EXCLUDED.vector,
@ -256,7 +386,7 @@ class PgvectorClient(VectorDBBase):
"vector": vector,
"collection_name": collection_name,
"text": item["text"],
"metadata": json.dumps(item["metadata"]),
"metadata_text": json_metadata,
"key": PGVECTOR_PGCRYPTO_KEY,
},
)
@ -273,7 +403,7 @@ class PgvectorClient(VectorDBBase):
if existing:
existing.vector = vector
existing.text = item["text"]
existing.vmetadata = item["metadata"]
existing.vmetadata = process_metadata(item["metadata"])
existing.collection_name = (
collection_name # Update collection_name if necessary
)
@ -283,7 +413,7 @@ class PgvectorClient(VectorDBBase):
vector=vector,
collection_name=collection_name,
text=item["text"],
vmetadata=item["metadata"],
vmetadata=process_metadata(item["metadata"]),
)
self.session.add(new_chunk)
self.session.commit()
@ -310,11 +440,11 @@ class PgvectorClient(VectorDBBase):
num_queries = len(vectors)
def vector_expr(vector):
return cast(array(vector), Vector(VECTOR_LENGTH))
return cast(array(vector), VECTOR_TYPE_FACTORY(VECTOR_LENGTH))
# Create the values for query vectors
qid_col = column("qid", Integer)
q_vector_col = column("q_vector", Vector(VECTOR_LENGTH))
q_vector_col = column("q_vector", VECTOR_TYPE_FACTORY(VECTOR_LENGTH))
query_vectors = (
values(qid_col, q_vector_col)
.data(
@ -397,10 +527,12 @@ class PgvectorClient(VectorDBBase):
documents[qid].append(row.text)
metadatas[qid].append(row.vmetadata)
self.session.rollback() # read-only transaction
return SearchResult(
ids=ids, distances=distances, documents=documents, metadatas=metadatas
)
except Exception as e:
self.session.rollback()
log.exception(f"Error during search: {e}")
return None
@ -453,12 +585,14 @@ class PgvectorClient(VectorDBBase):
documents = [[result.text for result in results]]
metadatas = [[result.vmetadata for result in results]]
self.session.rollback() # read-only transaction
return GetResult(
ids=ids,
documents=documents,
metadatas=metadatas,
)
except Exception as e:
self.session.rollback()
log.exception(f"Error during query: {e}")
return None
@ -499,8 +633,10 @@ class PgvectorClient(VectorDBBase):
documents = [[result.text for result in results]]
metadatas = [[result.vmetadata for result in results]]
self.session.rollback() # read-only transaction
return GetResult(ids=ids, documents=documents, metadatas=metadatas)
except Exception as e:
self.session.rollback()
log.exception(f"Error during get: {e}")
return None
@ -568,8 +704,10 @@ class PgvectorClient(VectorDBBase):
.first()
is not None
)
self.session.rollback() # read-only transaction
return exists
except Exception as e:
self.session.rollback()
log.exception(f"Error checking collection existence: {e}")
return False

4
backend/open_webui/retrieval/vector/dbs/pinecone.py
View file

@ -32,6 +32,8 @@ from open_webui.config import (
PINECONE_CLOUD,
)
from open_webui.env import SRC_LOG_LEVELS
from open_webui.retrieval.vector.utils import process_metadata
NO_LIMIT = 10000 # Reasonable limit to avoid overwhelming the system
BATCH_SIZE = 100 # Recommended batch size for Pinecone operations
@ -183,7 +185,7 @@ class PineconeClient(VectorDBBase):
point = {
"id": item["id"],
"values": item["vector"],
"metadata": metadata,
"metadata": process_metadata(metadata),
}
points.append(point)
return points

46
backend/open_webui/retrieval/vector/dbs/qdrant.py
View file

@ -18,6 +18,9 @@ from open_webui.config import (
QDRANT_ON_DISK,
QDRANT_GRPC_PORT,
QDRANT_PREFER_GRPC,
QDRANT_COLLECTION_PREFIX,
QDRANT_TIMEOUT,
QDRANT_HNSW_M,
)
from open_webui.env import SRC_LOG_LEVELS
@ -29,12 +32,14 @@ log.setLevel(SRC_LOG_LEVELS["RAG"])
class QdrantClient(VectorDBBase):
def __init__(self):
self.collection_prefix = "open-webui"
self.collection_prefix = QDRANT_COLLECTION_PREFIX
self.QDRANT_URI = QDRANT_URI
self.QDRANT_API_KEY = QDRANT_API_KEY
self.QDRANT_ON_DISK = QDRANT_ON_DISK
self.PREFER_GRPC = QDRANT_PREFER_GRPC
self.GRPC_PORT = QDRANT_GRPC_PORT
self.QDRANT_TIMEOUT = QDRANT_TIMEOUT
self.QDRANT_HNSW_M = QDRANT_HNSW_M
if not self.QDRANT_URI:
self.client = None
@ -52,9 +57,14 @@ class QdrantClient(VectorDBBase):
grpc_port=self.GRPC_PORT,
prefer_grpc=self.PREFER_GRPC,
api_key=self.QDRANT_API_KEY,
timeout=self.QDRANT_TIMEOUT,
)
else:
self.client = Qclient(url=self.QDRANT_URI, api_key=self.QDRANT_API_KEY)
self.client = Qclient(
url=self.QDRANT_URI,
api_key=self.QDRANT_API_KEY,
timeout=QDRANT_TIMEOUT,
)
def _result_to_get_result(self, points) -> GetResult:
ids = []
@ -84,8 +94,30 @@ class QdrantClient(VectorDBBase):
distance=models.Distance.COSINE,
on_disk=self.QDRANT_ON_DISK,
),
hnsw_config=models.HnswConfigDiff(
m=self.QDRANT_HNSW_M,
),
)
# Create payload indexes for efficient filtering
self.client.create_payload_index(
collection_name=collection_name_with_prefix,
field_name="metadata.hash",
field_schema=models.KeywordIndexParams(
type=models.KeywordIndexType.KEYWORD,
is_tenant=False,
on_disk=self.QDRANT_ON_DISK,
),
)
self.client.create_payload_index(
collection_name=collection_name_with_prefix,
field_name="metadata.file_id",
field_schema=models.KeywordIndexParams(
type=models.KeywordIndexType.KEYWORD,
is_tenant=False,
on_disk=self.QDRANT_ON_DISK,
),
)
log.info(f"collection {collection_name_with_prefix} successfully created!")
def _create_collection_if_not_exists(self, collection_name, dimension):
@ -151,23 +183,23 @@ class QdrantClient(VectorDBBase):
)
)
points = self.client.query_points(
points = self.client.scroll(
collection_name=f"{self.collection_prefix}_{collection_name}",
query_filter=models.Filter(should=field_conditions),
scroll_filter=models.Filter(should=field_conditions),
limit=limit,
)
return self._result_to_get_result(points.points)
return self._result_to_get_result(points[0])
except Exception as e:
log.exception(f"Error querying a collection '{collection_name}': {e}")
return None
def get(self, collection_name: str) -> Optional[GetResult]:
# Get all the items in the collection.
points = self.client.query_points(
points = self.client.scroll(
collection_name=f"{self.collection_prefix}_{collection_name}",
limit=NO_LIMIT, # otherwise qdrant would set limit to 10!
)
return self._result_to_get_result(points.points)
return self._result_to_get_result(points[0])
def insert(self, collection_name: str, items: list[VectorItem]):
# Insert the items into the collection, if the collection does not exist, it will be created.

682
backend/open_webui/retrieval/vector/dbs/qdrant_multitenancy.py
View file

@ -1,5 +1,5 @@
import logging
from typing import Optional, Tuple
from typing import Optional, Tuple, List, Dict, Any
from urllib.parse import urlparse
import grpc
@ -9,6 +9,9 @@ from open_webui.config import (
QDRANT_ON_DISK,
QDRANT_PREFER_GRPC,
QDRANT_URI,
QDRANT_COLLECTION_PREFIX,
QDRANT_TIMEOUT,
QDRANT_HNSW_M,
)
from open_webui.env import SRC_LOG_LEVELS
from open_webui.retrieval.vector.main import (
@ -23,39 +26,62 @@ from qdrant_client.http.models import PointStruct
from qdrant_client.models import models
NO_LIMIT = 999999999
TENANT_ID_FIELD = "tenant_id"
DEFAULT_DIMENSION = 384
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["RAG"])
def _tenant_filter(tenant_id: str) -> models.FieldCondition:
return models.FieldCondition(
key=TENANT_ID_FIELD, match=models.MatchValue(value=tenant_id)
)
def _metadata_filter(key: str, value: Any) -> models.FieldCondition:
return models.FieldCondition(
key=f"metadata.{key}", match=models.MatchValue(value=value)
)
class QdrantClient(VectorDBBase):
def __init__(self):
self.collection_prefix = "open-webui"
self.collection_prefix = QDRANT_COLLECTION_PREFIX
self.QDRANT_URI = QDRANT_URI
self.QDRANT_API_KEY = QDRANT_API_KEY
self.QDRANT_ON_DISK = QDRANT_ON_DISK
self.PREFER_GRPC = QDRANT_PREFER_GRPC
self.GRPC_PORT = QDRANT_GRPC_PORT
self.QDRANT_TIMEOUT = QDRANT_TIMEOUT
self.QDRANT_HNSW_M = QDRANT_HNSW_M
if not self.QDRANT_URI:
self.client = None
return
raise ValueError(
"QDRANT_URI is not set. Please configure it in the environment variables."
)
# Unified handling for either scheme
parsed = urlparse(self.QDRANT_URI)
host = parsed.hostname or self.QDRANT_URI
http_port = parsed.port or 6333 # default REST port
if self.PREFER_GRPC:
self.client = Qclient(
self.client = (
Qclient(
host=host,
port=http_port,
grpc_port=self.GRPC_PORT,
prefer_grpc=self.PREFER_GRPC,
api_key=self.QDRANT_API_KEY,
timeout=self.QDRANT_TIMEOUT,
)
else:
self.client = Qclient(url=self.QDRANT_URI, api_key=self.QDRANT_API_KEY)
if self.PREFER_GRPC
else Qclient(
url=self.QDRANT_URI,
api_key=self.QDRANT_API_KEY,
timeout=self.QDRANT_TIMEOUT,
)
)
# Main collection types for multi-tenancy
self.MEMORY_COLLECTION = f"{self.collection_prefix}_memories"
@ -65,23 +91,13 @@ class QdrantClient(VectorDBBase):
self.HASH_BASED_COLLECTION = f"{self.collection_prefix}_hash-based"
def _result_to_get_result(self, points) -> GetResult:
ids = []
documents = []
metadatas = []
ids, documents, metadatas = [], [], []
for point in points:
payload = point.payload
ids.append(point.id)
documents.append(payload["text"])
metadatas.append(payload["metadata"])
return GetResult(
**{
"ids": [ids],
"documents": [documents],
"metadatas": [metadatas],
}
)
return GetResult(ids=[ids], documents=[documents], metadatas=[metadatas])
def _get_collection_and_tenant_id(self, collection_name: str) -> Tuple[str, str]:
"""
@ -89,6 +105,13 @@ class QdrantClient(VectorDBBase):
Returns:
tuple: (collection_name, tenant_id)
WARNING: This mapping relies on current Open WebUI naming conventions for
collection names. If Open WebUI changes how it generates collection names
(e.g., "user-memory-" prefix, "file-" prefix, web search patterns, or hash
formats), this mapping will break and route data to incorrect collections.
POTENTIALLY CAUSING HUGE DATA CORRUPTION, DATA CONSISTENCY ISSUES AND INCORRECT
DATA MAPPING INSIDE THE DATABASE.
"""
# Check for user memory collections
tenant_id = collection_name
@ -113,143 +136,53 @@ class QdrantClient(VectorDBBase):
else:
return self.KNOWLEDGE_COLLECTION, tenant_id
def _extract_error_message(self, exception):
"""
Extract error message from either HTTP or gRPC exceptions
Returns:
tuple: (status_code, error_message)
"""
# Check if it's an HTTP exception
if isinstance(exception, UnexpectedResponse):
try:
error_data = exception.structured()
error_msg = error_data.get("status", {}).get("error", "")
return exception.status_code, error_msg
except Exception as inner_e:
log.error(f"Failed to parse HTTP error: {inner_e}")
return exception.status_code, str(exception)
# Check if it's a gRPC exception
elif isinstance(exception, grpc.RpcError):
# Extract status code from gRPC error
status_code = None
if hasattr(exception, "code") and callable(exception.code):
status_code = exception.code().value[0]
# Extract error message
error_msg = str(exception)
if "details =" in error_msg:
# Parse the details line which contains the actual error message
try:
details_line = [
line.strip()
for line in error_msg.split("\n")
if "details =" in line
][0]
error_msg = details_line.split("details =")[1].strip(' "')
except (IndexError, AttributeError):
# Fall back to full message if parsing fails
pass
return status_code, error_msg
# For any other type of exception
return None, str(exception)
def _is_collection_not_found_error(self, exception):
"""
Check if the exception is due to collection not found, supporting both HTTP and gRPC
"""
status_code, error_msg = self._extract_error_message(exception)
# HTTP error (404)
if (
status_code == 404
and "Collection" in error_msg
and "doesn't exist" in error_msg
):
return True
# gRPC error (NOT_FOUND status)
if (
isinstance(exception, grpc.RpcError)
and exception.code() == grpc.StatusCode.NOT_FOUND
):
return True
return False
def _is_dimension_mismatch_error(self, exception):
"""
Check if the exception is due to dimension mismatch, supporting both HTTP and gRPC
"""
status_code, error_msg = self._extract_error_message(exception)
# Common patterns in both HTTP and gRPC
return (
"Vector dimension error" in error_msg
or "dimensions mismatch" in error_msg
or "invalid vector size" in error_msg
)
def _create_multi_tenant_collection_if_not_exists(
self, mt_collection_name: str, dimension: int = 384
def _create_multi_tenant_collection(
self, mt_collection_name: str, dimension: int = DEFAULT_DIMENSION
):
"""
Creates a collection with multi-tenancy configuration if it doesn't exist.
Default dimension is set to 384 which corresponds to 'sentence-transformers/all-MiniLM-L6-v2'.
When creating collections dynamically (insert/upsert), the actual vector dimensions will be used.
Creates a collection with multi-tenancy configuration and payload indexes for tenant_id and metadata fields.
"""
try:
# Try to create the collection directly - will fail if it already exists
self.client.create_collection(
collection_name=mt_collection_name,
vectors_config=models.VectorParams(
size=dimension,
distance=models.Distance.COSINE,
on_disk=self.QDRANT_ON_DISK,
),
hnsw_config=models.HnswConfigDiff(
payload_m=16, # Enable per-tenant indexing
m=0,
on_disk=self.QDRANT_ON_DISK,
),
)
self.client.create_collection(
collection_name=mt_collection_name,
vectors_config=models.VectorParams(
size=dimension,
distance=models.Distance.COSINE,
on_disk=self.QDRANT_ON_DISK,
),
# Disable global index building due to multitenancy
# For more details https://qdrant.tech/documentation/guides/multiple-partitions/#calibrate-performance
hnsw_config=models.HnswConfigDiff(
payload_m=self.QDRANT_HNSW_M,
m=0,
),
)
log.info(
f"Multi-tenant collection {mt_collection_name} created with dimension {dimension}!"
)
# Create tenant ID payload index
self.client.create_payload_index(
collection_name=mt_collection_name,
field_name=TENANT_ID_FIELD,
field_schema=models.KeywordIndexParams(
type=models.KeywordIndexType.KEYWORD,
is_tenant=True,
on_disk=self.QDRANT_ON_DISK,
),
)
for field in ("metadata.hash", "metadata.file_id"):
self.client.create_payload_index(
collection_name=mt_collection_name,
field_name="tenant_id",
field_name=field,
field_schema=models.KeywordIndexParams(
type=models.KeywordIndexType.KEYWORD,
is_tenant=True,
on_disk=self.QDRANT_ON_DISK,
),
wait=True,
)
log.info(
f"Multi-tenant collection {mt_collection_name} created with dimension {dimension}!"
)
except (UnexpectedResponse, grpc.RpcError) as e:
# Check for the specific error indicating collection already exists
status_code, error_msg = self._extract_error_message(e)
# HTTP status code 409 or gRPC ALREADY_EXISTS
if (isinstance(e, UnexpectedResponse) and status_code == 409) or (
isinstance(e, grpc.RpcError)
and e.code() == grpc.StatusCode.ALREADY_EXISTS
):
if "already exists" in error_msg:
log.debug(f"Collection {mt_collection_name} already exists")
return
# If it's not an already exists error, re-raise
raise e
except Exception as e:
raise e
def _create_points(self, items: list[VectorItem], tenant_id: str):
def _create_points(
self, items: List[VectorItem], tenant_id: str
) -> List[PointStruct]:
"""
Create point structs from vector items with tenant ID.
"""
@ -260,56 +193,42 @@ class QdrantClient(VectorDBBase):
payload={
"text": item["text"],
"metadata": item["metadata"],
"tenant_id": tenant_id,
TENANT_ID_FIELD: tenant_id,
},
)
for item in items
]
def _ensure_collection(
self, mt_collection_name: str, dimension: int = DEFAULT_DIMENSION
):
"""
Ensure the collection exists and payload indexes are created for tenant_id and metadata fields.
"""
if not self.client.collection_exists(collection_name=mt_collection_name):
self._create_multi_tenant_collection(mt_collection_name, dimension)
def has_collection(self, collection_name: str) -> bool:
"""
Check if a logical collection exists by checking for any points with the tenant ID.
"""
if not self.client:
return False
# Map to multi-tenant collection and tenant ID
mt_collection, tenant_id = self._get_collection_and_tenant_id(collection_name)
# Create tenant filter
tenant_filter = models.FieldCondition(
key="tenant_id", match=models.MatchValue(value=tenant_id)
)
try:
# Try directly querying - most of the time collection should exist
response = self.client.query_points(
collection_name=mt_collection,
query_filter=models.Filter(must=[tenant_filter]),
limit=1,
)
# Collection exists with this tenant ID if there are points
return len(response.points) > 0
except (UnexpectedResponse, grpc.RpcError) as e:
if self._is_collection_not_found_error(e):
log.debug(f"Collection {mt_collection} doesn't exist")
return False
else:
# For other API errors, log and return False
_, error_msg = self._extract_error_message(e)
log.warning(f"Unexpected Qdrant error: {error_msg}")
return False
except Exception as e:
# For any other errors, log and return False
log.debug(f"Error checking collection {mt_collection}: {e}")
if not self.client.collection_exists(collection_name=mt_collection):
return False
tenant_filter = _tenant_filter(tenant_id)
count_result = self.client.count(
collection_name=mt_collection,
count_filter=models.Filter(must=[tenant_filter]),
)
return count_result.count > 0
def delete(
self,
collection_name: str,
ids: Optional[list[str]] = None,
filter: Optional[dict] = None,
ids: Optional[List[str]] = None,
filter: Optional[Dict[str, Any]] = None,
):
"""
Delete vectors by ID or filter from a collection with tenant isolation.
@ -317,189 +236,76 @@ class QdrantClient(VectorDBBase):
if not self.client:
return None
# Map to multi-tenant collection and tenant ID
mt_collection, tenant_id = self._get_collection_and_tenant_id(collection_name)
if not self.client.collection_exists(collection_name=mt_collection):
log.debug(f"Collection {mt_collection} doesn't exist, nothing to delete")
return None
# Create tenant filter
tenant_filter = models.FieldCondition(
key="tenant_id", match=models.MatchValue(value=tenant_id)
must_conditions = [_tenant_filter(tenant_id)]
should_conditions = []
if ids:
should_conditions = [_metadata_filter("id", id_value) for id_value in ids]
elif filter:
must_conditions += [_metadata_filter(k, v) for k, v in filter.items()]
return self.client.delete(
collection_name=mt_collection,
points_selector=models.FilterSelector(
filter=models.Filter(must=must_conditions, should=should_conditions)
),
)
must_conditions = [tenant_filter]
should_conditions = []
if ids:
for id_value in ids:
should_conditions.append(
models.FieldCondition(
key="metadata.id",
match=models.MatchValue(value=id_value),
),
)
elif filter:
for key, value in filter.items():
must_conditions.append(
models.FieldCondition(
key=f"metadata.{key}",
match=models.MatchValue(value=value),
),
)
try:
# Try to delete directly - most of the time collection should exist
update_result = self.client.delete(
collection_name=mt_collection,
points_selector=models.FilterSelector(
filter=models.Filter(must=must_conditions, should=should_conditions)
),
)
return update_result
except (UnexpectedResponse, grpc.RpcError) as e:
if self._is_collection_not_found_error(e):
log.debug(
f"Collection {mt_collection} doesn't exist, nothing to delete"
)
return None
else:
# For other API errors, log and re-raise
_, error_msg = self._extract_error_message(e)
log.warning(f"Unexpected Qdrant error: {error_msg}")
raise
except Exception as e:
# For non-Qdrant exceptions, re-raise
raise
def search(
self, collection_name: str, vectors: list[list[float | int]], limit: int
self, collection_name: str, vectors: List[List[float | int]], limit: int
) -> Optional[SearchResult]:
"""
Search for the nearest neighbor items based on the vectors with tenant isolation.
"""
if not self.client:
if not self.client or not vectors:
return None
# Map to multi-tenant collection and tenant ID
mt_collection, tenant_id = self._get_collection_and_tenant_id(collection_name)
# Get the vector dimension from the query vector
dimension = len(vectors[0]) if vectors and len(vectors) > 0 else None
try:
# Try the search operation directly - most of the time collection should exist
# Create tenant filter
tenant_filter = models.FieldCondition(
key="tenant_id", match=models.MatchValue(value=tenant_id)
)
# Ensure vector dimensions match the collection
collection_dim = self.client.get_collection(
mt_collection
).config.params.vectors.size
if collection_dim != dimension:
if collection_dim < dimension:
vectors = [vector[:collection_dim] for vector in vectors]
else:
vectors = [
vector + [0] * (collection_dim - dimension)
for vector in vectors
]
# Search with tenant filter
prefetch_query = models.Prefetch(
filter=models.Filter(must=[tenant_filter]),
limit=NO_LIMIT,
)
query_response = self.client.query_points(
collection_name=mt_collection,
query=vectors[0],
prefetch=prefetch_query,
limit=limit,
)
get_result = self._result_to_get_result(query_response.points)
return SearchResult(
ids=get_result.ids,
documents=get_result.documents,
metadatas=get_result.metadatas,
# qdrant distance is [-1, 1], normalize to [0, 1]
distances=[
[(point.score + 1.0) / 2.0 for point in query_response.points]
],
)
except (UnexpectedResponse, grpc.RpcError) as e:
if self._is_collection_not_found_error(e):
log.debug(
f"Collection {mt_collection} doesn't exist, search returns None"
)
return None
else:
# For other API errors, log and re-raise
_, error_msg = self._extract_error_message(e)
log.warning(f"Unexpected Qdrant error during search: {error_msg}")
raise
except Exception as e:
# For non-Qdrant exceptions, log and return None
log.exception(f"Error searching collection '{collection_name}': {e}")
if not self.client.collection_exists(collection_name=mt_collection):
log.debug(f"Collection {mt_collection} doesn't exist, search returns None")
return None
def query(self, collection_name: str, filter: dict, limit: Optional[int] = None):
tenant_filter = _tenant_filter(tenant_id)
query_response = self.client.query_points(
collection_name=mt_collection,
query=vectors[0],
limit=limit,
query_filter=models.Filter(must=[tenant_filter]),
)
get_result = self._result_to_get_result(query_response.points)
return SearchResult(
ids=get_result.ids,
documents=get_result.documents,
metadatas=get_result.metadatas,
distances=[[(point.score + 1.0) / 2.0 for point in query_response.points]],
)
def query(
self, collection_name: str, filter: Dict[str, Any], limit: Optional[int] = None
):
"""
Query points with filters and tenant isolation.
"""
if not self.client:
return None
# Map to multi-tenant collection and tenant ID
mt_collection, tenant_id = self._get_collection_and_tenant_id(collection_name)
# Set default limit if not provided
if not self.client.collection_exists(collection_name=mt_collection):
log.debug(f"Collection {mt_collection} doesn't exist, query returns None")
return None
if limit is None:
limit = NO_LIMIT
# Create tenant filter
tenant_filter = models.FieldCondition(
key="tenant_id", match=models.MatchValue(value=tenant_id)
)
# Create metadata filters
field_conditions = []
for key, value in filter.items():
field_conditions.append(
models.FieldCondition(
key=f"metadata.{key}", match=models.MatchValue(value=value)
)
)
# Combine tenant filter with metadata filters
tenant_filter = _tenant_filter(tenant_id)
field_conditions = [_metadata_filter(k, v) for k, v in filter.items()]
combined_filter = models.Filter(must=[tenant_filter, *field_conditions])
try:
# Try the query directly - most of the time collection should exist
points = self.client.query_points(
collection_name=mt_collection,
query_filter=combined_filter,
limit=limit,
)
return self._result_to_get_result(points.points)
except (UnexpectedResponse, grpc.RpcError) as e:
if self._is_collection_not_found_error(e):
log.debug(
f"Collection {mt_collection} doesn't exist, query returns None"
)
return None
else:
# For other API errors, log and re-raise
_, error_msg = self._extract_error_message(e)
log.warning(f"Unexpected Qdrant error during query: {error_msg}")
raise
except Exception as e:
# For non-Qdrant exceptions, log and re-raise
log.exception(f"Error querying collection '{collection_name}': {e}")
return None
points = self.client.scroll(
collection_name=mt_collection,
scroll_filter=combined_filter,
limit=limit,
)
return self._result_to_get_result(points[0])
def get(self, collection_name: str) -> Optional[GetResult]:
"""
@ -507,169 +313,36 @@ class QdrantClient(VectorDBBase):
"""
if not self.client:
return None
# Map to multi-tenant collection and tenant ID
mt_collection, tenant_id = self._get_collection_and_tenant_id(collection_name)
# Create tenant filter
tenant_filter = models.FieldCondition(
key="tenant_id", match=models.MatchValue(value=tenant_id)
)
try:
# Try to get points directly - most of the time collection should exist
points = self.client.query_points(
collection_name=mt_collection,
query_filter=models.Filter(must=[tenant_filter]),
limit=NO_LIMIT,
)
return self._result_to_get_result(points.points)
except (UnexpectedResponse, grpc.RpcError) as e:
if self._is_collection_not_found_error(e):
log.debug(f"Collection {mt_collection} doesn't exist, get returns None")
return None
else:
# For other API errors, log and re-raise
_, error_msg = self._extract_error_message(e)
log.warning(f"Unexpected Qdrant error during get: {error_msg}")
raise
except Exception as e:
# For non-Qdrant exceptions, log and return None
log.exception(f"Error getting collection '{collection_name}': {e}")
if not self.client.collection_exists(collection_name=mt_collection):
log.debug(f"Collection {mt_collection} doesn't exist, get returns None")
return None
def _handle_operation_with_error_retry(
self, operation_name, mt_collection, points, dimension
):
"""
Private helper to handle common error cases for insert and upsert operations.
Args:
operation_name: 'insert' or 'upsert'
mt_collection: The multi-tenant collection name
points: The vector points to insert/upsert
dimension: The dimension of the vectors
Returns:
The operation result (for upsert) or None (for insert)
"""
try:
if operation_name == "insert":
self.client.upload_points(mt_collection, points)
return None
else: # upsert
return self.client.upsert(mt_collection, points)
except (UnexpectedResponse, grpc.RpcError) as e:
# Handle collection not found
if self._is_collection_not_found_error(e):
log.info(
f"Collection {mt_collection} doesn't exist. Creating it with dimension {dimension}."
)
# Create collection with correct dimensions from our vectors
self._create_multi_tenant_collection_if_not_exists(
mt_collection_name=mt_collection, dimension=dimension
)
# Try operation again - no need for dimension adjustment since we just created with correct dimensions
if operation_name == "insert":
self.client.upload_points(mt_collection, points)
return None
else: # upsert
return self.client.upsert(mt_collection, points)
# Handle dimension mismatch
elif self._is_dimension_mismatch_error(e):
# For dimension errors, the collection must exist, so get its configuration
mt_collection_info = self.client.get_collection(mt_collection)
existing_size = mt_collection_info.config.params.vectors.size
log.info(
f"Dimension mismatch: Collection {mt_collection} expects {existing_size}, got {dimension}"
)
if existing_size < dimension:
# Truncate vectors to fit
log.info(
f"Truncating vectors from {dimension} to {existing_size} dimensions"
)
points = [
PointStruct(
id=point.id,
vector=point.vector[:existing_size],
payload=point.payload,
)
for point in points
]
elif existing_size > dimension:
# Pad vectors with zeros
log.info(
f"Padding vectors from {dimension} to {existing_size} dimensions with zeros"
)
points = [
PointStruct(
id=point.id,
vector=point.vector
+ [0] * (existing_size - len(point.vector)),
payload=point.payload,
)
for point in points
]
# Try operation again with adjusted dimensions
if operation_name == "insert":
self.client.upload_points(mt_collection, points)
return None
else: # upsert
return self.client.upsert(mt_collection, points)
else:
# Not a known error we can handle, log and re-raise
_, error_msg = self._extract_error_message(e)
log.warning(f"Unhandled Qdrant error: {error_msg}")
raise
except Exception as e:
# For non-Qdrant exceptions, re-raise
raise
def insert(self, collection_name: str, items: list[VectorItem]):
"""
Insert items with tenant ID.
"""
if not self.client or not items:
return None
# Map to multi-tenant collection and tenant ID
mt_collection, tenant_id = self._get_collection_and_tenant_id(collection_name)
# Get dimensions from the actual vectors
dimension = len(items[0]["vector"]) if items else None
# Create points with tenant ID
points = self._create_points(items, tenant_id)
# Handle the operation with error retry
return self._handle_operation_with_error_retry(
"insert", mt_collection, points, dimension
tenant_filter = _tenant_filter(tenant_id)
points = self.client.scroll(
collection_name=mt_collection,
scroll_filter=models.Filter(must=[tenant_filter]),
limit=NO_LIMIT,
)
return self._result_to_get_result(points[0])
def upsert(self, collection_name: str, items: list[VectorItem]):
def upsert(self, collection_name: str, items: List[VectorItem]):
"""
Upsert items with tenant ID.
"""
if not self.client or not items:
return None
# Map to multi-tenant collection and tenant ID
mt_collection, tenant_id = self._get_collection_and_tenant_id(collection_name)
# Get dimensions from the actual vectors
dimension = len(items[0]["vector"]) if items else None
# Create points with tenant ID
dimension = len(items[0]["vector"])
self._ensure_collection(mt_collection, dimension)
points = self._create_points(items, tenant_id)
self.client.upload_points(mt_collection, points)
return None
# Handle the operation with error retry
return self._handle_operation_with_error_retry(
"upsert", mt_collection, points, dimension
)
def insert(self, collection_name: str, items: List[VectorItem]):
"""
Insert items with tenant ID.
"""
return self.upsert(collection_name, items)
def reset(self):
"""
@ -677,11 +350,9 @@ class QdrantClient(VectorDBBase):
"""
if not self.client:
return None
collection_names = self.client.get_collections().collections
for collection_name in collection_names:
if collection_name.name.startswith(self.collection_prefix):
self.client.delete_collection(collection_name=collection_name.name)
for collection in self.client.get_collections().collections:
if collection.name.startswith(self.collection_prefix):
self.client.delete_collection(collection_name=collection.name)
def delete_collection(self, collection_name: str):
"""
@ -689,24 +360,13 @@ class QdrantClient(VectorDBBase):
"""
if not self.client:
return None
# Map to multi-tenant collection and tenant ID
mt_collection, tenant_id = self._get_collection_and_tenant_id(collection_name)
tenant_filter = models.FieldCondition(
key="tenant_id", match=models.MatchValue(value=tenant_id)
)
field_conditions = [tenant_filter]
update_result = self.client.delete(
if not self.client.collection_exists(collection_name=mt_collection):
log.debug(f"Collection {mt_collection} doesn't exist, nothing to delete")
return None
self.client.delete(
collection_name=mt_collection,
points_selector=models.FilterSelector(
filter=models.Filter(must=field_conditions)
filter=models.Filter(must=[_tenant_filter(tenant_id)])
),
)
if self.client.get_collection(mt_collection).points_count == 0:
self.client.delete_collection(mt_collection)
return update_result

776
backend/open_webui/retrieval/vector/dbs/s3vector.py Normal file
View file

@ -0,0 +1,776 @@
from open_webui.retrieval.vector.utils import process_metadata
from open_webui.retrieval.vector.main import (
VectorDBBase,
VectorItem,
GetResult,
SearchResult,
)
from open_webui.config import S3_VECTOR_BUCKET_NAME, S3_VECTOR_REGION
from open_webui.env import SRC_LOG_LEVELS
from typing import List, Optional, Dict, Any, Union
import logging
import boto3
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["RAG"])
class S3VectorClient(VectorDBBase):
"""
AWS S3 Vector integration for Open WebUI Knowledge.
"""
def __init__(self):
self.bucket_name = S3_VECTOR_BUCKET_NAME
self.region = S3_VECTOR_REGION
# Simple validation - log warnings instead of raising exceptions
if not self.bucket_name:
log.warning("S3_VECTOR_BUCKET_NAME not set - S3Vector will not work")
if not self.region:
log.warning("S3_VECTOR_REGION not set - S3Vector will not work")
if self.bucket_name and self.region:
try:
self.client = boto3.client("s3vectors", region_name=self.region)
log.info(
f"S3Vector client initialized for bucket '{self.bucket_name}' in region '{self.region}'"
)
except Exception as e:
log.error(f"Failed to initialize S3Vector client: {e}")
self.client = None
else:
self.client = None
def _create_index(
self,
index_name: str,
dimension: int,
data_type: str = "float32",
distance_metric: str = "cosine",
) -> None:
"""
Create a new index in the S3 vector bucket for the given collection if it does not exist.
"""
if self.has_collection(index_name):
log.debug(f"Index '{index_name}' already exists, skipping creation")
return
try:
self.client.create_index(
vectorBucketName=self.bucket_name,
indexName=index_name,
dataType=data_type,
dimension=dimension,
distanceMetric=distance_metric,
)
log.info(
f"Created S3 index: {index_name} (dim={dimension}, type={data_type}, metric={distance_metric})"
)
except Exception as e:
log.error(f"Error creating S3 index '{index_name}': {e}")
raise
def _filter_metadata(
self, metadata: Dict[str, Any], item_id: str
) -> Dict[str, Any]:
"""
Filter vector metadata keys to comply with S3 Vector API limit of 10 keys maximum.
"""
if not isinstance(metadata, dict) or len(metadata) <= 10:
return metadata
# Keep only the first 10 keys, prioritizing important ones based on actual Open WebUI metadata
important_keys = [
"text", # The actual document content
"file_id", # File ID
"source", # Document source file
"title", # Document title
"page", # Page number
"total_pages", # Total pages in document
"embedding_config", # Embedding configuration
"created_by", # User who created it
"name", # Document name
"hash", # Content hash
]
filtered_metadata = {}
# First, add important keys if they exist
for key in important_keys:
if key in metadata:
filtered_metadata[key] = metadata[key]
if len(filtered_metadata) >= 10:
break
# If we still have room, add other keys
if len(filtered_metadata) < 10:
for key, value in metadata.items():
if key not in filtered_metadata:
filtered_metadata[key] = value
if len(filtered_metadata) >= 10:
break
log.warning(
f"Metadata for key '{item_id}' had {len(metadata)} keys, limited to 10 keys"
)
return filtered_metadata
def has_collection(self, collection_name: str) -> bool:
"""
Check if a vector index exists using direct lookup.
This avoids pagination issues with list_indexes() and is significantly faster.
"""
try:
self.client.get_index(
vectorBucketName=self.bucket_name, indexName=collection_name
)
return True
except Exception as e:
log.error(f"Error checking if index '{collection_name}' exists: {e}")
return False
def delete_collection(self, collection_name: str) -> None:
"""
Delete an entire S3 Vector index/collection.
"""
if not self.has_collection(collection_name):
log.warning(
f"Collection '{collection_name}' does not exist, nothing to delete"
)
return
try:
log.info(f"Deleting collection '{collection_name}'")
self.client.delete_index(
vectorBucketName=self.bucket_name, indexName=collection_name
)
log.info(f"Successfully deleted collection '{collection_name}'")
except Exception as e:
log.error(f"Error deleting collection '{collection_name}': {e}")
raise
def insert(self, collection_name: str, items: List[VectorItem]) -> None:
"""
Insert vector items into the S3 Vector index. Create index if it does not exist.
"""
if not items:
log.warning("No items to insert")
return
dimension = len(items[0]["vector"])
try:
if not self.has_collection(collection_name):
log.info(f"Index '{collection_name}' does not exist. Creating index.")
self._create_index(
index_name=collection_name,
dimension=dimension,
data_type="float32",
distance_metric="cosine",
)
# Prepare vectors for insertion
vectors = []
for item in items:
# Ensure vector data is in the correct format for S3 Vector API
vector_data = item["vector"]
if isinstance(vector_data, list):
# Convert list to float32 values as required by S3 Vector API
vector_data = [float(x) for x in vector_data]
# Prepare metadata, ensuring the text field is preserved
metadata = item.get("metadata", {}).copy()
# Add the text field to metadata so it's available for retrieval
metadata["text"] = item["text"]
# Convert metadata to string format for consistency
metadata = process_metadata(metadata)
# Filter metadata to comply with S3 Vector API limit of 10 keys
metadata = self._filter_metadata(metadata, item["id"])
vectors.append(
{
"key": item["id"],
"data": {"float32": vector_data},
"metadata": metadata,
}
)
# Insert vectors in batches of 500 (S3 Vector API limit)
batch_size = 500
for i in range(0, len(vectors), batch_size):
batch = vectors[i : i + batch_size]
self.client.put_vectors(
vectorBucketName=self.bucket_name,
indexName=collection_name,
vectors=batch,
)
log.info(
f"Inserted batch {i//batch_size + 1}: {len(batch)} vectors into index '{collection_name}'."
)
log.info(
f"Completed insertion of {len(vectors)} vectors into index '{collection_name}'."
)
except Exception as e:
log.error(f"Error inserting vectors: {e}")
raise
def upsert(self, collection_name: str, items: List[VectorItem]) -> None:
"""
Insert or update vector items in the S3 Vector index. Create index if it does not exist.
"""
if not items:
log.warning("No items to upsert")
return
dimension = len(items[0]["vector"])
log.info(f"Upsert dimension: {dimension}")
try:
if not self.has_collection(collection_name):
log.info(
f"Index '{collection_name}' does not exist. Creating index for upsert."
)
self._create_index(
index_name=collection_name,
dimension=dimension,
data_type="float32",
distance_metric="cosine",
)
# Prepare vectors for upsert
vectors = []
for item in items:
# Ensure vector data is in the correct format for S3 Vector API
vector_data = item["vector"]
if isinstance(vector_data, list):
# Convert list to float32 values as required by S3 Vector API
vector_data = [float(x) for x in vector_data]
# Prepare metadata, ensuring the text field is preserved
metadata = item.get("metadata", {}).copy()
# Add the text field to metadata so it's available for retrieval
metadata["text"] = item["text"]
# Convert metadata to string format for consistency
metadata = process_metadata(metadata)
# Filter metadata to comply with S3 Vector API limit of 10 keys
metadata = self._filter_metadata(metadata, item["id"])
vectors.append(
{
"key": item["id"],
"data": {"float32": vector_data},
"metadata": metadata,
}
)
# Upsert vectors in batches of 500 (S3 Vector API limit)
batch_size = 500
for i in range(0, len(vectors), batch_size):
batch = vectors[i : i + batch_size]
if i == 0: # Log sample info for first batch only
log.info(
f"Upserting batch 1: {len(batch)} vectors. First vector sample: key={batch[0]['key']}, data_type={type(batch[0]['data']['float32'])}, data_len={len(batch[0]['data']['float32'])}"
)
else:
log.info(
f"Upserting batch {i//batch_size + 1}: {len(batch)} vectors."
)
self.client.put_vectors(
vectorBucketName=self.bucket_name,
indexName=collection_name,
vectors=batch,
)
log.info(
f"Completed upsert of {len(vectors)} vectors into index '{collection_name}'."
)
except Exception as e:
log.error(f"Error upserting vectors: {e}")
raise
def search(
self, collection_name: str, vectors: List[List[Union[float, int]]], limit: int
) -> Optional[SearchResult]:
"""
Search for similar vectors in a collection using multiple query vectors.
"""
if not self.has_collection(collection_name):
log.warning(f"Collection '{collection_name}' does not exist")
return None
if not vectors:
log.warning("No query vectors provided")
return None
try:
log.info(
f"Searching collection '{collection_name}' with {len(vectors)} query vectors, limit={limit}"
)
# Initialize result lists
all_ids = []
all_documents = []
all_metadatas = []
all_distances = []
# Process each query vector
for i, query_vector in enumerate(vectors):
log.debug(f"Processing query vector {i+1}/{len(vectors)}")
# Prepare the query vector in S3 Vector format
query_vector_dict = {"float32": [float(x) for x in query_vector]}
# Call S3 Vector query API
response = self.client.query_vectors(
vectorBucketName=self.bucket_name,
indexName=collection_name,
topK=limit,
queryVector=query_vector_dict,
returnMetadata=True,
returnDistance=True,
)
# Process results for this query
query_ids = []
query_documents = []
query_metadatas = []
query_distances = []
result_vectors = response.get("vectors", [])
for vector in result_vectors:
vector_id = vector.get("key")
vector_metadata = vector.get("metadata", {})
vector_distance = vector.get("distance", 0.0)
# Extract document text from metadata
document_text = ""
if isinstance(vector_metadata, dict):
# Get the text field first (highest priority)
document_text = vector_metadata.get("text")
if not document_text:
# Fallback to other possible text fields
document_text = (
vector_metadata.get("content")
or vector_metadata.get("document")
or vector_id
)
else:
document_text = vector_id
query_ids.append(vector_id)
query_documents.append(document_text)
query_metadatas.append(vector_metadata)
query_distances.append(vector_distance)
# Add this query's results to the overall results
all_ids.append(query_ids)
all_documents.append(query_documents)
all_metadatas.append(query_metadatas)
all_distances.append(query_distances)
log.info(f"Search completed. Found results for {len(all_ids)} queries")
# Return SearchResult format
return SearchResult(
ids=all_ids if all_ids else None,
documents=all_documents if all_documents else None,
metadatas=all_metadatas if all_metadatas else None,
distances=all_distances if all_distances else None,
)
except Exception as e:
log.error(f"Error searching collection '{collection_name}': {str(e)}")
# Handle specific AWS exceptions
if hasattr(e, "response") and "Error" in e.response:
error_code = e.response["Error"]["Code"]
if error_code == "NotFoundException":
log.warning(f"Collection '{collection_name}' not found")
return None
elif error_code == "ValidationException":
log.error(f"Invalid query vector dimensions or parameters")
return None
elif error_code == "AccessDeniedException":
log.error(
f"Access denied for collection '{collection_name}'. Check permissions."
)
return None
raise
def query(
self, collection_name: str, filter: Dict, limit: Optional[int] = None
) -> Optional[GetResult]:
"""
Query vectors from a collection using metadata filter.
"""
if not self.has_collection(collection_name):
log.warning(f"Collection '{collection_name}' does not exist")
return GetResult(ids=[[]], documents=[[]], metadatas=[[]])
if not filter:
log.warning("No filter provided, returning all vectors")
return self.get(collection_name)
try:
log.info(f"Querying collection '{collection_name}' with filter: {filter}")
# For S3 Vector, we need to use list_vectors and then filter results
# Since S3 Vector may not support complex server-side filtering,
# we'll retrieve all vectors and filter client-side
# Get all vectors first
all_vectors_result = self.get(collection_name)
if not all_vectors_result or not all_vectors_result.ids:
log.warning("No vectors found in collection")
return GetResult(ids=[[]], documents=[[]], metadatas=[[]])
# Extract the lists from the result
all_ids = all_vectors_result.ids[0] if all_vectors_result.ids else []
all_documents = (
all_vectors_result.documents[0] if all_vectors_result.documents else []
)
all_metadatas = (
all_vectors_result.metadatas[0] if all_vectors_result.metadatas else []
)
# Apply client-side filtering
filtered_ids = []
filtered_documents = []
filtered_metadatas = []
for i, metadata in enumerate(all_metadatas):
if self._matches_filter(metadata, filter):
if i < len(all_ids):
filtered_ids.append(all_ids[i])
if i < len(all_documents):
filtered_documents.append(all_documents[i])
filtered_metadatas.append(metadata)
# Apply limit if specified
if limit and len(filtered_ids) >= limit:
break
log.info(
f"Filter applied: {len(filtered_ids)} vectors match out of {len(all_ids)} total"
)
# Return GetResult format
if filtered_ids:
return GetResult(
ids=[filtered_ids],
documents=[filtered_documents],
metadatas=[filtered_metadatas],
)
else:
return GetResult(ids=[[]], documents=[[]], metadatas=[[]])
except Exception as e:
log.error(f"Error querying collection '{collection_name}': {str(e)}")
# Handle specific AWS exceptions
if hasattr(e, "response") and "Error" in e.response:
error_code = e.response["Error"]["Code"]
if error_code == "NotFoundException":
log.warning(f"Collection '{collection_name}' not found")
return GetResult(ids=[[]], documents=[[]], metadatas=[[]])
elif error_code == "AccessDeniedException":
log.error(
f"Access denied for collection '{collection_name}'. Check permissions."
)
return GetResult(ids=[[]], documents=[[]], metadatas=[[]])
raise
def get(self, collection_name: str) -> Optional[GetResult]:
"""
Retrieve all vectors from a collection.
"""
if not self.has_collection(collection_name):
log.warning(f"Collection '{collection_name}' does not exist")
return GetResult(ids=[[]], documents=[[]], metadatas=[[]])
try:
log.info(f"Retrieving all vectors from collection '{collection_name}'")
# Initialize result lists
all_ids = []
all_documents = []
all_metadatas = []
# Handle pagination
next_token = None
while True:
# Prepare request parameters
request_params = {
"vectorBucketName": self.bucket_name,
"indexName": collection_name,
"returnData": False, # Don't include vector data (not needed for get)
"returnMetadata": True, # Include metadata
"maxResults": 500, # Use reasonable page size
}
if next_token:
request_params["nextToken"] = next_token
# Call S3 Vector API
response = self.client.list_vectors(**request_params)
# Process vectors in this page
vectors = response.get("vectors", [])
for vector in vectors:
vector_id = vector.get("key")
vector_data = vector.get("data", {})
vector_metadata = vector.get("metadata", {})
# Extract the actual vector array
vector_array = vector_data.get("float32", [])
# For documents, we try to extract text from metadata or use the vector ID
document_text = ""
if isinstance(vector_metadata, dict):
# Get the text field first (highest priority)
document_text = vector_metadata.get("text")
if not document_text:
# Fallback to other possible text fields
document_text = (
vector_metadata.get("content")
or vector_metadata.get("document")
or vector_id
)
# Log the actual content for debugging
log.debug(
f"Document text preview (first 200 chars): {str(document_text)[:200]}"
)
else:
document_text = vector_id
all_ids.append(vector_id)
all_documents.append(document_text)
all_metadatas.append(vector_metadata)
# Check if there are more pages
next_token = response.get("nextToken")
if not next_token:
break
log.info(
f"Retrieved {len(all_ids)} vectors from collection '{collection_name}'"
)
# Return in GetResult format
# The Open WebUI GetResult expects lists of lists, so we wrap each list
if all_ids:
return GetResult(
ids=[all_ids], documents=[all_documents], metadatas=[all_metadatas]
)
else:
return GetResult(ids=[[]], documents=[[]], metadatas=[[]])
except Exception as e:
log.error(
f"Error retrieving vectors from collection '{collection_name}': {str(e)}"
)
# Handle specific AWS exceptions
if hasattr(e, "response") and "Error" in e.response:
error_code = e.response["Error"]["Code"]
if error_code == "NotFoundException":
log.warning(f"Collection '{collection_name}' not found")
return GetResult(ids=[[]], documents=[[]], metadatas=[[]])
elif error_code == "AccessDeniedException":
log.error(
f"Access denied for collection '{collection_name}'. Check permissions."
)
return GetResult(ids=[[]], documents=[[]], metadatas=[[]])
raise
def delete(
self,
collection_name: str,
ids: Optional[List[str]] = None,
filter: Optional[Dict] = None,
) -> None:
"""
Delete vectors by ID or filter from a collection.
"""
if not self.has_collection(collection_name):
log.warning(
f"Collection '{collection_name}' does not exist, nothing to delete"
)
return
# Check if this is a knowledge collection (not file-specific)
is_knowledge_collection = not collection_name.startswith("file-")
try:
if ids:
# Delete by specific vector IDs/keys
log.info(
f"Deleting {len(ids)} vectors by IDs from collection '{collection_name}'"
)
self.client.delete_vectors(
vectorBucketName=self.bucket_name,
indexName=collection_name,
keys=ids,
)
log.info(f"Deleted {len(ids)} vectors from index '{collection_name}'")
elif filter:
# Handle filter-based deletion
log.info(
f"Deleting vectors by filter from collection '{collection_name}': {filter}"
)
# If this is a knowledge collection and we have a file_id filter,
# also clean up the corresponding file-specific collection
if is_knowledge_collection and "file_id" in filter:
file_id = filter["file_id"]
file_collection_name = f"file-{file_id}"
if self.has_collection(file_collection_name):
log.info(
f"Found related file-specific collection '{file_collection_name}', deleting it to prevent duplicates"
)
self.delete_collection(file_collection_name)
# For the main collection, implement query-then-delete
# First, query to get IDs matching the filter
query_result = self.query(collection_name, filter)
if query_result and query_result.ids and query_result.ids[0]:
matching_ids = query_result.ids[0]
log.info(
f"Found {len(matching_ids)} vectors matching filter, deleting them"
)
# Delete the matching vectors by ID
self.client.delete_vectors(
vectorBucketName=self.bucket_name,
indexName=collection_name,
keys=matching_ids,
)
log.info(
f"Deleted {len(matching_ids)} vectors from index '{collection_name}' using filter"
)
else:
log.warning("No vectors found matching the filter criteria")
else:
log.warning("No IDs or filter provided for deletion")
except Exception as e:
log.error(
f"Error deleting vectors from collection '{collection_name}': {e}"
)
raise
def reset(self) -> None:
"""
Reset/clear all vector data. For S3 Vector, this deletes all indexes.
"""
try:
log.warning(
"Reset called - this will delete all vector indexes in the S3 bucket"
)
# List all indexes
response = self.client.list_indexes(vectorBucketName=self.bucket_name)
indexes = response.get("indexes", [])
if not indexes:
log.warning("No indexes found to delete")
return
# Delete all indexes
deleted_count = 0
for index in indexes:
index_name = index.get("indexName")
if index_name:
try:
self.client.delete_index(
vectorBucketName=self.bucket_name, indexName=index_name
)
deleted_count += 1
log.info(f"Deleted index: {index_name}")
except Exception as e:
log.error(f"Error deleting index '{index_name}': {e}")
log.info(f"Reset completed: deleted {deleted_count} indexes")
except Exception as e:
log.error(f"Error during reset: {e}")
raise
def _matches_filter(self, metadata: Dict[str, Any], filter: Dict[str, Any]) -> bool:
"""
Check if metadata matches the given filter conditions.
"""
if not isinstance(metadata, dict) or not isinstance(filter, dict):
return False
# Check each filter condition
for key, expected_value in filter.items():
# Handle special operators
if key.startswith("$"):
if key == "$and":
# All conditions must match
if not isinstance(expected_value, list):
continue
for condition in expected_value:
if not self._matches_filter(metadata, condition):
return False
elif key == "$or":
# At least one condition must match
if not isinstance(expected_value, list):
continue
any_match = False
for condition in expected_value:
if self._matches_filter(metadata, condition):
any_match = True
break
if not any_match:
return False
continue
# Get the actual value from metadata
actual_value = metadata.get(key)
# Handle different types of expected values
if isinstance(expected_value, dict):
# Handle comparison operators
for op, op_value in expected_value.items():
if op == "$eq":
if actual_value != op_value:
return False
elif op == "$ne":
if actual_value == op_value:
return False
elif op == "$in":
if (
not isinstance(op_value, list)
or actual_value not in op_value
):
return False
elif op == "$nin":
if isinstance(op_value, list) and actual_value in op_value:
return False
elif op == "$exists":
if bool(op_value) != (key in metadata):
return False
# Add more operators as needed
else:
# Simple equality check
if actual_value != expected_value:
return False
return True

340
backend/open_webui/retrieval/vector/dbs/weaviate.py Normal file
View file

@ -0,0 +1,340 @@
import weaviate
import re
import uuid
from typing import Any, Dict, List, Optional, Union
from open_webui.retrieval.vector.main import (
VectorDBBase,
VectorItem,
SearchResult,
GetResult,
)
from open_webui.retrieval.vector.utils import process_metadata
from open_webui.config import (
WEAVIATE_HTTP_HOST,
WEAVIATE_HTTP_PORT,
WEAVIATE_GRPC_PORT,
WEAVIATE_API_KEY,
)
def _convert_uuids_to_strings(obj: Any) -> Any:
"""
Recursively convert UUID objects to strings in nested data structures.
This function handles:
- UUID objects -> string
- Dictionaries with UUID values
- Lists/Tuples with UUID values
- Nested combinations of the above
Args:
obj: Any object that might contain UUIDs
Returns:
The same object structure with UUIDs converted to strings
"""
if isinstance(obj, uuid.UUID):
return str(obj)
elif isinstance(obj, dict):
return {key: _convert_uuids_to_strings(value) for key, value in obj.items()}
elif isinstance(obj, (list, tuple)):
return type(obj)(_convert_uuids_to_strings(item) for item in obj)
elif isinstance(obj, (str, int, float, bool, type(None))):
return obj
else:
return obj
class WeaviateClient(VectorDBBase):
def __init__(self):
self.url = WEAVIATE_HTTP_HOST
try:
# Build connection parameters
connection_params = {
"host": WEAVIATE_HTTP_HOST,
"port": WEAVIATE_HTTP_PORT,
"grpc_port": WEAVIATE_GRPC_PORT,
}
# Only add auth_credentials if WEAVIATE_API_KEY exists and is not empty
if WEAVIATE_API_KEY:
connection_params["auth_credentials"] = (
weaviate.classes.init.Auth.api_key(WEAVIATE_API_KEY)
)
self.client = weaviate.connect_to_local(**connection_params)
self.client.connect()
except Exception as e:
raise ConnectionError(f"Failed to connect to Weaviate: {e}") from e
def _sanitize_collection_name(self, collection_name: str) -> str:
"""Sanitize collection name to be a valid Weaviate class name."""
if not isinstance(collection_name, str) or not collection_name.strip():
raise ValueError("Collection name must be a non-empty string")
# Requirements for a valid Weaviate class name:
# The collection name must begin with a capital letter.
# The name can only contain letters, numbers, and the underscore (_) character. Spaces are not allowed.
# Replace hyphens with underscores and keep only alphanumeric characters
name = re.sub(r"[^a-zA-Z0-9_]", "", collection_name.replace("-", "_"))
name = name.strip("_")
if not name:
raise ValueError(
"Could not sanitize collection name to be a valid Weaviate class name"
)
# Ensure it starts with a letter and is capitalized
if not name[0].isalpha():
name = "C" + name
return name[0].upper() + name[1:]
def has_collection(self, collection_name: str) -> bool:
sane_collection_name = self._sanitize_collection_name(collection_name)
return self.client.collections.exists(sane_collection_name)
def delete_collection(self, collection_name: str) -> None:
sane_collection_name = self._sanitize_collection_name(collection_name)
if self.client.collections.exists(sane_collection_name):
self.client.collections.delete(sane_collection_name)
def _create_collection(self, collection_name: str) -> None:
self.client.collections.create(
name=collection_name,
vector_config=weaviate.classes.config.Configure.Vectors.self_provided(),
properties=[
weaviate.classes.config.Property(
name="text", data_type=weaviate.classes.config.DataType.TEXT
),
],
)
def insert(self, collection_name: str, items: List[VectorItem]) -> None:
sane_collection_name = self._sanitize_collection_name(collection_name)
if not self.client.collections.exists(sane_collection_name):
self._create_collection(sane_collection_name)
collection = self.client.collections.get(sane_collection_name)
with collection.batch.fixed_size(batch_size=100) as batch:
for item in items:
item_uuid = str(uuid.uuid4()) if not item["id"] else str(item["id"])
properties = {"text": item["text"]}
if item["metadata"]:
clean_metadata = _convert_uuids_to_strings(
process_metadata(item["metadata"])
)
clean_metadata.pop("text", None)
properties.update(clean_metadata)
batch.add_object(
properties=properties, uuid=item_uuid, vector=item["vector"]
)
def upsert(self, collection_name: str, items: List[VectorItem]) -> None:
sane_collection_name = self._sanitize_collection_name(collection_name)
if not self.client.collections.exists(sane_collection_name):
self._create_collection(sane_collection_name)
collection = self.client.collections.get(sane_collection_name)
with collection.batch.fixed_size(batch_size=100) as batch:
for item in items:
item_uuid = str(item["id"]) if item["id"] else None
properties = {"text": item["text"]}
if item["metadata"]:
clean_metadata = _convert_uuids_to_strings(
process_metadata(item["metadata"])
)
clean_metadata.pop("text", None)
properties.update(clean_metadata)
batch.add_object(
properties=properties, uuid=item_uuid, vector=item["vector"]
)
def search(
self, collection_name: str, vectors: List[List[Union[float, int]]], limit: int
) -> Optional[SearchResult]:
sane_collection_name = self._sanitize_collection_name(collection_name)
if not self.client.collections.exists(sane_collection_name):
return None
collection = self.client.collections.get(sane_collection_name)
result_ids, result_documents, result_metadatas, result_distances = (
[],
[],
[],
[],
)
for vector_embedding in vectors:
try:
response = collection.query.near_vector(
near_vector=vector_embedding,
limit=limit,
return_metadata=weaviate.classes.query.MetadataQuery(distance=True),
)
ids = [str(obj.uuid) for obj in response.objects]
documents = []
metadatas = []
distances = []
for obj in response.objects:
properties = dict(obj.properties) if obj.properties else {}
documents.append(properties.pop("text", ""))
metadatas.append(_convert_uuids_to_strings(properties))
# Weaviate has cosine distance, 2 (worst) -> 0 (best). Re-ordering to 0 -> 1
raw_distances = [
(
obj.metadata.distance
if obj.metadata and obj.metadata.distance
else 2.0
)
for obj in response.objects
]
distances = [(2 - dist) / 2 for dist in raw_distances]
result_ids.append(ids)
result_documents.append(documents)
result_metadatas.append(metadatas)
result_distances.append(distances)
except Exception:
result_ids.append([])
result_documents.append([])
result_metadatas.append([])
result_distances.append([])
return SearchResult(
**{
"ids": result_ids,
"documents": result_documents,
"metadatas": result_metadatas,
"distances": result_distances,
}
)
def query(
self, collection_name: str, filter: Dict, limit: Optional[int] = None
) -> Optional[GetResult]:
sane_collection_name = self._sanitize_collection_name(collection_name)
if not self.client.collections.exists(sane_collection_name):
return None
collection = self.client.collections.get(sane_collection_name)
weaviate_filter = None
if filter:
for key, value in filter.items():
prop_filter = weaviate.classes.query.Filter.by_property(name=key).equal(
value
)
weaviate_filter = (
prop_filter
if weaviate_filter is None
else weaviate.classes.query.Filter.all_of(
[weaviate_filter, prop_filter]
)
)
try:
response = collection.query.fetch_objects(
filters=weaviate_filter, limit=limit
)
ids = [str(obj.uuid) for obj in response.objects]
documents = []
metadatas = []
for obj in response.objects:
properties = dict(obj.properties) if obj.properties else {}
documents.append(properties.pop("text", ""))
metadatas.append(_convert_uuids_to_strings(properties))
return GetResult(
**{
"ids": [ids],
"documents": [documents],
"metadatas": [metadatas],
}
)
except Exception:
return None
def get(self, collection_name: str) -> Optional[GetResult]:
sane_collection_name = self._sanitize_collection_name(collection_name)
if not self.client.collections.exists(sane_collection_name):
return None
collection = self.client.collections.get(sane_collection_name)
ids, documents, metadatas = [], [], []
try:
for item in collection.iterator():
ids.append(str(item.uuid))
properties = dict(item.properties) if item.properties else {}
documents.append(properties.pop("text", ""))
metadatas.append(_convert_uuids_to_strings(properties))
if not ids:
return None
return GetResult(
**{
"ids": [ids],
"documents": [documents],
"metadatas": [metadatas],
}
)
except Exception:
return None
def delete(
self,
collection_name: str,
ids: Optional[List[str]] = None,
filter: Optional[Dict] = None,
) -> None:
sane_collection_name = self._sanitize_collection_name(collection_name)
if not self.client.collections.exists(sane_collection_name):
return
collection = self.client.collections.get(sane_collection_name)
try:
if ids:
for item_id in ids:
collection.data.delete_by_id(uuid=item_id)
elif filter:
weaviate_filter = None
for key, value in filter.items():
prop_filter = weaviate.classes.query.Filter.by_property(
name=key
).equal(value)
weaviate_filter = (
prop_filter
if weaviate_filter is None
else weaviate.classes.query.Filter.all_of(
[weaviate_filter, prop_filter]
)
)
if weaviate_filter:
collection.data.delete_many(where=weaviate_filter)
except Exception:
pass
def reset(self) -> None:
try:
for collection_name in self.client.collections.list_all().keys():
self.client.collections.delete(collection_name)
except Exception:
pass

29
backend/open_webui/retrieval/vector/factory.py
View file

@ -1,6 +1,10 @@
from open_webui.retrieval.vector.main import VectorDBBase
from open_webui.retrieval.vector.type import VectorType
from open_webui.config import VECTOR_DB, ENABLE_QDRANT_MULTITENANCY_MODE
from open_webui.config import (
VECTOR_DB,
ENABLE_QDRANT_MULTITENANCY_MODE,
ENABLE_MILVUS_MULTITENANCY_MODE,
)
class Vector:
@ -12,9 +16,16 @@ class Vector:
"""
match vector_type:
case VectorType.MILVUS:
from open_webui.retrieval.vector.dbs.milvus import MilvusClient
if ENABLE_MILVUS_MULTITENANCY_MODE:
from open_webui.retrieval.vector.dbs.milvus_multitenancy import (
MilvusClient,
)
return MilvusClient()
return MilvusClient()
else:
from open_webui.retrieval.vector.dbs.milvus import MilvusClient
return MilvusClient()
case VectorType.QDRANT:
if ENABLE_QDRANT_MULTITENANCY_MODE:
from open_webui.retrieval.vector.dbs.qdrant_multitenancy import (
@ -30,6 +41,10 @@ class Vector:
from open_webui.retrieval.vector.dbs.pinecone import PineconeClient
return PineconeClient()
case VectorType.S3VECTOR:
from open_webui.retrieval.vector.dbs.s3vector import S3VectorClient
return S3VectorClient()
case VectorType.OPENSEARCH:
from open_webui.retrieval.vector.dbs.opensearch import OpenSearchClient
@ -48,6 +63,14 @@ class Vector:
from open_webui.retrieval.vector.dbs.chroma import ChromaClient
return ChromaClient()
case VectorType.ORACLE23AI:
from open_webui.retrieval.vector.dbs.oracle23ai import Oracle23aiClient
return Oracle23aiClient()
case VectorType.WEAVIATE:
from open_webui.retrieval.vector.dbs.weaviate import WeaviateClient
return WeaviateClient()
case _:
raise ValueError(f"Unsupported vector type: {vector_type}")

3
backend/open_webui/retrieval/vector/type.py
View file

@ -9,3 +9,6 @@ class VectorType(StrEnum):
ELASTICSEARCH = "elasticsearch"
OPENSEARCH = "opensearch"
PGVECTOR = "pgvector"
ORACLE23AI = "oracle23ai"
S3VECTOR = "s3vector"
WEAVIATE = "weaviate"

28
backend/open_webui/retrieval/vector/utils.py Normal file
View file

@ -0,0 +1,28 @@
from datetime import datetime
KEYS_TO_EXCLUDE = ["content", "pages", "tables", "paragraphs", "sections", "figures"]
def filter_metadata(metadata: dict[str, any]) -> dict[str, any]:
metadata = {
key: value for key, value in metadata.items() if key not in KEYS_TO_EXCLUDE
}
return metadata
def process_metadata(
metadata: dict[str, any],
) -> dict[str, any]:
for key, value in metadata.items():
# Remove large fields
if key in KEYS_TO_EXCLUDE:
del metadata[key]
# Convert non-serializable fields to strings
if (
isinstance(value, datetime)
or isinstance(value, list)
or isinstance(value, dict)
):
metadata[key] = str(value)
return metadata

128
backend/open_webui/retrieval/web/azure.py Normal file
View file

@ -0,0 +1,128 @@
import logging
from typing import Optional
from open_webui.retrieval.web.main import SearchResult, get_filtered_results
from open_webui.env import SRC_LOG_LEVELS
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["RAG"])
"""
Azure AI Search integration for Open WebUI.
Documentation: https://learn.microsoft.com/en-us/python/api/overview/azure/search-documents-readme?view=azure-python
Required package: azure-search-documents
Install: pip install azure-search-documents
"""
def search_azure(
api_key: str,
endpoint: str,
index_name: str,
query: str,
count: int,
filter_list: Optional[list[str]] = None,
) -> list[SearchResult]:
"""
Search using Azure AI Search.
Args:
api_key: Azure Search API key (query key or admin key)
endpoint: Azure Search service endpoint (e.g., https://myservice.search.windows.net)
index_name: Name of the search index to query
query: Search query string
count: Number of results to return
filter_list: Optional list of domains to filter results
Returns:
List of SearchResult objects with link, title, and snippet
"""
try:
from azure.core.credentials import AzureKeyCredential
from azure.search.documents import SearchClient
except ImportError:
log.error(
"azure-search-documents package is not installed. "
"Install it with: pip install azure-search-documents"
)
raise ImportError(
"azure-search-documents is required for Azure AI Search. "
"Install it with: pip install azure-search-documents"
)
try:
# Create search client with API key authentication
credential = AzureKeyCredential(api_key)
search_client = SearchClient(
endpoint=endpoint, index_name=index_name, credential=credential
)
# Perform the search
results = search_client.search(search_text=query, top=count)
# Convert results to list and extract fields
search_results = []
for result in results:
# Azure AI Search returns documents with custom schemas
# We need to extract common fields that might represent URL, title, and content
# Common field names to look for:
result_dict = dict(result)
# Try to find URL field (common names)
link = (
result_dict.get("url")
or result_dict.get("link")
or result_dict.get("uri")
or result_dict.get("metadata_storage_path")
or ""
)
# Try to find title field (common names)
title = (
result_dict.get("title")
or result_dict.get("name")
or result_dict.get("metadata_title")
or result_dict.get("metadata_storage_name")
or None
)
# Try to find content/snippet field (common names)
snippet = (
result_dict.get("content")
or result_dict.get("snippet")
or result_dict.get("description")
or result_dict.get("summary")
or result_dict.get("text")
or None
)
# Truncate snippet if too long
if snippet and len(snippet) > 500:
snippet = snippet[:497] + "..."
if link: # Only add if we found a valid link
search_results.append(
{
"link": link,
"title": title,
"snippet": snippet,
}
)
# Apply domain filtering if specified
if filter_list:
search_results = get_filtered_results(search_results, filter_list)
# Convert to SearchResult objects
return [
SearchResult(
link=result["link"],
title=result.get("title"),
snippet=result.get("snippet"),
)
for result in search_results
]
except Exception as ex:
log.error(f"Azure AI Search error: {ex}")
raise ex

4
backend/open_webui/retrieval/web/brave.py
View file

@ -36,7 +36,9 @@ def search_brave(
return [
SearchResult(
link=result["url"], title=result.get("title"), snippet=result.get("snippet")
link=result["url"],
title=result.get("title"),
snippet=result.get("description"),
)
for result in results[:count]
]

12
backend/open_webui/retrieval/web/duckduckgo.py
View file

@ -2,8 +2,8 @@ import logging
from typing import Optional
from open_webui.retrieval.web.main import SearchResult, get_filtered_results
from duckduckgo_search import DDGS
from duckduckgo_search.exceptions import RatelimitException
from ddgs import DDGS
from ddgs.exceptions import RatelimitException
from open_webui.env import SRC_LOG_LEVELS
log = logging.getLogger(__name__)
@ -11,7 +11,10 @@ log.setLevel(SRC_LOG_LEVELS["RAG"])
def search_duckduckgo(
query: str, count: int, filter_list: Optional[list[str]] = None
query: str,
count: int,
filter_list: Optional[list[str]] = None,
concurrent_requests: Optional[int] = None,
) -> list[SearchResult]:
"""
Search using DuckDuckGo's Search API and return the results as a list of SearchResult objects.
@ -25,6 +28,9 @@ def search_duckduckgo(
# Use the DDGS context manager to create a DDGS object
search_results = []
with DDGS() as ddgs:
if concurrent_requests:
ddgs.threads = concurrent_requests
# Use the ddgs.text() method to perform the search
try:
search_results = ddgs.text(

25
backend/open_webui/retrieval/web/external.py
View file

@ -2,27 +2,42 @@ import logging
from typing import Optional, List
import requests
from open_webui.retrieval.web.main import SearchResult, get_filtered_results
from fastapi import Request
from open_webui.env import SRC_LOG_LEVELS
from open_webui.retrieval.web.main import SearchResult, get_filtered_results
from open_webui.utils.headers import include_user_info_headers
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["RAG"])
def search_external(
request: Request,
external_url: str,
external_api_key: str,
query: str,
count: int,
filter_list: Optional[List[str]] = None,
user=None,
) -> List[SearchResult]:
try:
headers = {
"User-Agent": "Open WebUI (https://github.com/open-webui/open-webui) RAG Bot",
"Authorization": f"Bearer {external_api_key}",
}
headers = include_user_info_headers(headers, user)
chat_id = getattr(request.state, "chat_id", None)
if chat_id:
headers["X-OpenWebUI-Chat-Id"] = str(chat_id)
response = requests.post(
external_url,
headers={
"User-Agent": "Open WebUI (https://github.com/open-webui/open-webui) RAG Bot",
"Authorization": f"Bearer {external_api_key}",
},
headers=headers,
json={
"query": query,
"count": count,

28
backend/open_webui/retrieval/web/firecrawl.py
View file

@ -1,11 +1,10 @@
import logging
from typing import Optional, List
from urllib.parse import urljoin
import requests
from open_webui.retrieval.web.main import SearchResult, get_filtered_results
from open_webui.env import SRC_LOG_LEVELS
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["RAG"])
@ -18,27 +17,20 @@ def search_firecrawl(
filter_list: Optional[List[str]] = None,
) -> List[SearchResult]:
try:
firecrawl_search_url = urljoin(firecrawl_url, "/v1/search")
response = requests.post(
firecrawl_search_url,
headers={
"User-Agent": "Open WebUI (https://github.com/open-webui/open-webui) RAG Bot",
"Authorization": f"Bearer {firecrawl_api_key}",
},
json={
"query": query,
"limit": count,
},
from firecrawl import FirecrawlApp
firecrawl = FirecrawlApp(api_key=firecrawl_api_key, api_url=firecrawl_url)
response = firecrawl.search(
query=query, limit=count, ignore_invalid_urls=True, timeout=count * 3
)
response.raise_for_status()
results = response.json().get("data", [])
results = response.web
if filter_list:
results = get_filtered_results(results, filter_list)
results = [
SearchResult(
link=result.get("url"),
title=result.get("title"),
snippet=result.get("description"),
link=result.url,
title=result.title,
snippet=result.description,
)
for result in results[:count]
]

5
backend/open_webui/retrieval/web/google_pse.py
View file

@ -15,6 +15,7 @@ def search_google_pse(
query: str,
count: int,
filter_list: Optional[list[str]] = None,
referer: Optional[str] = None,
) -> list[SearchResult]:
"""Search using Google's Programmable Search Engine API and return the results as a list of SearchResult objects.
Handles pagination for counts greater than 10.
@ -30,7 +31,11 @@ def search_google_pse(
list[SearchResult]: A list of SearchResult objects.
"""
url = "https://www.googleapis.com/customsearch/v1"
headers = {"Content-Type": "application/json"}
if referer:
headers["Referer"] = referer
all_results = []
start_index = 1 # Google PSE start parameter is 1-based

24
backend/open_webui/retrieval/web/main.py
View file

@ -5,18 +5,38 @@ from urllib.parse import urlparse
from pydantic import BaseModel
from open_webui.retrieval.web.utils import resolve_hostname
from open_webui.utils.misc import is_string_allowed
def get_filtered_results(results, filter_list):
if not filter_list:
return results
filtered_results = []
for result in results:
url = result.get("url") or result.get("link", "")
url = result.get("url") or result.get("link", "") or result.get("href", "")
if not validators.url(url):
continue
domain = urlparse(url).netloc
if any(domain.endswith(filtered_domain) for filtered_domain in filter_list):
if not domain:
continue
hostnames = [domain]
try:
ipv4_addresses, ipv6_addresses = resolve_hostname(domain)
hostnames.extend(ipv4_addresses)
hostnames.extend(ipv6_addresses)
except Exception:
pass
if any(is_string_allowed(hostname, filter_list) for hostname in hostnames):
filtered_results.append(result)
continue
return filtered_results

51
backend/open_webui/retrieval/web/ollama.py Normal file
View file

@ -0,0 +1,51 @@
import logging
from dataclasses import dataclass
from typing import Optional
import requests
from open_webui.env import SRC_LOG_LEVELS
from open_webui.retrieval.web.main import SearchResult
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["RAG"])
def search_ollama_cloud(
url: str,
api_key: str,
query: str,
count: int,
filter_list: Optional[list[str]] = None,
) -> list[SearchResult]:
"""Search using Ollama Search API and return the results as a list of SearchResult objects.
Args:
api_key (str): A Ollama Search API key
query (str): The query to search for
count (int): Number of results to return
filter_list (Optional[list[str]]): List of domains to filter results by
"""
log.info(f"Searching with Ollama for query: {query}")
headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
payload = {"query": query, "max_results": count}
try:
response = requests.post(f"{url}/api/web_search", headers=headers, json=payload)
response.raise_for_status()
data = response.json()
results = data.get("results", [])
log.info(f"Found {len(results)} results")
return [
SearchResult(
link=result.get("url", ""),
title=result.get("title", ""),
snippet=result.get("content", ""),
)
for result in results
]
except Exception as e:
log.error(f"Error searching Ollama: {e}")
return []

76
backend/open_webui/retrieval/web/perplexity_search.py Normal file
View file

@ -0,0 +1,76 @@
import logging
from typing import Optional, Literal
import requests
from open_webui.retrieval.web.main import SearchResult, get_filtered_results
from open_webui.utils.headers import include_user_info_headers
from open_webui.env import SRC_LOG_LEVELS
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["RAG"])
def search_perplexity_search(
api_key: str,
query: str,
count: int,
filter_list: Optional[list[str]] = None,
api_url: str = "https://api.perplexity.ai/search",
user=None,
) -> list[SearchResult]:
"""Search using Perplexity API and return the results as a list of SearchResult objects.
Args:
api_key (str): A Perplexity API key
query (str): The query to search for
count (int): Maximum number of results to return
filter_list (Optional[list[str]]): List of domains to filter results
api_url (str): Custom API URL (defaults to https://api.perplexity.ai/search)
user: Optional user object for forwarding user info headers
"""
# Handle PersistentConfig object
if hasattr(api_key, "__str__"):
api_key = str(api_key)
if hasattr(api_url, "__str__"):
api_url = str(api_url)
try:
url = api_url
# Create payload for the API call
payload = {
"query": query,
"max_results": count,
}
headers = {
"Authorization": f"Bearer {api_key}",
"Content-Type": "application/json",
}
# Forward user info headers if user is provided
if user is not None:
headers = include_user_info_headers(headers, user)
# Make the API request
response = requests.request("POST", url, json=payload, headers=headers)
# Parse the JSON response
json_response = response.json()
# Extract citations from the response
results = json_response.get("results", [])
return [
SearchResult(
link=result["url"], title=result["title"], snippet=result["snippet"]
)
for result in results
]
except Exception as e:
log.error(f"Error searching with Perplexity Search API: {e}")
return []

180
backend/open_webui/retrieval/web/utils.py
View file

@ -4,7 +4,6 @@ import socket
import ssl
import urllib.parse
import urllib.request
from collections import defaultdict
from datetime import datetime, time, timedelta
from typing import (
Any,
@ -17,13 +16,15 @@ from typing import (
Union,
Literal,
)
from fastapi.concurrency import run_in_threadpool
import aiohttp
import certifi
import validators
from langchain_community.document_loaders import PlaywrightURLLoader, WebBaseLoader
from langchain_community.document_loaders.firecrawl import FireCrawlLoader
from langchain_community.document_loaders.base import BaseLoader
from langchain_core.documents import Document
from open_webui.retrieval.loaders.tavily import TavilyLoader
from open_webui.retrieval.loaders.external_web import ExternalWebLoader
from open_webui.constants import ERROR_MESSAGES
@ -38,17 +39,46 @@ from open_webui.config import (
TAVILY_EXTRACT_DEPTH,
EXTERNAL_WEB_LOADER_URL,
EXTERNAL_WEB_LOADER_API_KEY,
WEB_FETCH_FILTER_LIST,
)
from open_webui.env import SRC_LOG_LEVELS, AIOHTTP_CLIENT_SESSION_SSL
from open_webui.env import SRC_LOG_LEVELS
from open_webui.utils.misc import is_string_allowed
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["RAG"])
def resolve_hostname(hostname):
# Get address information
addr_info = socket.getaddrinfo(hostname, None)
# Extract IP addresses from address information
ipv4_addresses = [info[4][0] for info in addr_info if info[0] == socket.AF_INET]
ipv6_addresses = [info[4][0] for info in addr_info if info[0] == socket.AF_INET6]
return ipv4_addresses, ipv6_addresses
def validate_url(url: Union[str, Sequence[str]]):
if isinstance(url, str):
if isinstance(validators.url(url), validators.ValidationError):
raise ValueError(ERROR_MESSAGES.INVALID_URL)
parsed_url = urllib.parse.urlparse(url)
# Protocol validation - only allow http/https
if parsed_url.scheme not in ["http", "https"]:
log.warning(
f"Blocked non-HTTP(S) protocol: {parsed_url.scheme} in URL: {url}"
)
raise ValueError(ERROR_MESSAGES.INVALID_URL)
# Blocklist check using unified filtering logic
if WEB_FETCH_FILTER_LIST:
if not is_string_allowed(url, WEB_FETCH_FILTER_LIST):
log.warning(f"URL blocked by filter list: {url}")
raise ValueError(ERROR_MESSAGES.INVALID_URL)
if not ENABLE_RAG_LOCAL_WEB_FETCH:
# Local web fetch is disabled, filter out any URLs that resolve to private IP addresses
parsed_url = urllib.parse.urlparse(url)
@ -75,22 +105,12 @@ def safe_validate_urls(url: Sequence[str]) -> Sequence[str]:
try:
if validate_url(u):
valid_urls.append(u)
except ValueError:
except Exception as e:
log.debug(f"Invalid URL {u}: {str(e)}")
continue
return valid_urls
def resolve_hostname(hostname):
# Get address information
addr_info = socket.getaddrinfo(hostname, None)
# Extract IP addresses from address information
ipv4_addresses = [info[4][0] for info in addr_info if info[0] == socket.AF_INET]
ipv6_addresses = [info[4][0] for info in addr_info if info[0] == socket.AF_INET6]
return ipv4_addresses, ipv6_addresses
def extract_metadata(soup, url):
metadata = {"source": url}
if title := soup.find("title"):
@ -141,13 +161,13 @@ class RateLimitMixin:
class URLProcessingMixin:
def _verify_ssl_cert(self, url: str) -> bool:
async def _verify_ssl_cert(self, url: str) -> bool:
"""Verify SSL certificate for a URL."""
return verify_ssl_cert(url)
return await run_in_threadpool(verify_ssl_cert, url)
async def _safe_process_url(self, url: str) -> bool:
"""Perform safety checks before processing a URL."""
if self.verify_ssl and not self._verify_ssl_cert(url):
if self.verify_ssl and not await self._verify_ssl_cert(url):
raise ValueError(f"SSL certificate verification failed for {url}")
await self._wait_for_rate_limit()
return True
@ -188,13 +208,12 @@ class SafeFireCrawlLoader(BaseLoader, RateLimitMixin, URLProcessingMixin):
(uses FIRE_CRAWL_API_KEY environment variable if not provided).
api_url: Base URL for FireCrawl API. Defaults to official API endpoint.
mode: Operation mode selection:
- 'crawl': Website crawling mode (default)
- 'scrape': Direct page scraping
- 'crawl': Website crawling mode
- 'scrape': Direct page scraping (default)
- 'map': Site map generation
proxy: Proxy override settings for the FireCrawl API.
params: The parameters to pass to the Firecrawl API.
Examples include crawlerOptions.
For more details, visit: https://github.com/mendableai/firecrawl-py
For more details, visit: https://docs.firecrawl.dev/sdks/python#batch-scrape
"""
proxy_server = proxy.get("server") if proxy else None
if trust_env and not proxy_server:
@ -214,50 +233,88 @@ class SafeFireCrawlLoader(BaseLoader, RateLimitMixin, URLProcessingMixin):
self.api_key = api_key
self.api_url = api_url
self.mode = mode
self.params = params
self.params = params or {}
def lazy_load(self) -> Iterator[Document]:
"""Load documents concurrently using FireCrawl."""
for url in self.web_paths:
try:
self._safe_process_url_sync(url)
loader = FireCrawlLoader(
url=url,
api_key=self.api_key,
api_url=self.api_url,
mode=self.mode,
params=self.params,
"""Load documents using FireCrawl batch_scrape."""
log.debug(
"Starting FireCrawl batch scrape for %d URLs, mode: %s, params: %s",
len(self.web_paths),
self.mode,
self.params,
)
try:
from firecrawl import FirecrawlApp
firecrawl = FirecrawlApp(api_key=self.api_key, api_url=self.api_url)
result = firecrawl.batch_scrape(
self.web_paths,
formats=["markdown"],
skip_tls_verification=not self.verify_ssl,
ignore_invalid_urls=True,
remove_base64_images=True,
max_age=300000, # 5 minutes https://docs.firecrawl.dev/features/fast-scraping#common-maxage-values
wait_timeout=len(self.web_paths) * 3,
**self.params,
)
if result.status != "completed":
raise RuntimeError(
f"FireCrawl batch scrape did not complete successfully. result: {result}"
)
for document in loader.lazy_load():
if not document.metadata.get("source"):
document.metadata["source"] = document.metadata.get("sourceURL")
yield document
except Exception as e:
if self.continue_on_failure:
log.exception(f"Error loading {url}: {e}")
continue
for data in result.data:
metadata = data.metadata or {}
yield Document(
page_content=data.markdown or "",
metadata={"source": metadata.url or metadata.source_url or ""},
)
except Exception as e:
if self.continue_on_failure:
log.exception(f"Error extracting content from URLs: {e}")
else:
raise e
async def alazy_load(self):
"""Async version of lazy_load."""
for url in self.web_paths:
try:
await self._safe_process_url(url)
loader = FireCrawlLoader(
url=url,
api_key=self.api_key,
api_url=self.api_url,
mode=self.mode,
params=self.params,
log.debug(
"Starting FireCrawl batch scrape for %d URLs, mode: %s, params: %s",
len(self.web_paths),
self.mode,
self.params,
)
try:
from firecrawl import FirecrawlApp
firecrawl = FirecrawlApp(api_key=self.api_key, api_url=self.api_url)
result = firecrawl.batch_scrape(
self.web_paths,
formats=["markdown"],
skip_tls_verification=not self.verify_ssl,
ignore_invalid_urls=True,
remove_base64_images=True,
max_age=300000, # 5 minutes https://docs.firecrawl.dev/features/fast-scraping#common-maxage-values
wait_timeout=len(self.web_paths) * 3,
**self.params,
)
if result.status != "completed":
raise RuntimeError(
f"FireCrawl batch scrape did not complete successfully. result: {result}"
)
async for document in loader.alazy_load():
if not document.metadata.get("source"):
document.metadata["source"] = document.metadata.get("sourceURL")
yield document
except Exception as e:
if self.continue_on_failure:
log.exception(f"Error loading {url}: {e}")
continue
for data in result.data:
metadata = data.metadata or {}
yield Document(
page_content=data.markdown or "",
metadata={"source": metadata.url or metadata.source_url or ""},
)
except Exception as e:
if self.continue_on_failure:
log.exception(f"Error extracting content from URLs: {e}")
else:
raise e
@ -517,6 +574,7 @@ class SafeWebBaseLoader(WebBaseLoader):
async with session.get(
url,
**(self.requests_kwargs | kwargs),
allow_redirects=False,
) as response:
if self.raise_for_status:
response.raise_for_status()
@ -602,6 +660,10 @@ def get_web_loader(
# Check if the URLs are valid
safe_urls = safe_validate_urls([urls] if isinstance(urls, str) else urls)
if not safe_urls:
log.warning(f"All provided URLs were blocked or invalid: {urls}")
raise ValueError(ERROR_MESSAGES.INVALID_URL)
web_loader_args = {
"web_paths": safe_urls,
"verify_ssl": verify_ssl,
@ -614,7 +676,7 @@ def get_web_loader(
WebLoaderClass = SafeWebBaseLoader
if WEB_LOADER_ENGINE.value == "playwright":
WebLoaderClass = SafePlaywrightURLLoader
web_loader_args["playwright_timeout"] = PLAYWRIGHT_TIMEOUT.value * 1000
web_loader_args["playwright_timeout"] = PLAYWRIGHT_TIMEOUT.value
if PLAYWRIGHT_WS_URL.value:
web_loader_args["playwright_ws_url"] = PLAYWRIGHT_WS_URL.value

385
backend/open_webui/routers/audio.py
View file

@ -3,8 +3,9 @@ import json
import logging
import os
import uuid
import html
import base64
from functools import lru_cache
from pathlib import Path
from pydub import AudioSegment
from pydub.silence import split_on_silence
from concurrent.futures import ThreadPoolExecutor
@ -15,7 +16,6 @@ import aiohttp
import aiofiles
import requests
import mimetypes
from urllib.parse import quote
from fastapi import (
Depends,
@ -34,18 +34,20 @@ from pydantic import BaseModel
from open_webui.utils.auth import get_admin_user, get_verified_user
from open_webui.utils.headers import include_user_info_headers
from open_webui.config import (
WHISPER_MODEL_AUTO_UPDATE,
WHISPER_MODEL_DIR,
CACHE_DIR,
WHISPER_LANGUAGE,
ELEVENLABS_API_BASE_URL,
)
from open_webui.constants import ERROR_MESSAGES
from open_webui.env import (
ENV,
AIOHTTP_CLIENT_SESSION_SSL,
AIOHTTP_CLIENT_TIMEOUT,
ENV,
SRC_LOG_LEVELS,
DEVICE_TYPE,
ENABLE_FORWARD_USER_INFO_HEADERS,
@ -154,6 +156,7 @@ def set_faster_whisper_model(model: str, auto_update: bool = False):
class TTSConfigForm(BaseModel):
OPENAI_API_BASE_URL: str
OPENAI_API_KEY: str
OPENAI_PARAMS: Optional[dict] = None
API_KEY: str
ENGINE: str
MODEL: str
@ -177,6 +180,9 @@ class STTConfigForm(BaseModel):
AZURE_LOCALES: str
AZURE_BASE_URL: str
AZURE_MAX_SPEAKERS: str
MISTRAL_API_KEY: str
MISTRAL_API_BASE_URL: str
MISTRAL_USE_CHAT_COMPLETIONS: bool
class AudioConfigUpdateForm(BaseModel):
@ -190,6 +196,7 @@ async def get_audio_config(request: Request, user=Depends(get_admin_user)):
"tts": {
"OPENAI_API_BASE_URL": request.app.state.config.TTS_OPENAI_API_BASE_URL,
"OPENAI_API_KEY": request.app.state.config.TTS_OPENAI_API_KEY,
"OPENAI_PARAMS": request.app.state.config.TTS_OPENAI_PARAMS,
"API_KEY": request.app.state.config.TTS_API_KEY,
"ENGINE": request.app.state.config.TTS_ENGINE,
"MODEL": request.app.state.config.TTS_MODEL,
@ -212,6 +219,9 @@ async def get_audio_config(request: Request, user=Depends(get_admin_user)):
"AZURE_LOCALES": request.app.state.config.AUDIO_STT_AZURE_LOCALES,
"AZURE_BASE_URL": request.app.state.config.AUDIO_STT_AZURE_BASE_URL,
"AZURE_MAX_SPEAKERS": request.app.state.config.AUDIO_STT_AZURE_MAX_SPEAKERS,
"MISTRAL_API_KEY": request.app.state.config.AUDIO_STT_MISTRAL_API_KEY,
"MISTRAL_API_BASE_URL": request.app.state.config.AUDIO_STT_MISTRAL_API_BASE_URL,
"MISTRAL_USE_CHAT_COMPLETIONS": request.app.state.config.AUDIO_STT_MISTRAL_USE_CHAT_COMPLETIONS,
},
}
@ -222,6 +232,7 @@ async def update_audio_config(
):
request.app.state.config.TTS_OPENAI_API_BASE_URL = form_data.tts.OPENAI_API_BASE_URL
request.app.state.config.TTS_OPENAI_API_KEY = form_data.tts.OPENAI_API_KEY
request.app.state.config.TTS_OPENAI_PARAMS = form_data.tts.OPENAI_PARAMS
request.app.state.config.TTS_API_KEY = form_data.tts.API_KEY
request.app.state.config.TTS_ENGINE = form_data.tts.ENGINE
request.app.state.config.TTS_MODEL = form_data.tts.MODEL
@ -252,6 +263,13 @@ async def update_audio_config(
request.app.state.config.AUDIO_STT_AZURE_MAX_SPEAKERS = (
form_data.stt.AZURE_MAX_SPEAKERS
)
request.app.state.config.AUDIO_STT_MISTRAL_API_KEY = form_data.stt.MISTRAL_API_KEY
request.app.state.config.AUDIO_STT_MISTRAL_API_BASE_URL = (
form_data.stt.MISTRAL_API_BASE_URL
)
request.app.state.config.AUDIO_STT_MISTRAL_USE_CHAT_COMPLETIONS = (
form_data.stt.MISTRAL_USE_CHAT_COMPLETIONS
)
if request.app.state.config.STT_ENGINE == "":
request.app.state.faster_whisper_model = set_faster_whisper_model(
@ -262,12 +280,13 @@ async def update_audio_config(
return {
"tts": {
"OPENAI_API_BASE_URL": request.app.state.config.TTS_OPENAI_API_BASE_URL,
"OPENAI_API_KEY": request.app.state.config.TTS_OPENAI_API_KEY,
"API_KEY": request.app.state.config.TTS_API_KEY,
"ENGINE": request.app.state.config.TTS_ENGINE,
"MODEL": request.app.state.config.TTS_MODEL,
"VOICE": request.app.state.config.TTS_VOICE,
"OPENAI_API_BASE_URL": request.app.state.config.TTS_OPENAI_API_BASE_URL,
"OPENAI_API_KEY": request.app.state.config.TTS_OPENAI_API_KEY,
"OPENAI_PARAMS": request.app.state.config.TTS_OPENAI_PARAMS,
"API_KEY": request.app.state.config.TTS_API_KEY,
"SPLIT_ON": request.app.state.config.TTS_SPLIT_ON,
"AZURE_SPEECH_REGION": request.app.state.config.TTS_AZURE_SPEECH_REGION,
"AZURE_SPEECH_BASE_URL": request.app.state.config.TTS_AZURE_SPEECH_BASE_URL,
@ -286,6 +305,9 @@ async def update_audio_config(
"AZURE_LOCALES": request.app.state.config.AUDIO_STT_AZURE_LOCALES,
"AZURE_BASE_URL": request.app.state.config.AUDIO_STT_AZURE_BASE_URL,
"AZURE_MAX_SPEAKERS": request.app.state.config.AUDIO_STT_AZURE_MAX_SPEAKERS,
"MISTRAL_API_KEY": request.app.state.config.AUDIO_STT_MISTRAL_API_KEY,
"MISTRAL_API_BASE_URL": request.app.state.config.AUDIO_STT_MISTRAL_API_BASE_URL,
"MISTRAL_USE_CHAT_COMPLETIONS": request.app.state.config.AUDIO_STT_MISTRAL_USE_CHAT_COMPLETIONS,
},
}
@ -328,6 +350,7 @@ async def speech(request: Request, user=Depends(get_verified_user)):
log.exception(e)
raise HTTPException(status_code=400, detail="Invalid JSON payload")
r = None
if request.app.state.config.TTS_ENGINE == "openai":
payload["model"] = request.app.state.config.TTS_MODEL
@ -336,32 +359,32 @@ async def speech(request: Request, user=Depends(get_verified_user)):
async with aiohttp.ClientSession(
timeout=timeout, trust_env=True
) as session:
async with session.post(
payload = {
**payload,
**(request.app.state.config.TTS_OPENAI_PARAMS or {}),
}
headers = {
"Content-Type": "application/json",
"Authorization": f"Bearer {request.app.state.config.TTS_OPENAI_API_KEY}",
}
if ENABLE_FORWARD_USER_INFO_HEADERS:
headers = include_user_info_headers(headers, user)
r = await session.post(
url=f"{request.app.state.config.TTS_OPENAI_API_BASE_URL}/audio/speech",
json=payload,
headers={
"Content-Type": "application/json",
"Authorization": f"Bearer {request.app.state.config.TTS_OPENAI_API_KEY}",
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS
else {}
),
},
headers=headers,
ssl=AIOHTTP_CLIENT_SESSION_SSL,
) as r:
r.raise_for_status()
)
async with aiofiles.open(file_path, "wb") as f:
await f.write(await r.read())
r.raise_for_status()
async with aiofiles.open(file_body_path, "w") as f:
await f.write(json.dumps(payload))
async with aiofiles.open(file_path, "wb") as f:
await f.write(await r.read())
async with aiofiles.open(file_body_path, "w") as f:
await f.write(json.dumps(payload))
return FileResponse(file_path)
@ -369,18 +392,22 @@ async def speech(request: Request, user=Depends(get_verified_user)):
log.exception(e)
detail = None
try:
if r.status != 200:
res = await r.json()
status_code = 500
detail = f"Open WebUI: Server Connection Error"
if r is not None:
status_code = r.status
try:
res = await r.json()
if "error" in res:
detail = f"External: {res['error'].get('message', '')}"
except Exception:
detail = f"External: {e}"
detail = f"External: {res['error']}"
except Exception:
detail = f"External: {e}"
raise HTTPException(
status_code=getattr(r, "status", 500) if r else 500,
detail=detail if detail else "Open WebUI: Server Connection Error",
status_code=status_code,
detail=detail,
)
elif request.app.state.config.TTS_ENGINE == "elevenlabs":
@ -398,7 +425,7 @@ async def speech(request: Request, user=Depends(get_verified_user)):
timeout=timeout, trust_env=True
) as session:
async with session.post(
f"https://api.elevenlabs.io/v1/text-to-speech/{voice_id}",
f"{ELEVENLABS_API_BASE_URL}/v1/text-to-speech/{voice_id}",
json={
"text": payload["input"],
"model_id": request.app.state.config.TTS_MODEL,
@ -453,7 +480,7 @@ async def speech(request: Request, user=Depends(get_verified_user)):
try:
data = f"""<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xml:lang="{locale}">
<voice name="{language}">{payload["input"]}</voice>
<voice name="{language}">{html.escape(payload["input"])}</voice>
</speak>"""
timeout = aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT)
async with aiohttp.ClientSession(
@ -537,13 +564,18 @@ async def speech(request: Request, user=Depends(get_verified_user)):
return FileResponse(file_path)
def transcription_handler(request, file_path, metadata):
def transcription_handler(request, file_path, metadata, user=None):
filename = os.path.basename(file_path)
file_dir = os.path.dirname(file_path)
id = filename.split(".")[0]
metadata = metadata or {}
languages = [
metadata.get("language", None) if not WHISPER_LANGUAGE else WHISPER_LANGUAGE,
None, # Always fallback to None in case transcription fails
]
if request.app.state.config.STT_ENGINE == "":
if request.app.state.faster_whisper_model is None:
request.app.state.faster_whisper_model = set_faster_whisper_model(
@ -555,7 +587,7 @@ def transcription_handler(request, file_path, metadata):
file_path,
beam_size=5,
vad_filter=request.app.state.config.WHISPER_VAD_FILTER,
language=metadata.get("language") or WHISPER_LANGUAGE,
language=languages[0],
)
log.info(
"Detected language '%s' with probability %f"
@ -575,21 +607,30 @@ def transcription_handler(request, file_path, metadata):
elif request.app.state.config.STT_ENGINE == "openai":
r = None
try:
r = requests.post(
url=f"{request.app.state.config.STT_OPENAI_API_BASE_URL}/audio/transcriptions",
headers={
"Authorization": f"Bearer {request.app.state.config.STT_OPENAI_API_KEY}"
},
files={"file": (filename, open(file_path, "rb"))},
data={
for language in languages:
payload = {
"model": request.app.state.config.STT_MODEL,
**(
{"language": metadata.get("language")}
if metadata.get("language")
else {}
),
},
)
}
if language:
payload["language"] = language
headers = {
"Authorization": f"Bearer {request.app.state.config.STT_OPENAI_API_KEY}"
}
if user and ENABLE_FORWARD_USER_INFO_HEADERS:
headers = include_user_info_headers(headers, user)
r = requests.post(
url=f"{request.app.state.config.STT_OPENAI_API_BASE_URL}/audio/transcriptions",
headers=headers,
files={"file": (filename, open(file_path, "rb"))},
data=payload,
)
if r.status_code == 200:
# Successful transcription
break
r.raise_for_status()
data = r.json()
@ -631,18 +672,26 @@ def transcription_handler(request, file_path, metadata):
"Content-Type": mime,
}
# Add model if specified
params = {}
if request.app.state.config.STT_MODEL:
params["model"] = request.app.state.config.STT_MODEL
for language in languages:
params = {}
if request.app.state.config.STT_MODEL:
params["model"] = request.app.state.config.STT_MODEL
if language:
params["language"] = language
# Make request to Deepgram API
r = requests.post(
"https://api.deepgram.com/v1/listen?smart_format=true",
headers=headers,
params=params,
data=file_data,
)
if r.status_code == 200:
# Successful transcription
break
# Make request to Deepgram API
r = requests.post(
"https://api.deepgram.com/v1/listen?smart_format=true",
headers=headers,
params=params,
data=file_data,
)
r.raise_for_status()
response_data = r.json()
@ -795,8 +844,190 @@ def transcription_handler(request, file_path, metadata):
detail=detail if detail else "Open WebUI: Server Connection Error",
)
elif request.app.state.config.STT_ENGINE == "mistral":
# Check file exists
if not os.path.exists(file_path):
raise HTTPException(status_code=400, detail="Audio file not found")
def transcribe(request: Request, file_path: str, metadata: Optional[dict] = None):
# Check file size
file_size = os.path.getsize(file_path)
if file_size > MAX_FILE_SIZE:
raise HTTPException(
status_code=400,
detail=f"File size exceeds limit of {MAX_FILE_SIZE_MB}MB",
)
api_key = request.app.state.config.AUDIO_STT_MISTRAL_API_KEY
api_base_url = (
request.app.state.config.AUDIO_STT_MISTRAL_API_BASE_URL
or "https://api.mistral.ai/v1"
)
use_chat_completions = (
request.app.state.config.AUDIO_STT_MISTRAL_USE_CHAT_COMPLETIONS
)
if not api_key:
raise HTTPException(
status_code=400,
detail="Mistral API key is required for Mistral STT",
)
r = None
try:
# Use voxtral-mini-latest as the default model for transcription
model = request.app.state.config.STT_MODEL or "voxtral-mini-latest"
log.info(
f"Mistral STT - model: {model}, "
f"method: {'chat_completions' if use_chat_completions else 'transcriptions'}"
)
if use_chat_completions:
# Use chat completions API with audio input
# This method requires mp3 or wav format
audio_file_to_use = file_path
if is_audio_conversion_required(file_path):
log.debug("Converting audio to mp3 for chat completions API")
converted_path = convert_audio_to_mp3(file_path)
if converted_path:
audio_file_to_use = converted_path
else:
log.error("Audio conversion failed")
raise HTTPException(
status_code=500,
detail="Audio conversion failed. Chat completions API requires mp3 or wav format.",
)
# Read and encode audio file as base64
with open(audio_file_to_use, "rb") as audio_file:
audio_base64 = base64.b64encode(audio_file.read()).decode("utf-8")
# Prepare chat completions request
url = f"{api_base_url}/chat/completions"
# Add language instruction if specified
language = metadata.get("language", None) if metadata else None
if language:
text_instruction = f"Transcribe this audio exactly as spoken in {language}. Do not translate it."
else:
text_instruction = "Transcribe this audio exactly as spoken in its original language. Do not translate it to another language."
payload = {
"model": model,
"messages": [
{
"role": "user",
"content": [
{
"type": "input_audio",
"input_audio": audio_base64,
},
{"type": "text", "text": text_instruction},
],
}
],
}
r = requests.post(
url=url,
json=payload,
headers={
"Authorization": f"Bearer {api_key}",
"Content-Type": "application/json",
},
)
r.raise_for_status()
response = r.json()
# Extract transcript from chat completion response
transcript = (
response.get("choices", [{}])[0]
.get("message", {})
.get("content", "")
.strip()
)
if not transcript:
raise ValueError("Empty transcript in response")
data = {"text": transcript}
else:
# Use dedicated transcriptions API
url = f"{api_base_url}/audio/transcriptions"
# Determine the MIME type
mime_type, _ = mimetypes.guess_type(file_path)
if not mime_type:
mime_type = "audio/webm"
# Use context manager to ensure file is properly closed
with open(file_path, "rb") as audio_file:
files = {"file": (filename, audio_file, mime_type)}
data_form = {"model": model}
# Add language if specified in metadata
language = metadata.get("language", None) if metadata else None
if language:
data_form["language"] = language
r = requests.post(
url=url,
files=files,
data=data_form,
headers={
"Authorization": f"Bearer {api_key}",
},
)
r.raise_for_status()
response = r.json()
# Extract transcript from response
transcript = response.get("text", "").strip()
if not transcript:
raise ValueError("Empty transcript in response")
data = {"text": transcript}
# Save transcript to json file (consistent with other providers)
transcript_file = f"{file_dir}/{id}.json"
with open(transcript_file, "w") as f:
json.dump(data, f)
log.debug(data)
return data
except ValueError as e:
log.exception("Error parsing Mistral response")
raise HTTPException(
status_code=500,
detail=f"Failed to parse Mistral response: {str(e)}",
)
except requests.exceptions.RequestException as e:
log.exception(e)
detail = None
try:
if r is not None and r.status_code != 200:
res = r.json()
if "error" in res:
detail = f"External: {res['error'].get('message', '')}"
else:
detail = f"External: {r.text}"
except Exception:
detail = f"External: {e}"
raise HTTPException(
status_code=getattr(r, "status_code", 500) if r else 500,
detail=detail if detail else "Open WebUI: Server Connection Error",
)
def transcribe(
request: Request, file_path: str, metadata: Optional[dict] = None, user=None
):
log.info(f"transcribe: {file_path} {metadata}")
if is_audio_conversion_required(file_path):
@ -823,7 +1054,9 @@ def transcribe(request: Request, file_path: str, metadata: Optional[dict] = None
with ThreadPoolExecutor() as executor:
# Submit tasks for each chunk_path
futures = [
executor.submit(transcription_handler, request, chunk_path, metadata)
executor.submit(
transcription_handler, request, chunk_path, metadata, user
)
for chunk_path in chunk_paths
]
# Gather results as they complete
@ -920,14 +1153,18 @@ def transcription(
):
log.info(f"file.content_type: {file.content_type}")
supported_content_types = request.app.state.config.STT_SUPPORTED_CONTENT_TYPES or [
"audio/*",
"video/webm",
]
stt_supported_content_types = getattr(
request.app.state.config, "STT_SUPPORTED_CONTENT_TYPES", []
)
if not any(
fnmatch(file.content_type, content_type)
for content_type in supported_content_types
for content_type in (
stt_supported_content_types
if stt_supported_content_types
and any(t.strip() for t in stt_supported_content_types)
else ["audio/*", "video/webm"]
)
):
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
@ -954,7 +1191,7 @@ def transcription(
if language:
metadata = {"language": language}
result = transcribe(request, file_path, metadata)
result = transcribe(request, file_path, metadata, user)
return {
**result,
@ -1000,7 +1237,7 @@ def get_available_models(request: Request) -> list[dict]:
elif request.app.state.config.TTS_ENGINE == "elevenlabs":
try:
response = requests.get(
"https://api.elevenlabs.io/v1/models",
f"{ELEVENLABS_API_BASE_URL}/v1/models",
headers={
"xi-api-key": request.app.state.config.TTS_API_KEY,
"Content-Type": "application/json",
@ -1104,7 +1341,7 @@ def get_elevenlabs_voices(api_key: str) -> dict:
try:
# TODO: Add retries
response = requests.get(
"https://api.elevenlabs.io/v1/voices",
f"{ELEVENLABS_API_BASE_URL}/v1/voices",
headers={
"xi-api-key": api_key,
"Content-Type": "application/json",

228
backend/open_webui/routers/auths.py
View file

@ -4,6 +4,7 @@ import time
import datetime
import logging
from aiohttp import ClientSession
import urllib
from open_webui.models.auths import (
AddUserForm,
@ -15,11 +16,11 @@ from open_webui.models.auths import (
SigninResponse,
SignupForm,
UpdatePasswordForm,
UpdateProfileForm,
UserResponse,
)
from open_webui.models.users import Users
from open_webui.models.users import Users, UpdateProfileForm
from open_webui.models.groups import Groups
from open_webui.models.oauth_sessions import OAuthSessions
from open_webui.constants import ERROR_MESSAGES, WEBHOOK_MESSAGES
from open_webui.env import (
@ -30,16 +31,25 @@ from open_webui.env import (
WEBUI_AUTH_COOKIE_SAME_SITE,
WEBUI_AUTH_COOKIE_SECURE,
WEBUI_AUTH_SIGNOUT_REDIRECT_URL,
ENABLE_INITIAL_ADMIN_SIGNUP,
SRC_LOG_LEVELS,
)
from fastapi import APIRouter, Depends, HTTPException, Request, status
from fastapi.responses import RedirectResponse, Response, JSONResponse
from open_webui.config import OPENID_PROVIDER_URL, ENABLE_OAUTH_SIGNUP, ENABLE_LDAP
from open_webui.config import (
OPENID_PROVIDER_URL,
ENABLE_OAUTH_SIGNUP,
ENABLE_LDAP,
ENABLE_PASSWORD_AUTH,
)
from pydantic import BaseModel
from open_webui.utils.misc import parse_duration, validate_email_format
from open_webui.utils.auth import (
validate_password,
verify_password,
decode_token,
invalidate_token,
create_api_key,
create_token,
get_admin_user,
@ -49,7 +59,7 @@ from open_webui.utils.auth import (
get_http_authorization_cred,
)
from open_webui.utils.webhook import post_webhook
from open_webui.utils.access_control import get_permissions
from open_webui.utils.access_control import get_permissions, has_permission
from typing import Optional, List
@ -73,7 +83,13 @@ class SessionUserResponse(Token, UserResponse):
permissions: Optional[dict] = None
@router.get("/", response_model=SessionUserResponse)
class SessionUserInfoResponse(SessionUserResponse):
bio: Optional[str] = None
gender: Optional[str] = None
date_of_birth: Optional[datetime.date] = None
@router.get("/", response_model=SessionUserInfoResponse)
async def get_session_user(
request: Request, response: Response, user=Depends(get_current_user)
):
@ -121,6 +137,9 @@ async def get_session_user(
"name": user.name,
"role": user.role,
"profile_image_url": user.profile_image_url,
"bio": user.bio,
"gender": user.gender,
"date_of_birth": user.date_of_birth,
"permissions": user_permissions,
}
@ -137,7 +156,7 @@ async def update_profile(
if session_user:
user = Users.update_user_by_id(
session_user.id,
{"profile_image_url": form_data.profile_image_url, "name": form_data.name},
form_data.model_dump(),
)
if user:
return user
@ -159,13 +178,19 @@ async def update_password(
if WEBUI_AUTH_TRUSTED_EMAIL_HEADER:
raise HTTPException(400, detail=ERROR_MESSAGES.ACTION_PROHIBITED)
if session_user:
user = Auths.authenticate_user(session_user.email, form_data.password)
user = Auths.authenticate_user(
session_user.email, lambda pw: verify_password(form_data.password, pw)
)
if user:
try:
validate_password(form_data.password)
except Exception as e:
raise HTTPException(400, detail=str(e))
hashed = get_password_hash(form_data.new_password)
return Auths.update_user_password_by_id(user.id, hashed)
else:
raise HTTPException(400, detail=ERROR_MESSAGES.INVALID_PASSWORD)
raise HTTPException(400, detail=ERROR_MESSAGES.INCORRECT_PASSWORD)
else:
raise HTTPException(400, detail=ERROR_MESSAGES.INVALID_CRED)
@ -175,7 +200,17 @@ async def update_password(
############################
@router.post("/ldap", response_model=SessionUserResponse)
async def ldap_auth(request: Request, response: Response, form_data: LdapForm):
ENABLE_LDAP = request.app.state.config.ENABLE_LDAP
# Security checks FIRST - before loading any config
if not request.app.state.config.ENABLE_LDAP:
raise HTTPException(400, detail="LDAP authentication is not enabled")
if not ENABLE_PASSWORD_AUTH:
raise HTTPException(
status_code=status.HTTP_403_FORBIDDEN,
detail=ERROR_MESSAGES.ACTION_PROHIBITED,
)
# NOW load LDAP config variables
LDAP_SERVER_LABEL = request.app.state.config.LDAP_SERVER_LABEL
LDAP_SERVER_HOST = request.app.state.config.LDAP_SERVER_HOST
LDAP_SERVER_PORT = request.app.state.config.LDAP_SERVER_PORT
@ -196,9 +231,6 @@ async def ldap_auth(request: Request, response: Response, form_data: LdapForm):
else "ALL"
)
if not ENABLE_LDAP:
raise HTTPException(400, detail="LDAP authentication is not enabled")
try:
tls = Tls(
validate=LDAP_VALIDATE_CERT,
@ -351,11 +383,9 @@ async def ldap_auth(request: Request, response: Response, form_data: LdapForm):
user = Users.get_user_by_email(email)
if not user:
try:
user_count = Users.get_num_users()
role = (
"admin"
if user_count == 0
if not Users.has_users()
else request.app.state.config.DEFAULT_USER_ROLE
)
@ -455,6 +485,12 @@ async def ldap_auth(request: Request, response: Response, form_data: LdapForm):
@router.post("/signin", response_model=SessionUserResponse)
async def signin(request: Request, response: Response, form_data: SigninForm):
if not ENABLE_PASSWORD_AUTH:
raise HTTPException(
status_code=status.HTTP_403_FORBIDDEN,
detail=ERROR_MESSAGES.ACTION_PROHIBITED,
)
if WEBUI_AUTH_TRUSTED_EMAIL_HEADER:
if WEBUI_AUTH_TRUSTED_EMAIL_HEADER not in request.headers:
raise HTTPException(400, detail=ERROR_MESSAGES.INVALID_TRUSTED_HEADER)
@ -464,6 +500,10 @@ async def signin(request: Request, response: Response, form_data: SigninForm):
if WEBUI_AUTH_TRUSTED_NAME_HEADER:
name = request.headers.get(WEBUI_AUTH_TRUSTED_NAME_HEADER, email)
try:
name = urllib.parse.unquote(name, encoding="utf-8")
except Exception as e:
pass
if not Users.get_user_by_email(email.lower()):
await signup(
@ -487,9 +527,11 @@ async def signin(request: Request, response: Response, form_data: SigninForm):
admin_password = "admin"
if Users.get_user_by_email(admin_email.lower()):
user = Auths.authenticate_user(admin_email.lower(), admin_password)
user = Auths.authenticate_user(
admin_email.lower(), lambda pw: verify_password(admin_password, pw)
)
else:
if Users.get_num_users() != 0:
if Users.has_users():
raise HTTPException(400, detail=ERROR_MESSAGES.EXISTING_USERS)
await signup(
@ -498,9 +540,22 @@ async def signin(request: Request, response: Response, form_data: SigninForm):
SignupForm(email=admin_email, password=admin_password, name="User"),
)
user = Auths.authenticate_user(admin_email.lower(), admin_password)
user = Auths.authenticate_user(
admin_email.lower(), lambda pw: verify_password(admin_password, pw)
)
else:
user = Auths.authenticate_user(form_data.email.lower(), form_data.password)
password_bytes = form_data.password.encode("utf-8")
if len(password_bytes) > 72:
# TODO: Implement other hashing algorithms that support longer passwords
log.info("Password too long, truncating to 72 bytes for bcrypt")
password_bytes = password_bytes[:72]
# decode safely — ignore incomplete UTF-8 sequences
form_data.password = password_bytes.decode("utf-8", errors="ignore")
user = Auths.authenticate_user(
form_data.email.lower(), lambda pw: verify_password(form_data.password, pw)
)
if user:
@ -556,22 +611,23 @@ async def signin(request: Request, response: Response, form_data: SigninForm):
@router.post("/signup", response_model=SessionUserResponse)
async def signup(request: Request, response: Response, form_data: SignupForm):
has_users = Users.has_users()
if WEBUI_AUTH:
if (
not request.app.state.config.ENABLE_SIGNUP
or not request.app.state.config.ENABLE_LOGIN_FORM
):
raise HTTPException(
status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.ACCESS_PROHIBITED
)
if has_users or not ENABLE_INITIAL_ADMIN_SIGNUP:
raise HTTPException(
status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.ACCESS_PROHIBITED
)
else:
if Users.get_num_users() != 0:
if has_users:
raise HTTPException(
status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.ACCESS_PROHIBITED
)
user_count = Users.get_num_users()
if not validate_email_format(form_data.email.lower()):
raise HTTPException(
status.HTTP_400_BAD_REQUEST, detail=ERROR_MESSAGES.INVALID_EMAIL_FORMAT
@ -581,18 +637,14 @@ async def signup(request: Request, response: Response, form_data: SignupForm):
raise HTTPException(400, detail=ERROR_MESSAGES.EMAIL_TAKEN)
try:
role = (
"admin" if user_count == 0 else request.app.state.config.DEFAULT_USER_ROLE
)
# The password passed to bcrypt must be 72 bytes or fewer. If it is longer, it will be truncated before hashing.
if len(form_data.password.encode("utf-8")) > 72:
raise HTTPException(
status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.PASSWORD_TOO_LONG,
)
try:
validate_password(form_data.password)
except Exception as e:
raise HTTPException(400, detail=str(e))
hashed = get_password_hash(form_data.password)
role = "admin" if not has_users else request.app.state.config.DEFAULT_USER_ROLE
user = Auths.insert_new_auth(
form_data.email.lower(),
hashed,
@ -629,7 +681,7 @@ async def signup(request: Request, response: Response, form_data: SignupForm):
)
if request.app.state.config.WEBHOOK_URL:
post_webhook(
await post_webhook(
request.app.state.WEBUI_NAME,
request.app.state.config.WEBHOOK_URL,
WEBHOOK_MESSAGES.USER_SIGNUP(user.name),
@ -644,10 +696,14 @@ async def signup(request: Request, response: Response, form_data: SignupForm):
user.id, request.app.state.config.USER_PERMISSIONS
)
if user_count == 0:
if not has_users:
# Disable signup after the first user is created
request.app.state.config.ENABLE_SIGNUP = False
default_group_id = getattr(request.app.state.config, "DEFAULT_GROUP_ID", "")
if default_group_id and default_group_id:
Groups.add_users_to_group(default_group_id, [user.id])
return {
"token": token,
"token_type": "Bearer",
@ -668,38 +724,66 @@ async def signup(request: Request, response: Response, form_data: SignupForm):
@router.get("/signout")
async def signout(request: Request, response: Response):
# get auth token from headers or cookies
token = None
auth_header = request.headers.get("Authorization")
if auth_header:
auth_cred = get_http_authorization_cred(auth_header)
token = auth_cred.credentials
else:
token = request.cookies.get("token")
if token:
await invalidate_token(request, token)
response.delete_cookie("token")
response.delete_cookie("oui-session")
response.delete_cookie("oauth_id_token")
if ENABLE_OAUTH_SIGNUP.value:
oauth_id_token = request.cookies.get("oauth_id_token")
if oauth_id_token:
oauth_session_id = request.cookies.get("oauth_session_id")
if oauth_session_id:
response.delete_cookie("oauth_session_id")
session = OAuthSessions.get_session_by_id(oauth_session_id)
oauth_server_metadata_url = (
request.app.state.oauth_manager.get_server_metadata_url(session.provider)
if session
else None
) or OPENID_PROVIDER_URL.value
if session and oauth_server_metadata_url:
oauth_id_token = session.token.get("id_token")
try:
async with ClientSession() as session:
async with session.get(OPENID_PROVIDER_URL.value) as resp:
if resp.status == 200:
openid_data = await resp.json()
async with ClientSession(trust_env=True) as session:
async with session.get(oauth_server_metadata_url) as r:
if r.status == 200:
openid_data = await r.json()
logout_url = openid_data.get("end_session_endpoint")
if logout_url:
response.delete_cookie("oauth_id_token")
if logout_url:
return JSONResponse(
status_code=200,
content={
"status": True,
"redirect_url": f"{logout_url}?id_token_hint={oauth_id_token}",
"redirect_url": f"{logout_url}?id_token_hint={oauth_id_token}"
+ (
f"&post_logout_redirect_uri={WEBUI_AUTH_SIGNOUT_REDIRECT_URL}"
if WEBUI_AUTH_SIGNOUT_REDIRECT_URL
else ""
),
},
headers=response.headers,
)
else:
raise HTTPException(
status_code=resp.status,
detail="Failed to fetch OpenID configuration",
)
raise Exception("Failed to fetch OpenID configuration")
except Exception as e:
log.error(f"OpenID signout error: {str(e)}")
raise HTTPException(
status_code=500,
detail="Failed to sign out from the OpenID provider.",
headers=response.headers,
)
if WEBUI_AUTH_SIGNOUT_REDIRECT_URL:
@ -733,6 +817,11 @@ async def add_user(form_data: AddUserForm, user=Depends(get_admin_user)):
raise HTTPException(400, detail=ERROR_MESSAGES.EMAIL_TAKEN)
try:
try:
validate_password(form_data.password)
except Exception as e:
raise HTTPException(400, detail=str(e))
hashed = get_password_hash(form_data.password)
user = Auths.insert_new_auth(
form_data.email.lower(),
@ -804,10 +893,11 @@ async def get_admin_config(request: Request, user=Depends(get_admin_user)):
"SHOW_ADMIN_DETAILS": request.app.state.config.SHOW_ADMIN_DETAILS,
"WEBUI_URL": request.app.state.config.WEBUI_URL,
"ENABLE_SIGNUP": request.app.state.config.ENABLE_SIGNUP,
"ENABLE_API_KEY": request.app.state.config.ENABLE_API_KEY,
"ENABLE_API_KEY_ENDPOINT_RESTRICTIONS": request.app.state.config.ENABLE_API_KEY_ENDPOINT_RESTRICTIONS,
"API_KEY_ALLOWED_ENDPOINTS": request.app.state.config.API_KEY_ALLOWED_ENDPOINTS,
"ENABLE_API_KEYS": request.app.state.config.ENABLE_API_KEYS,
"ENABLE_API_KEYS_ENDPOINT_RESTRICTIONS": request.app.state.config.ENABLE_API_KEYS_ENDPOINT_RESTRICTIONS,
"API_KEYS_ALLOWED_ENDPOINTS": request.app.state.config.API_KEYS_ALLOWED_ENDPOINTS,
"DEFAULT_USER_ROLE": request.app.state.config.DEFAULT_USER_ROLE,
"DEFAULT_GROUP_ID": request.app.state.config.DEFAULT_GROUP_ID,
"JWT_EXPIRES_IN": request.app.state.config.JWT_EXPIRES_IN,
"ENABLE_COMMUNITY_SHARING": request.app.state.config.ENABLE_COMMUNITY_SHARING,
"ENABLE_MESSAGE_RATING": request.app.state.config.ENABLE_MESSAGE_RATING,
@ -824,10 +914,11 @@ class AdminConfig(BaseModel):
SHOW_ADMIN_DETAILS: bool
WEBUI_URL: str
ENABLE_SIGNUP: bool
ENABLE_API_KEY: bool
ENABLE_API_KEY_ENDPOINT_RESTRICTIONS: bool
API_KEY_ALLOWED_ENDPOINTS: str
ENABLE_API_KEYS: bool
ENABLE_API_KEYS_ENDPOINT_RESTRICTIONS: bool
API_KEYS_ALLOWED_ENDPOINTS: str
DEFAULT_USER_ROLE: str
DEFAULT_GROUP_ID: str
JWT_EXPIRES_IN: str
ENABLE_COMMUNITY_SHARING: bool
ENABLE_MESSAGE_RATING: bool
@ -847,12 +938,12 @@ async def update_admin_config(
request.app.state.config.WEBUI_URL = form_data.WEBUI_URL
request.app.state.config.ENABLE_SIGNUP = form_data.ENABLE_SIGNUP
request.app.state.config.ENABLE_API_KEY = form_data.ENABLE_API_KEY
request.app.state.config.ENABLE_API_KEY_ENDPOINT_RESTRICTIONS = (
form_data.ENABLE_API_KEY_ENDPOINT_RESTRICTIONS
request.app.state.config.ENABLE_API_KEYS = form_data.ENABLE_API_KEYS
request.app.state.config.ENABLE_API_KEYS_ENDPOINT_RESTRICTIONS = (
form_data.ENABLE_API_KEYS_ENDPOINT_RESTRICTIONS
)
request.app.state.config.API_KEY_ALLOWED_ENDPOINTS = (
form_data.API_KEY_ALLOWED_ENDPOINTS
request.app.state.config.API_KEYS_ALLOWED_ENDPOINTS = (
form_data.API_KEYS_ALLOWED_ENDPOINTS
)
request.app.state.config.ENABLE_CHANNELS = form_data.ENABLE_CHANNELS
@ -861,6 +952,8 @@ async def update_admin_config(
if form_data.DEFAULT_USER_ROLE in ["pending", "user", "admin"]:
request.app.state.config.DEFAULT_USER_ROLE = form_data.DEFAULT_USER_ROLE
request.app.state.config.DEFAULT_GROUP_ID = form_data.DEFAULT_GROUP_ID
pattern = r"^(-1|0|(-?\d+(\.\d+)?)(ms|s|m|h|d|w))$"
# Check if the input string matches the pattern
@ -887,10 +980,11 @@ async def update_admin_config(
"SHOW_ADMIN_DETAILS": request.app.state.config.SHOW_ADMIN_DETAILS,
"WEBUI_URL": request.app.state.config.WEBUI_URL,
"ENABLE_SIGNUP": request.app.state.config.ENABLE_SIGNUP,
"ENABLE_API_KEY": request.app.state.config.ENABLE_API_KEY,
"ENABLE_API_KEY_ENDPOINT_RESTRICTIONS": request.app.state.config.ENABLE_API_KEY_ENDPOINT_RESTRICTIONS,
"API_KEY_ALLOWED_ENDPOINTS": request.app.state.config.API_KEY_ALLOWED_ENDPOINTS,
"ENABLE_API_KEYS": request.app.state.config.ENABLE_API_KEYS,
"ENABLE_API_KEYS_ENDPOINT_RESTRICTIONS": request.app.state.config.ENABLE_API_KEYS_ENDPOINT_RESTRICTIONS,
"API_KEYS_ALLOWED_ENDPOINTS": request.app.state.config.API_KEYS_ALLOWED_ENDPOINTS,
"DEFAULT_USER_ROLE": request.app.state.config.DEFAULT_USER_ROLE,
"DEFAULT_GROUP_ID": request.app.state.config.DEFAULT_GROUP_ID,
"JWT_EXPIRES_IN": request.app.state.config.JWT_EXPIRES_IN,
"ENABLE_COMMUNITY_SHARING": request.app.state.config.ENABLE_COMMUNITY_SHARING,
"ENABLE_MESSAGE_RATING": request.app.state.config.ENABLE_MESSAGE_RATING,
@ -1014,9 +1108,11 @@ async def update_ldap_config(
# create api key
@router.post("/api_key", response_model=ApiKey)
async def generate_api_key(request: Request, user=Depends(get_current_user)):
if not request.app.state.config.ENABLE_API_KEY:
if not request.app.state.config.ENABLE_API_KEYS or not has_permission(
user.id, "features.api_keys", request.app.state.config.USER_PERMISSIONS
):
raise HTTPException(
status.HTTP_403_FORBIDDEN,
status_code=status.HTTP_403_FORBIDDEN,
detail=ERROR_MESSAGES.API_KEY_CREATION_NOT_ALLOWED,
)

458
backend/open_webui/routers/channels.py
View file

@ -7,10 +7,25 @@ from fastapi import APIRouter, Depends, HTTPException, Request, status, Backgrou
from pydantic import BaseModel
from open_webui.socket.main import sio, get_user_ids_from_room
from open_webui.models.users import Users, UserNameResponse
from open_webui.socket.main import (
sio,
get_user_ids_from_room,
get_active_status_by_user_id,
)
from open_webui.models.users import (
UserListResponse,
UserModelResponse,
Users,
UserNameResponse,
)
from open_webui.models.channels import Channels, ChannelModel, ChannelForm
from open_webui.models.groups import Groups
from open_webui.models.channels import (
Channels,
ChannelModel,
ChannelForm,
ChannelResponse,
)
from open_webui.models.messages import (
Messages,
MessageModel,
@ -24,9 +39,21 @@ from open_webui.constants import ERROR_MESSAGES
from open_webui.env import SRC_LOG_LEVELS
from open_webui.utils.models import (
get_all_models,
get_filtered_models,
)
from open_webui.utils.chat import generate_chat_completion
from open_webui.utils.auth import get_admin_user, get_verified_user
from open_webui.utils.access_control import has_access, get_users_with_access
from open_webui.utils.access_control import (
has_access,
get_users_with_access,
get_permitted_group_and_user_ids,
)
from open_webui.utils.webhook import post_webhook
from open_webui.utils.channels import extract_mentions, replace_mentions
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["MODELS"])
@ -40,10 +67,14 @@ router = APIRouter()
@router.get("/", response_model=list[ChannelModel])
async def get_channels(user=Depends(get_verified_user)):
return Channels.get_channels_by_user_id(user.id)
@router.get("/list", response_model=list[ChannelModel])
async def get_all_channels(user=Depends(get_verified_user)):
if user.role == "admin":
return Channels.get_channels()
else:
return Channels.get_channels_by_user_id(user.id)
return Channels.get_channels_by_user_id(user.id)
############################
@ -68,7 +99,7 @@ async def create_new_channel(form_data: ChannelForm, user=Depends(get_admin_user
############################
@router.get("/{id}", response_model=Optional[ChannelModel])
@router.get("/{id}", response_model=Optional[ChannelResponse])
async def get_channel_by_id(id: str, user=Depends(get_verified_user)):
channel = Channels.get_channel_by_id(id)
if not channel:
@ -83,7 +114,75 @@ async def get_channel_by_id(id: str, user=Depends(get_verified_user)):
status_code=status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.DEFAULT()
)
return ChannelModel(**channel.model_dump())
write_access = has_access(
user.id, type="write", access_control=channel.access_control, strict=False
)
user_count = len(get_users_with_access("read", channel.access_control))
return ChannelResponse(
**{
**channel.model_dump(),
"write_access": write_access or user.role == "admin",
"user_count": user_count,
}
)
PAGE_ITEM_COUNT = 30
@router.get("/{id}/users", response_model=UserListResponse)
async def get_channel_users_by_id(
id: str,
query: Optional[str] = None,
order_by: Optional[str] = None,
direction: Optional[str] = None,
page: Optional[int] = 1,
user=Depends(get_verified_user),
):
channel = Channels.get_channel_by_id(id)
if not channel:
raise HTTPException(
status_code=status.HTTP_404_NOT_FOUND, detail=ERROR_MESSAGES.NOT_FOUND
)
limit = PAGE_ITEM_COUNT
page = max(1, page)
skip = (page - 1) * limit
filter = {
"roles": ["!pending"],
}
if query:
filter["query"] = query
if order_by:
filter["order_by"] = order_by
if direction:
filter["direction"] = direction
permitted_ids = get_permitted_group_and_user_ids("read", channel.access_control)
if permitted_ids:
filter["user_ids"] = permitted_ids.get("user_ids")
filter["group_ids"] = permitted_ids.get("group_ids")
result = Users.get_users(filter=filter, skip=skip, limit=limit)
users = result["users"]
total = result["total"]
return {
"users": [
UserModelResponse(
**user.model_dump(), is_active=get_active_status_by_user_id(user.id)
)
for user in users
],
"total": total,
}
############################
@ -140,7 +239,7 @@ async def delete_channel_by_id(id: str, user=Depends(get_admin_user)):
class MessageUserResponse(MessageResponse):
user: UserNameResponse
pass
@router.get("/{id}/messages", response_model=list[MessageUserResponse])
@ -169,15 +268,17 @@ async def get_channel_messages(
user = Users.get_user_by_id(message.user_id)
users[message.user_id] = user
replies = Messages.get_replies_by_message_id(message.id)
latest_reply_at = replies[0].created_at if replies else None
thread_replies = Messages.get_thread_replies_by_message_id(message.id)
latest_thread_reply_at = (
thread_replies[0].created_at if thread_replies else None
)
messages.append(
MessageUserResponse(
**{
**message.model_dump(),
"reply_count": len(replies),
"latest_reply_at": latest_reply_at,
"reply_count": len(thread_replies),
"latest_reply_at": latest_thread_reply_at,
"reactions": Messages.get_reactions_by_message_id(message.id),
"user": UserNameResponse(**users[message.user_id].model_dump()),
}
@ -196,16 +297,13 @@ async def send_notification(name, webui_url, channel, message, active_user_ids):
users = get_users_with_access("read", channel.access_control)
for user in users:
if user.id in active_user_ids:
continue
else:
if user.id not in active_user_ids:
if user.settings:
webhook_url = user.settings.ui.get("notifications", {}).get(
"webhook_url", None
)
if webhook_url:
post_webhook(
await post_webhook(
name,
webhook_url,
f"#{channel.name} - {webui_url}/channels/{channel.id}\n\n{message.content}",
@ -217,14 +315,185 @@ async def send_notification(name, webui_url, channel, message, active_user_ids):
},
)
return True
@router.post("/{id}/messages/post", response_model=Optional[MessageModel])
async def post_new_message(
request: Request,
id: str,
form_data: MessageForm,
background_tasks: BackgroundTasks,
user=Depends(get_verified_user),
async def model_response_handler(request, channel, message, user):
MODELS = {
model["id"]: model
for model in get_filtered_models(await get_all_models(request, user=user), user)
}
mentions = extract_mentions(message.content)
message_content = replace_mentions(message.content)
model_mentions = {}
# check if the message is a reply to a message sent by a model
if (
message.reply_to_message
and message.reply_to_message.meta
and message.reply_to_message.meta.get("model_id", None)
):
model_id = message.reply_to_message.meta.get("model_id", None)
model_mentions[model_id] = {"id": model_id, "id_type": "M"}
# check if any of the mentions are models
for mention in mentions:
if mention["id_type"] == "M" and mention["id"] not in model_mentions:
model_mentions[mention["id"]] = mention
if not model_mentions:
return False
for mention in model_mentions.values():
model_id = mention["id"]
model = MODELS.get(model_id, None)
if model:
try:
# reverse to get in chronological order
thread_messages = Messages.get_messages_by_parent_id(
channel.id,
message.parent_id if message.parent_id else message.id,
)[::-1]
response_message, channel = await new_message_handler(
request,
channel.id,
MessageForm(
**{
"parent_id": (
message.parent_id if message.parent_id else message.id
),
"content": f"",
"data": {},
"meta": {
"model_id": model_id,
"model_name": model.get("name", model_id),
},
}
),
user,
)
thread_history = []
images = []
message_users = {}
for thread_message in thread_messages:
message_user = None
if thread_message.user_id not in message_users:
message_user = Users.get_user_by_id(thread_message.user_id)
message_users[thread_message.user_id] = message_user
else:
message_user = message_users[thread_message.user_id]
if thread_message.meta and thread_message.meta.get(
"model_id", None
):
# If the message was sent by a model, use the model name
message_model_id = thread_message.meta.get("model_id", None)
message_model = MODELS.get(message_model_id, None)
username = (
message_model.get("name", message_model_id)
if message_model
else message_model_id
)
else:
username = message_user.name if message_user else "Unknown"
thread_history.append(
f"{username}: {replace_mentions(thread_message.content)}"
)
thread_message_files = thread_message.data.get("files", [])
for file in thread_message_files:
if file.get("type", "") == "image":
images.append(file.get("url", ""))
thread_history_string = "\n\n".join(thread_history)
system_message = {
"role": "system",
"content": f"You are {model.get('name', model_id)}, participating in a threaded conversation. Be concise and conversational."
+ (
f"Here's the thread history:\n\n\n{thread_history_string}\n\n\nContinue the conversation naturally as {model.get('name', model_id)}, addressing the most recent message while being aware of the full context."
if thread_history
else ""
),
}
content = f"{user.name if user else 'User'}: {message_content}"
if images:
content = [
{
"type": "text",
"text": content,
},
*[
{
"type": "image_url",
"image_url": {
"url": image,
},
}
for image in images
],
]
form_data = {
"model": model_id,
"messages": [
system_message,
{"role": "user", "content": content},
],
"stream": False,
}
res = await generate_chat_completion(
request,
form_data=form_data,
user=user,
)
if res:
if res.get("choices", []) and len(res["choices"]) > 0:
await update_message_by_id(
channel.id,
response_message.id,
MessageForm(
**{
"content": res["choices"][0]["message"]["content"],
"meta": {
"done": True,
},
}
),
user,
)
elif res.get("error", None):
await update_message_by_id(
channel.id,
response_message.id,
MessageForm(
**{
"content": f"Error: {res['error']}",
"meta": {
"done": True,
},
}
),
user,
)
except Exception as e:
log.info(e)
pass
return True
async def new_message_handler(
request: Request, id: str, form_data: MessageForm, user=Depends(get_verified_user)
):
channel = Channels.get_channel_by_id(id)
if not channel:
@ -233,7 +502,7 @@ async def post_new_message(
)
if user.role != "admin" and not has_access(
user.id, type="read", access_control=channel.access_control
user.id, type="write", access_control=channel.access_control, strict=False
):
raise HTTPException(
status_code=status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.DEFAULT()
@ -241,31 +510,21 @@ async def post_new_message(
try:
message = Messages.insert_new_message(form_data, channel.id, user.id)
if message:
message = Messages.get_message_by_id(message.id)
event_data = {
"channel_id": channel.id,
"message_id": message.id,
"data": {
"type": "message",
"data": MessageUserResponse(
**{
**message.model_dump(),
"reply_count": 0,
"latest_reply_at": None,
"reactions": Messages.get_reactions_by_message_id(
message.id
),
"user": UserNameResponse(**user.model_dump()),
}
).model_dump(),
"data": message.model_dump(),
},
"user": UserNameResponse(**user.model_dump()).model_dump(),
"channel": channel.model_dump(),
}
await sio.emit(
"channel-events",
"events:channel",
event_data,
to=f"channel:{channel.id}",
)
@ -276,33 +535,45 @@ async def post_new_message(
if parent_message:
await sio.emit(
"channel-events",
"events:channel",
{
"channel_id": channel.id,
"message_id": parent_message.id,
"data": {
"type": "message:reply",
"data": MessageUserResponse(
**{
**parent_message.model_dump(),
"user": UserNameResponse(
**Users.get_user_by_id(
parent_message.user_id
).model_dump()
),
}
).model_dump(),
"data": parent_message.model_dump(),
},
"user": UserNameResponse(**user.model_dump()).model_dump(),
"channel": channel.model_dump(),
},
to=f"channel:{channel.id}",
)
return message, channel
else:
raise Exception("Error creating message")
except Exception as e:
log.exception(e)
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST, detail=ERROR_MESSAGES.DEFAULT()
)
active_user_ids = get_user_ids_from_room(f"channel:{channel.id}")
background_tasks.add_task(
send_notification,
@router.post("/{id}/messages/post", response_model=Optional[MessageModel])
async def post_new_message(
request: Request,
id: str,
form_data: MessageForm,
background_tasks: BackgroundTasks,
user=Depends(get_verified_user),
):
try:
message, channel = await new_message_handler(request, id, form_data, user)
active_user_ids = get_user_ids_from_room(f"channel:{channel.id}")
async def background_handler():
await model_response_handler(request, channel, message, user)
await send_notification(
request.app.state.WEBUI_NAME,
request.app.state.config.WEBUI_URL,
channel,
@ -310,7 +581,12 @@ async def post_new_message(
active_user_ids,
)
return MessageModel(**message.model_dump())
background_tasks.add_task(background_handler)
return message
except HTTPException as e:
raise e
except Exception as e:
log.exception(e)
raise HTTPException(
@ -430,13 +706,6 @@ async def update_message_by_id(
status_code=status.HTTP_404_NOT_FOUND, detail=ERROR_MESSAGES.NOT_FOUND
)
if user.role != "admin" and not has_access(
user.id, type="read", access_control=channel.access_control
):
raise HTTPException(
status_code=status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.DEFAULT()
)
message = Messages.get_message_by_id(message_id)
if not message:
raise HTTPException(
@ -448,26 +717,28 @@ async def update_message_by_id(
status_code=status.HTTP_400_BAD_REQUEST, detail=ERROR_MESSAGES.DEFAULT()
)
if (
user.role != "admin"
and message.user_id != user.id
and not has_access(user.id, type="read", access_control=channel.access_control)
):
raise HTTPException(
status_code=status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.DEFAULT()
)
try:
message = Messages.update_message_by_id(message_id, form_data)
message = Messages.get_message_by_id(message_id)
if message:
await sio.emit(
"channel-events",
"events:channel",
{
"channel_id": channel.id,
"message_id": message.id,
"data": {
"type": "message:update",
"data": MessageUserResponse(
**{
**message.model_dump(),
"user": UserNameResponse(
**user.model_dump()
).model_dump(),
}
).model_dump(),
"data": message.model_dump(),
},
"user": UserNameResponse(**user.model_dump()).model_dump(),
"channel": channel.model_dump(),
@ -503,7 +774,7 @@ async def add_reaction_to_message(
)
if user.role != "admin" and not has_access(
user.id, type="read", access_control=channel.access_control
user.id, type="write", access_control=channel.access_control, strict=False
):
raise HTTPException(
status_code=status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.DEFAULT()
@ -525,7 +796,7 @@ async def add_reaction_to_message(
message = Messages.get_message_by_id(message_id)
await sio.emit(
"channel-events",
"events:channel",
{
"channel_id": channel.id,
"message_id": message.id,
@ -533,9 +804,6 @@ async def add_reaction_to_message(
"type": "message:reaction:add",
"data": {
**message.model_dump(),
"user": UserNameResponse(
**Users.get_user_by_id(message.user_id).model_dump()
).model_dump(),
"name": form_data.name,
},
},
@ -569,7 +837,7 @@ async def remove_reaction_by_id_and_user_id_and_name(
)
if user.role != "admin" and not has_access(
user.id, type="read", access_control=channel.access_control
user.id, type="write", access_control=channel.access_control, strict=False
):
raise HTTPException(
status_code=status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.DEFAULT()
@ -594,7 +862,7 @@ async def remove_reaction_by_id_and_user_id_and_name(
message = Messages.get_message_by_id(message_id)
await sio.emit(
"channel-events",
"events:channel",
{
"channel_id": channel.id,
"message_id": message.id,
@ -602,9 +870,6 @@ async def remove_reaction_by_id_and_user_id_and_name(
"type": "message:reaction:remove",
"data": {
**message.model_dump(),
"user": UserNameResponse(
**Users.get_user_by_id(message.user_id).model_dump()
).model_dump(),
"name": form_data.name,
},
},
@ -637,13 +902,6 @@ async def delete_message_by_id(
status_code=status.HTTP_404_NOT_FOUND, detail=ERROR_MESSAGES.NOT_FOUND
)
if user.role != "admin" and not has_access(
user.id, type="read", access_control=channel.access_control
):
raise HTTPException(
status_code=status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.DEFAULT()
)
message = Messages.get_message_by_id(message_id)
if not message:
raise HTTPException(
@ -655,10 +913,21 @@ async def delete_message_by_id(
status_code=status.HTTP_400_BAD_REQUEST, detail=ERROR_MESSAGES.DEFAULT()
)
if (
user.role != "admin"
and message.user_id != user.id
and not has_access(
user.id, type="write", access_control=channel.access_control, strict=False
)
):
raise HTTPException(
status_code=status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.DEFAULT()
)
try:
Messages.delete_message_by_id(message_id)
await sio.emit(
"channel-events",
"events:channel",
{
"channel_id": channel.id,
"message_id": message.id,
@ -681,22 +950,13 @@ async def delete_message_by_id(
if parent_message:
await sio.emit(
"channel-events",
"events:channel",
{
"channel_id": channel.id,
"message_id": parent_message.id,
"data": {
"type": "message:reply",
"data": MessageUserResponse(
**{
**parent_message.model_dump(),
"user": UserNameResponse(
**Users.get_user_by_id(
parent_message.user_id
).model_dump()
),
}
).model_dump(),
"data": parent_message.model_dump(),
},
"user": UserNameResponse(**user.model_dump()).model_dump(),
"channel": channel.model_dump(),

143
backend/open_webui/routers/chats.py
View file

@ -7,6 +7,7 @@ from open_webui.socket.main import get_event_emitter
from open_webui.models.chats import (
ChatForm,
ChatImportForm,
ChatsImportForm,
ChatResponse,
Chats,
ChatTitleIdResponse,
@ -36,16 +37,33 @@ router = APIRouter()
@router.get("/", response_model=list[ChatTitleIdResponse])
@router.get("/list", response_model=list[ChatTitleIdResponse])
async def get_session_user_chat_list(
user=Depends(get_verified_user), page: Optional[int] = None
def get_session_user_chat_list(
user=Depends(get_verified_user),
page: Optional[int] = None,
include_pinned: Optional[bool] = False,
include_folders: Optional[bool] = False,
):
if page is not None:
limit = 60
skip = (page - 1) * limit
try:
if page is not None:
limit = 60
skip = (page - 1) * limit
return Chats.get_chat_title_id_list_by_user_id(user.id, skip=skip, limit=limit)
else:
return Chats.get_chat_title_id_list_by_user_id(user.id)
return Chats.get_chat_title_id_list_by_user_id(
user.id,
include_folders=include_folders,
include_pinned=include_pinned,
skip=skip,
limit=limit,
)
else:
return Chats.get_chat_title_id_list_by_user_id(
user.id, include_folders=include_folders, include_pinned=include_pinned
)
except Exception as e:
log.exception(e)
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST, detail=ERROR_MESSAGES.DEFAULT()
)
############################
@ -125,26 +143,15 @@ async def create_new_chat(form_data: ChatForm, user=Depends(get_verified_user)):
############################
# ImportChat
# ImportChats
############################
@router.post("/import", response_model=Optional[ChatResponse])
async def import_chat(form_data: ChatImportForm, user=Depends(get_verified_user)):
@router.post("/import", response_model=list[ChatResponse])
async def import_chats(form_data: ChatsImportForm, user=Depends(get_verified_user)):
try:
chat = Chats.import_chat(user.id, form_data)
if chat:
tags = chat.meta.get("tags", [])
for tag_id in tags:
tag_id = tag_id.replace(" ", "_").lower()
tag_name = " ".join([word.capitalize() for word in tag_id.split("_")])
if (
tag_id != "none"
and Tags.get_tag_by_name_and_user_id(tag_name, user.id) is None
):
Tags.insert_new_tag(tag_name, user.id)
return ChatResponse(**chat.model_dump())
chats = Chats.import_chats(user.id, form_data.chats)
return chats
except Exception as e:
log.exception(e)
raise HTTPException(
@ -158,7 +165,7 @@ async def import_chat(form_data: ChatImportForm, user=Depends(get_verified_user)
@router.get("/search", response_model=list[ChatTitleIdResponse])
async def search_user_chats(
def search_user_chats(
text: str, page: Optional[int] = None, user=Depends(get_verified_user)
):
if page is None:
@ -206,6 +213,28 @@ async def get_chats_by_folder_id(folder_id: str, user=Depends(get_verified_user)
]
@router.get("/folder/{folder_id}/list")
async def get_chat_list_by_folder_id(
folder_id: str, page: Optional[int] = 1, user=Depends(get_verified_user)
):
try:
limit = 10
skip = (page - 1) * limit
return [
{"title": chat.title, "id": chat.id, "updated_at": chat.updated_at}
for chat in Chats.get_chats_by_folder_id_and_user_id(
folder_id, user.id, skip=skip, limit=limit
)
]
except Exception as e:
log.exception(e)
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST, detail=ERROR_MESSAGES.DEFAULT()
)
############################
# GetPinnedChats
############################
@ -327,6 +356,16 @@ async def archive_all_chats(user=Depends(get_verified_user)):
return Chats.archive_all_chats_by_user_id(user.id)
############################
# UnarchiveAllChats
############################
@router.post("/unarchive/all", response_model=bool)
async def unarchive_all_chats(user=Depends(get_verified_user)):
return Chats.unarchive_all_chats_by_user_id(user.id)
############################
# GetSharedChatById
############################
@ -609,8 +648,28 @@ async def clone_chat_by_id(
"title": form_data.title if form_data.title else f"Clone of {chat.title}",
}
chat = Chats.insert_new_chat(user.id, ChatForm(**{"chat": updated_chat}))
return ChatResponse(**chat.model_dump())
chats = Chats.import_chats(
user.id,
[
ChatImportForm(
**{
"chat": updated_chat,
"meta": chat.meta,
"pinned": chat.pinned,
"folder_id": chat.folder_id,
}
)
],
)
if chats:
chat = chats[0]
return ChatResponse(**chat.model_dump())
else:
raise HTTPException(
status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
detail=ERROR_MESSAGES.DEFAULT(),
)
else:
raise HTTPException(
status_code=status.HTTP_401_UNAUTHORIZED, detail=ERROR_MESSAGES.DEFAULT()
@ -638,8 +697,28 @@ async def clone_shared_chat_by_id(id: str, user=Depends(get_verified_user)):
"title": f"Clone of {chat.title}",
}
chat = Chats.insert_new_chat(user.id, ChatForm(**{"chat": updated_chat}))
return ChatResponse(**chat.model_dump())
chats = Chats.import_chats(
user.id,
[
ChatImportForm(
**{
"chat": updated_chat,
"meta": chat.meta,
"pinned": chat.pinned,
"folder_id": chat.folder_id,
}
)
],
)
if chats:
chat = chats[0]
return ChatResponse(**chat.model_dump())
else:
raise HTTPException(
status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
detail=ERROR_MESSAGES.DEFAULT(),
)
else:
raise HTTPException(
status_code=status.HTTP_401_UNAUTHORIZED, detail=ERROR_MESSAGES.DEFAULT()
@ -684,8 +763,10 @@ async def archive_chat_by_id(id: str, user=Depends(get_verified_user)):
@router.post("/{id}/share", response_model=Optional[ChatResponse])
async def share_chat_by_id(request: Request, id: str, user=Depends(get_verified_user)):
if not has_permission(
user.id, "chat.share", request.app.state.config.USER_PERMISSIONS
if (user.role != "admin") and (
not has_permission(
user.id, "chat.share", request.app.state.config.USER_PERMISSIONS
)
):
raise HTTPException(
status_code=status.HTTP_401_UNAUTHORIZED,

254
backend/open_webui/routers/configs.py
View file

@ -1,5 +1,8 @@
import logging
import copy
from fastapi import APIRouter, Depends, Request, HTTPException
from pydantic import BaseModel, ConfigDict
import aiohttp
from typing import Optional
@ -7,11 +10,30 @@ from open_webui.utils.auth import get_admin_user, get_verified_user
from open_webui.config import get_config, save_config
from open_webui.config import BannerModel
from open_webui.utils.tools import get_tool_server_data, get_tool_servers_data
from open_webui.utils.tools import (
get_tool_server_data,
get_tool_server_url,
set_tool_servers,
)
from open_webui.utils.mcp.client import MCPClient
from open_webui.models.oauth_sessions import OAuthSessions
from open_webui.env import SRC_LOG_LEVELS
from open_webui.utils.oauth import (
get_discovery_urls,
get_oauth_client_info_with_dynamic_client_registration,
encrypt_data,
decrypt_data,
OAuthClientInformationFull,
)
from mcp.shared.auth import OAuthMetadata
router = APIRouter()
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["MAIN"])
############################
# ImportConfig
@ -39,35 +61,79 @@ async def export_config(user=Depends(get_admin_user)):
############################
# Direct Connections Config
# Connections Config
############################
class DirectConnectionsConfigForm(BaseModel):
class ConnectionsConfigForm(BaseModel):
ENABLE_DIRECT_CONNECTIONS: bool
ENABLE_BASE_MODELS_CACHE: bool
@router.get("/direct_connections", response_model=DirectConnectionsConfigForm)
async def get_direct_connections_config(request: Request, user=Depends(get_admin_user)):
@router.get("/connections", response_model=ConnectionsConfigForm)
async def get_connections_config(request: Request, user=Depends(get_admin_user)):
return {
"ENABLE_DIRECT_CONNECTIONS": request.app.state.config.ENABLE_DIRECT_CONNECTIONS,
"ENABLE_BASE_MODELS_CACHE": request.app.state.config.ENABLE_BASE_MODELS_CACHE,
}
@router.post("/direct_connections", response_model=DirectConnectionsConfigForm)
async def set_direct_connections_config(
@router.post("/connections", response_model=ConnectionsConfigForm)
async def set_connections_config(
request: Request,
form_data: DirectConnectionsConfigForm,
form_data: ConnectionsConfigForm,
user=Depends(get_admin_user),
):
request.app.state.config.ENABLE_DIRECT_CONNECTIONS = (
form_data.ENABLE_DIRECT_CONNECTIONS
)
request.app.state.config.ENABLE_BASE_MODELS_CACHE = (
form_data.ENABLE_BASE_MODELS_CACHE
)
return {
"ENABLE_DIRECT_CONNECTIONS": request.app.state.config.ENABLE_DIRECT_CONNECTIONS,
"ENABLE_BASE_MODELS_CACHE": request.app.state.config.ENABLE_BASE_MODELS_CACHE,
}
class OAuthClientRegistrationForm(BaseModel):
url: str
client_id: str
client_name: Optional[str] = None
@router.post("/oauth/clients/register")
async def register_oauth_client(
request: Request,
form_data: OAuthClientRegistrationForm,
type: Optional[str] = None,
user=Depends(get_admin_user),
):
try:
oauth_client_id = form_data.client_id
if type:
oauth_client_id = f"{type}:{form_data.client_id}"
oauth_client_info = (
await get_oauth_client_info_with_dynamic_client_registration(
request, oauth_client_id, form_data.url
)
)
return {
"status": True,
"oauth_client_info": encrypt_data(
oauth_client_info.model_dump(mode="json")
),
}
except Exception as e:
log.debug(f"Failed to register OAuth client: {e}")
raise HTTPException(
status_code=400,
detail=f"Failed to register OAuth client",
)
############################
# ToolServers Config
############################
@ -76,7 +142,9 @@ async def set_direct_connections_config(
class ToolServerConnection(BaseModel):
url: str
path: str
type: Optional[str] = "openapi" # openapi, mcp
auth_type: Optional[str]
headers: Optional[dict | str] = None
key: Optional[str]
config: Optional[dict]
@ -100,13 +168,47 @@ async def set_tool_servers_config(
form_data: ToolServersConfigForm,
user=Depends(get_admin_user),
):
for connection in request.app.state.config.TOOL_SERVER_CONNECTIONS:
server_type = connection.get("type", "openapi")
auth_type = connection.get("auth_type", "none")
if auth_type == "oauth_2.1":
# Remove existing OAuth clients for tool servers
server_id = connection.get("info", {}).get("id")
client_key = f"{server_type}:{server_id}"
try:
request.app.state.oauth_client_manager.remove_client(client_key)
except:
pass
# Set new tool server connections
request.app.state.config.TOOL_SERVER_CONNECTIONS = [
connection.model_dump() for connection in form_data.TOOL_SERVER_CONNECTIONS
]
request.app.state.TOOL_SERVERS = await get_tool_servers_data(
request.app.state.config.TOOL_SERVER_CONNECTIONS
)
await set_tool_servers(request)
for connection in request.app.state.config.TOOL_SERVER_CONNECTIONS:
server_type = connection.get("type", "openapi")
if server_type == "mcp":
server_id = connection.get("info", {}).get("id")
auth_type = connection.get("auth_type", "none")
if auth_type == "oauth_2.1" and server_id:
try:
oauth_client_info = connection.get("info", {}).get(
"oauth_client_info", ""
)
oauth_client_info = decrypt_data(oauth_client_info)
request.app.state.oauth_client_manager.add_client(
f"{server_type}:{server_id}",
OAuthClientInformationFull(**oauth_client_info),
)
except Exception as e:
log.debug(f"Failed to add OAuth client for MCP tool server: {e}")
continue
return {
"TOOL_SERVER_CONNECTIONS": request.app.state.config.TOOL_SERVER_CONNECTIONS,
@ -121,19 +223,129 @@ async def verify_tool_servers_config(
Verify the connection to the tool server.
"""
try:
if form_data.type == "mcp":
if form_data.auth_type == "oauth_2.1":
discovery_urls = get_discovery_urls(form_data.url)
for discovery_url in discovery_urls:
log.debug(
f"Trying to fetch OAuth 2.1 discovery document from {discovery_url}"
)
async with aiohttp.ClientSession(trust_env=True) as session:
async with session.get(
discovery_url
) as oauth_server_metadata_response:
if oauth_server_metadata_response.status == 200:
try:
oauth_server_metadata = (
OAuthMetadata.model_validate(
await oauth_server_metadata_response.json()
)
)
return {
"status": True,
"oauth_server_metadata": oauth_server_metadata.model_dump(
mode="json"
),
}
except Exception as e:
log.info(
f"Failed to parse OAuth 2.1 discovery document: {e}"
)
raise HTTPException(
status_code=400,
detail=f"Failed to parse OAuth 2.1 discovery document from {discovery_url}",
)
token = None
if form_data.auth_type == "bearer":
token = form_data.key
elif form_data.auth_type == "session":
token = request.state.token.credentials
raise HTTPException(
status_code=400,
detail=f"Failed to fetch OAuth 2.1 discovery document from {discovery_urls}",
)
else:
try:
client = MCPClient()
headers = None
url = f"{form_data.url}/{form_data.path}"
return await get_tool_server_data(token, url)
token = None
if form_data.auth_type == "bearer":
token = form_data.key
elif form_data.auth_type == "session":
token = request.state.token.credentials
elif form_data.auth_type == "system_oauth":
oauth_token = None
try:
if request.cookies.get("oauth_session_id", None):
oauth_token = await request.app.state.oauth_manager.get_oauth_token(
user.id,
request.cookies.get("oauth_session_id", None),
)
if oauth_token:
token = oauth_token.get("access_token", "")
except Exception as e:
pass
if token:
headers = {"Authorization": f"Bearer {token}"}
if form_data.headers and isinstance(form_data.headers, dict):
if headers is None:
headers = {}
headers.update(form_data.headers)
await client.connect(form_data.url, headers=headers)
specs = await client.list_tool_specs()
return {
"status": True,
"specs": specs,
}
except Exception as e:
log.debug(f"Failed to create MCP client: {e}")
raise HTTPException(
status_code=400,
detail=f"Failed to create MCP client",
)
finally:
if client:
await client.disconnect()
else: # openapi
token = None
headers = None
if form_data.auth_type == "bearer":
token = form_data.key
elif form_data.auth_type == "session":
token = request.state.token.credentials
elif form_data.auth_type == "system_oauth":
try:
if request.cookies.get("oauth_session_id", None):
oauth_token = (
await request.app.state.oauth_manager.get_oauth_token(
user.id,
request.cookies.get("oauth_session_id", None),
)
)
if oauth_token:
token = oauth_token.get("access_token", "")
except Exception as e:
pass
if token:
headers = {"Authorization": f"Bearer {token}"}
if form_data.headers and isinstance(form_data.headers, dict):
if headers is None:
headers = {}
headers.update(form_data.headers)
url = get_tool_server_url(form_data.url, form_data.path)
return await get_tool_server_data(url, headers=headers)
except HTTPException as e:
raise e
except Exception as e:
log.debug(f"Failed to connect to the tool server: {e}")
raise HTTPException(
status_code=400,
detail=f"Failed to connect to the tool server: {str(e)}",
detail=f"Failed to connect to the tool server",
)
@ -251,6 +463,7 @@ async def set_code_execution_config(
############################
class ModelsConfigForm(BaseModel):
DEFAULT_MODELS: Optional[str]
DEFAULT_PINNED_MODELS: Optional[str]
MODEL_ORDER_LIST: Optional[list[str]]
@ -258,6 +471,7 @@ class ModelsConfigForm(BaseModel):
async def get_models_config(request: Request, user=Depends(get_admin_user)):
return {
"DEFAULT_MODELS": request.app.state.config.DEFAULT_MODELS,
"DEFAULT_PINNED_MODELS": request.app.state.config.DEFAULT_PINNED_MODELS,
"MODEL_ORDER_LIST": request.app.state.config.MODEL_ORDER_LIST,
}
@ -267,9 +481,11 @@ async def set_models_config(
request: Request, form_data: ModelsConfigForm, user=Depends(get_admin_user)
):
request.app.state.config.DEFAULT_MODELS = form_data.DEFAULT_MODELS
request.app.state.config.DEFAULT_PINNED_MODELS = form_data.DEFAULT_PINNED_MODELS
request.app.state.config.MODEL_ORDER_LIST = form_data.MODEL_ORDER_LIST
return {
"DEFAULT_MODELS": request.app.state.config.DEFAULT_MODELS,
"DEFAULT_PINNED_MODELS": request.app.state.config.DEFAULT_PINNED_MODELS,
"MODEL_ORDER_LIST": request.app.state.config.MODEL_ORDER_LIST,
}

70
backend/open_webui/routers/evaluations.py
View file

@ -7,6 +7,8 @@ from open_webui.models.feedbacks import (
FeedbackModel,
FeedbackResponse,
FeedbackForm,
FeedbackUserResponse,
FeedbackListResponse,
Feedbacks,
)
@ -56,35 +58,10 @@ async def update_config(
}
class UserResponse(BaseModel):
id: str
name: str
email: str
role: str = "pending"
last_active_at: int # timestamp in epoch
updated_at: int # timestamp in epoch
created_at: int # timestamp in epoch
class FeedbackUserResponse(FeedbackResponse):
user: Optional[UserResponse] = None
@router.get("/feedbacks/all", response_model=list[FeedbackUserResponse])
@router.get("/feedbacks/all", response_model=list[FeedbackResponse])
async def get_all_feedbacks(user=Depends(get_admin_user)):
feedbacks = Feedbacks.get_all_feedbacks()
feedback_list = []
for feedback in feedbacks:
user = Users.get_user_by_id(feedback.user_id)
feedback_list.append(
FeedbackUserResponse(
**feedback.model_dump(),
user=UserResponse(**user.model_dump()) if user else None,
)
)
return feedback_list
return feedbacks
@router.delete("/feedbacks/all")
@ -111,6 +88,31 @@ async def delete_feedbacks(user=Depends(get_verified_user)):
return success
PAGE_ITEM_COUNT = 30
@router.get("/feedbacks/list", response_model=FeedbackListResponse)
async def get_feedbacks(
order_by: Optional[str] = None,
direction: Optional[str] = None,
page: Optional[int] = 1,
user=Depends(get_admin_user),
):
limit = PAGE_ITEM_COUNT
page = max(1, page)
skip = (page - 1) * limit
filter = {}
if order_by:
filter["order_by"] = order_by
if direction:
filter["direction"] = direction
result = Feedbacks.get_feedback_items(filter=filter, skip=skip, limit=limit)
return result
@router.post("/feedback", response_model=FeedbackModel)
async def create_feedback(
request: Request,
@ -129,7 +131,10 @@ async def create_feedback(
@router.get("/feedback/{id}", response_model=FeedbackModel)
async def get_feedback_by_id(id: str, user=Depends(get_verified_user)):
feedback = Feedbacks.get_feedback_by_id_and_user_id(id=id, user_id=user.id)
if user.role == "admin":
feedback = Feedbacks.get_feedback_by_id(id=id)
else:
feedback = Feedbacks.get_feedback_by_id_and_user_id(id=id, user_id=user.id)
if not feedback:
raise HTTPException(
@ -143,9 +148,12 @@ async def get_feedback_by_id(id: str, user=Depends(get_verified_user)):
async def update_feedback_by_id(
id: str, form_data: FeedbackForm, user=Depends(get_verified_user)
):
feedback = Feedbacks.update_feedback_by_id_and_user_id(
id=id, user_id=user.id, form_data=form_data
)
if user.role == "admin":
feedback = Feedbacks.update_feedback_by_id(id=id, form_data=form_data)
else:
feedback = Feedbacks.update_feedback_by_id_and_user_id(
id=id, user_id=user.id, form_data=form_data
)
if not feedback:
raise HTTPException(

236
backend/open_webui/routers/files.py
View file

@ -6,8 +6,10 @@ from fnmatch import fnmatch
from pathlib import Path
from typing import Optional
from urllib.parse import quote
import asyncio
from fastapi import (
BackgroundTasks,
APIRouter,
Depends,
File,
@ -18,9 +20,11 @@ from fastapi import (
status,
Query,
)
from fastapi.responses import FileResponse, StreamingResponse
from open_webui.constants import ERROR_MESSAGES
from open_webui.env import SRC_LOG_LEVELS
from open_webui.retrieval.vector.factory import VECTOR_DB_CLIENT
from open_webui.models.users import Users
from open_webui.models.files import (
@ -41,7 +45,6 @@ from pydantic import BaseModel
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["MODELS"])
router = APIRouter()
@ -82,14 +85,85 @@ def has_access_to_file(
############################
def process_uploaded_file(request, file, file_path, file_item, file_metadata, user):
try:
if file.content_type:
stt_supported_content_types = getattr(
request.app.state.config, "STT_SUPPORTED_CONTENT_TYPES", []
)
if any(
fnmatch(file.content_type, content_type)
for content_type in (
stt_supported_content_types
if stt_supported_content_types
and any(t.strip() for t in stt_supported_content_types)
else ["audio/*", "video/webm"]
)
):
file_path = Storage.get_file(file_path)
result = transcribe(request, file_path, file_metadata, user)
process_file(
request,
ProcessFileForm(
file_id=file_item.id, content=result.get("text", "")
),
user=user,
)
elif (not file.content_type.startswith(("image/", "video/"))) or (
request.app.state.config.CONTENT_EXTRACTION_ENGINE == "external"
):
process_file(request, ProcessFileForm(file_id=file_item.id), user=user)
else:
raise Exception(
f"File type {file.content_type} is not supported for processing"
)
else:
log.info(
f"File type {file.content_type} is not provided, but trying to process anyway"
)
process_file(request, ProcessFileForm(file_id=file_item.id), user=user)
except Exception as e:
log.error(f"Error processing file: {file_item.id}")
Files.update_file_data_by_id(
file_item.id,
{
"status": "failed",
"error": str(e.detail) if hasattr(e, "detail") else str(e),
},
)
@router.post("/", response_model=FileModelResponse)
def upload_file(
request: Request,
background_tasks: BackgroundTasks,
file: UploadFile = File(...),
metadata: Optional[dict | str] = Form(None),
process: bool = Query(True),
process_in_background: bool = Query(True),
user=Depends(get_verified_user),
):
return upload_file_handler(
request,
file=file,
metadata=metadata,
process=process,
process_in_background=process_in_background,
user=user,
background_tasks=background_tasks,
)
def upload_file_handler(
request: Request,
file: UploadFile = File(...),
metadata: Optional[dict | str] = Form(None),
process: bool = Query(True),
internal: bool = False,
process_in_background: bool = Query(True),
user=Depends(get_verified_user),
background_tasks: Optional[BackgroundTasks] = None,
):
log.info(f"file.content_type: {file.content_type}")
@ -111,7 +185,7 @@ def upload_file(
# Remove the leading dot from the file extension
file_extension = file_extension[1:] if file_extension else ""
if (not internal) and request.app.state.config.ALLOWED_FILE_EXTENSIONS:
if process and request.app.state.config.ALLOWED_FILE_EXTENSIONS:
request.app.state.config.ALLOWED_FILE_EXTENSIONS = [
ext for ext in request.app.state.config.ALLOWED_FILE_EXTENSIONS if ext
]
@ -128,13 +202,16 @@ def upload_file(
id = str(uuid.uuid4())
name = filename
filename = f"{id}_{filename}"
tags = {
"OpenWebUI-User-Email": user.email,
"OpenWebUI-User-Id": user.id,
"OpenWebUI-User-Name": user.name,
"OpenWebUI-File-Id": id,
}
contents, file_path = Storage.upload_file(file.file, filename, tags)
contents, file_path = Storage.upload_file(
file.file,
filename,
{
"OpenWebUI-User-Email": user.email,
"OpenWebUI-User-Id": user.id,
"OpenWebUI-User-Name": user.name,
"OpenWebUI-File-Id": id,
},
)
file_item = Files.insert_new_file(
user.id,
@ -143,6 +220,9 @@ def upload_file(
"id": id,
"filename": name,
"path": file_path,
"data": {
**({"status": "pending"} if process else {}),
},
"meta": {
"name": name,
"content_type": file.content_type,
@ -152,57 +232,37 @@ def upload_file(
}
),
)
if process:
try:
if file.content_type:
stt_supported_content_types = (
request.app.state.config.STT_SUPPORTED_CONTENT_TYPES
or [
"audio/*",
"video/webm",
]
)
if any(
fnmatch(file.content_type, content_type)
for content_type in stt_supported_content_types
):
file_path = Storage.get_file(file_path)
result = transcribe(request, file_path, file_metadata)
process_file(
request,
ProcessFileForm(file_id=id, content=result.get("text", "")),
user=user,
)
elif (not file.content_type.startswith(("image/", "video/"))) or (
request.app.state.config.CONTENT_EXTRACTION_ENGINE == "external"
):
process_file(request, ProcessFileForm(file_id=id), user=user)
else:
log.info(
f"File type {file.content_type} is not provided, but trying to process anyway"
)
process_file(request, ProcessFileForm(file_id=id), user=user)
file_item = Files.get_file_by_id(id=id)
except Exception as e:
log.exception(e)
log.error(f"Error processing file: {file_item.id}")
file_item = FileModelResponse(
**{
**file_item.model_dump(),
"error": str(e.detail) if hasattr(e, "detail") else str(e),
}
if background_tasks and process_in_background:
background_tasks.add_task(
process_uploaded_file,
request,
file,
file_path,
file_item,
file_metadata,
user,
)
if file_item:
return file_item
return {"status": True, **file_item.model_dump()}
else:
process_uploaded_file(
request,
file,
file_path,
file_item,
file_metadata,
user,
)
return {"status": True, **file_item.model_dump()}
else:
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.DEFAULT("Error uploading file"),
)
if file_item:
return file_item
else:
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.DEFAULT("Error uploading file"),
)
except Exception as e:
log.exception(e)
@ -285,6 +345,7 @@ async def delete_all_files(user=Depends(get_admin_user)):
if result:
try:
Storage.delete_all_files()
VECTOR_DB_CLIENT.reset()
except Exception as e:
log.exception(e)
log.error("Error deleting files")
@ -328,6 +389,63 @@ async def get_file_by_id(id: str, user=Depends(get_verified_user)):
)
@router.get("/{id}/process/status")
async def get_file_process_status(
id: str, stream: bool = Query(False), user=Depends(get_verified_user)
):
file = Files.get_file_by_id(id)
if not file:
raise HTTPException(
status_code=status.HTTP_404_NOT_FOUND,
detail=ERROR_MESSAGES.NOT_FOUND,
)
if (
file.user_id == user.id
or user.role == "admin"
or has_access_to_file(id, "read", user)
):
if stream:
MAX_FILE_PROCESSING_DURATION = 3600 * 2
async def event_stream(file_item):
if file_item:
for _ in range(MAX_FILE_PROCESSING_DURATION):
file_item = Files.get_file_by_id(file_item.id)
if file_item:
data = file_item.model_dump().get("data", {})
status = data.get("status")
if status:
event = {"status": status}
if status == "failed":
event["error"] = data.get("error")
yield f"data: {json.dumps(event)}\n\n"
if status in ("completed", "failed"):
break
else:
# Legacy
break
await asyncio.sleep(0.5)
else:
yield f"data: {json.dumps({'status': 'not_found'})}\n\n"
return StreamingResponse(
event_stream(file),
media_type="text/event-stream",
)
else:
return {"status": file.data.get("status", "pending")}
else:
raise HTTPException(
status_code=status.HTTP_404_NOT_FOUND,
detail=ERROR_MESSAGES.NOT_FOUND,
)
############################
# Get File Data Content By Id
############################
@ -602,12 +720,12 @@ async def delete_file_by_id(id: str, user=Depends(get_verified_user)):
or user.role == "admin"
or has_access_to_file(id, "write", user)
):
# We should add Chroma cleanup here
result = Files.delete_file_by_id(id)
if result:
try:
Storage.delete_file(file.path)
VECTOR_DB_CLIENT.delete(collection_name=f"file-{id}")
except Exception as e:
log.exception(e)
log.error("Error deleting files")

148
backend/open_webui/routers/folders.py
View file

@ -10,10 +10,15 @@ import mimetypes
from open_webui.models.folders import (
FolderForm,
FolderUpdateForm,
FolderModel,
FolderNameIdResponse,
Folders,
)
from open_webui.models.chats import Chats
from open_webui.models.files import Files
from open_webui.models.knowledge import Knowledges
from open_webui.config import UPLOAD_DIR
from open_webui.env import SRC_LOG_LEVELS
@ -40,24 +45,46 @@ router = APIRouter()
############################
@router.get("/", response_model=list[FolderModel])
@router.get("/", response_model=list[FolderNameIdResponse])
async def get_folders(user=Depends(get_verified_user)):
folders = Folders.get_folders_by_user_id(user.id)
return [
{
**folder.model_dump(),
"items": {
"chats": [
{"title": chat.title, "id": chat.id}
for chat in Chats.get_chats_by_folder_id_and_user_id(
folder.id, user.id
)
]
},
}
for folder in folders
]
# Verify folder data integrity
folder_list = []
for folder in folders:
if folder.parent_id and not Folders.get_folder_by_id_and_user_id(
folder.parent_id, user.id
):
folder = Folders.update_folder_parent_id_by_id_and_user_id(
folder.id, user.id, None
)
if folder.data:
if "files" in folder.data:
valid_files = []
for file in folder.data["files"]:
if file.get("type") == "file":
if Files.check_access_by_user_id(
file.get("id"), user.id, "read"
):
valid_files.append(file)
elif file.get("type") == "collection":
if Knowledges.check_access_by_user_id(
file.get("id"), user.id, "read"
):
valid_files.append(file)
else:
valid_files.append(file)
folder.data["files"] = valid_files
Folders.update_folder_by_id_and_user_id(
folder.id, user.id, FolderUpdateForm(data=folder.data)
)
folder_list.append(FolderNameIdResponse(**folder.model_dump()))
return folder_list
############################
@ -78,7 +105,7 @@ def create_folder(form_data: FolderForm, user=Depends(get_verified_user)):
)
try:
folder = Folders.insert_new_folder(user.id, form_data.name)
folder = Folders.insert_new_folder(user.id, form_data)
return folder
except Exception as e:
log.exception(e)
@ -113,24 +140,24 @@ async def get_folder_by_id(id: str, user=Depends(get_verified_user)):
@router.post("/{id}/update")
async def update_folder_name_by_id(
id: str, form_data: FolderForm, user=Depends(get_verified_user)
id: str, form_data: FolderUpdateForm, user=Depends(get_verified_user)
):
folder = Folders.get_folder_by_id_and_user_id(id, user.id)
if folder:
existing_folder = Folders.get_folder_by_parent_id_and_user_id_and_name(
folder.parent_id, user.id, form_data.name
)
if existing_folder:
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.DEFAULT("Folder already exists"),
if form_data.name is not None:
# Check if folder with same name exists
existing_folder = Folders.get_folder_by_parent_id_and_user_id_and_name(
folder.parent_id, user.id, form_data.name
)
if existing_folder and existing_folder.id != id:
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.DEFAULT("Folder already exists"),
)
try:
folder = Folders.update_folder_name_by_id_and_user_id(
id, user.id, form_data.name
)
folder = Folders.update_folder_by_id_and_user_id(id, user.id, form_data)
return folder
except Exception as e:
log.exception(e)
@ -231,33 +258,52 @@ async def update_folder_is_expanded_by_id(
@router.delete("/{id}")
async def delete_folder_by_id(
request: Request, id: str, user=Depends(get_verified_user)
request: Request,
id: str,
delete_contents: Optional[bool] = True,
user=Depends(get_verified_user),
):
chat_delete_permission = has_permission(
user.id, "chat.delete", request.app.state.config.USER_PERMISSIONS
)
if user.role != "admin" and not chat_delete_permission:
raise HTTPException(
status_code=status.HTTP_403_FORBIDDEN,
detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
if Chats.count_chats_by_folder_id_and_user_id(id, user.id):
chat_delete_permission = has_permission(
user.id, "chat.delete", request.app.state.config.USER_PERMISSIONS
)
folder = Folders.get_folder_by_id_and_user_id(id, user.id)
if folder:
try:
result = Folders.delete_folder_by_id_and_user_id(id, user.id)
if result:
return result
else:
raise Exception("Error deleting folder")
except Exception as e:
log.exception(e)
log.error(f"Error deleting folder: {id}")
if user.role != "admin" and not chat_delete_permission:
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.DEFAULT("Error deleting folder"),
status_code=status.HTTP_403_FORBIDDEN,
detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
)
folders = []
folders.append(Folders.get_folder_by_id_and_user_id(id, user.id))
while folders:
folder = folders.pop()
if folder:
try:
folder_ids = Folders.delete_folder_by_id_and_user_id(id, user.id)
for folder_id in folder_ids:
if delete_contents:
Chats.delete_chats_by_user_id_and_folder_id(user.id, folder_id)
else:
Chats.move_chats_by_user_id_and_folder_id(
user.id, folder_id, None
)
return True
except Exception as e:
log.exception(e)
log.error(f"Error deleting folder: {id}")
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.DEFAULT("Error deleting folder"),
)
finally:
# Get all subfolders
subfolders = Folders.get_folders_by_parent_id_and_user_id(
folder.id, user.id
)
folders.extend(subfolders)
else:
raise HTTPException(
status_code=status.HTTP_404_NOT_FOUND,

66
backend/open_webui/routers/functions.py
View file

@ -10,6 +10,8 @@ from open_webui.models.functions import (
FunctionForm,
FunctionModel,
FunctionResponse,
FunctionUserResponse,
FunctionWithValvesModel,
Functions,
)
from open_webui.utils.plugin import (
@ -41,14 +43,19 @@ async def get_functions(user=Depends(get_verified_user)):
return Functions.get_functions()
@router.get("/list", response_model=list[FunctionUserResponse])
async def get_function_list(user=Depends(get_admin_user)):
return Functions.get_function_list()
############################
# ExportFunctions
############################
@router.get("/export", response_model=list[FunctionModel])
async def get_functions(user=Depends(get_admin_user)):
return Functions.get_functions()
@router.get("/export", response_model=list[FunctionModel | FunctionWithValvesModel])
async def get_functions(include_valves: bool = False, user=Depends(get_admin_user)):
return Functions.get_functions(include_valves=include_valves)
############################
@ -105,7 +112,7 @@ async def load_function_from_url(
)
try:
async with aiohttp.ClientSession() as session:
async with aiohttp.ClientSession(trust_env=True) as session:
async with session.get(
url, headers={"Content-Type": "application/json"}
) as resp:
@ -131,15 +138,41 @@ async def load_function_from_url(
############################
class SyncFunctionsForm(FunctionForm):
functions: list[FunctionModel] = []
class SyncFunctionsForm(BaseModel):
functions: list[FunctionWithValvesModel] = []
@router.post("/sync", response_model=Optional[FunctionModel])
@router.post("/sync", response_model=list[FunctionWithValvesModel])
async def sync_functions(
request: Request, form_data: SyncFunctionsForm, user=Depends(get_admin_user)
):
return Functions.sync_functions(user.id, form_data.functions)
try:
for function in form_data.functions:
function.content = replace_imports(function.content)
function_module, function_type, frontmatter = load_function_module_by_id(
function.id,
content=function.content,
)
if hasattr(function_module, "Valves") and function.valves:
Valves = function_module.Valves
try:
Valves(
**{k: v for k, v in function.valves.items() if v is not None}
)
except Exception as e:
log.exception(
f"Error validating valves for function {function.id}: {e}"
)
raise e
return Functions.sync_functions(user.id, form_data.functions)
except Exception as e:
log.exception(f"Failed to load a function: {e}")
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.DEFAULT(e),
)
############################
@ -177,6 +210,9 @@ async def create_new_function(
function_cache_dir = CACHE_DIR / "functions" / form_data.id
function_cache_dir.mkdir(parents=True, exist_ok=True)
if function_type == "filter" and getattr(function_module, "toggle", None):
Functions.update_function_metadata_by_id(id, {"toggle": True})
if function:
return function
else:
@ -293,6 +329,9 @@ async def update_function_by_id(
function = Functions.update_function_by_id(id, updated)
if function_type == "filter" and getattr(function_module, "toggle", None):
Functions.update_function_metadata_by_id(id, {"toggle": True})
if function:
return function
else:
@ -398,8 +437,10 @@ async def update_function_valves_by_id(
try:
form_data = {k: v for k, v in form_data.items() if v is not None}
valves = Valves(**form_data)
Functions.update_function_valves_by_id(id, valves.model_dump())
return valves.model_dump()
valves_dict = valves.model_dump(exclude_unset=True)
Functions.update_function_valves_by_id(id, valves_dict)
return valves_dict
except Exception as e:
log.exception(f"Error updating function values by id {id}: {e}")
raise HTTPException(
@ -481,10 +522,11 @@ async def update_function_user_valves_by_id(
try:
form_data = {k: v for k, v in form_data.items() if v is not None}
user_valves = UserValves(**form_data)
user_valves_dict = user_valves.model_dump(exclude_unset=True)
Functions.update_user_valves_by_id_and_user_id(
id, user.id, user_valves.model_dump()
id, user.id, user_valves_dict
)
return user_valves.model_dump()
return user_valves_dict
except Exception as e:
log.exception(f"Error updating function user valves by id {id}: {e}")
raise HTTPException(

102
backend/open_webui/routers/groups.py
View file

@ -9,6 +9,7 @@ from open_webui.models.groups import (
GroupForm,
GroupUpdateForm,
GroupResponse,
UserIdsForm,
)
from open_webui.config import CACHE_DIR
@ -30,11 +31,32 @@ router = APIRouter()
@router.get("/", response_model=list[GroupResponse])
async def get_groups(user=Depends(get_verified_user)):
async def get_groups(share: Optional[bool] = None, user=Depends(get_verified_user)):
if user.role == "admin":
return Groups.get_groups()
groups = Groups.get_groups()
else:
return Groups.get_groups_by_member_id(user.id)
groups = Groups.get_groups_by_member_id(user.id)
group_list = []
for group in groups:
if share is not None:
# Check if the group has data and a config with share key
if (
group.data
and "share" in group.data.get("config", {})
and group.data["config"]["share"] != share
):
continue
group_list.append(
GroupResponse(
**group.model_dump(),
member_count=Groups.get_group_member_count_by_id(group.id),
)
)
return group_list
############################
@ -47,7 +69,10 @@ async def create_new_group(form_data: GroupForm, user=Depends(get_admin_user)):
try:
group = Groups.insert_new_group(user.id, form_data)
if group:
return group
return GroupResponse(
**group.model_dump(),
member_count=Groups.get_group_member_count_by_id(group.id),
)
else:
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
@ -70,7 +95,10 @@ async def create_new_group(form_data: GroupForm, user=Depends(get_admin_user)):
async def get_group_by_id(id: str, user=Depends(get_admin_user)):
group = Groups.get_group_by_id(id)
if group:
return group
return GroupResponse(
**group.model_dump(),
member_count=Groups.get_group_member_count_by_id(group.id),
)
else:
raise HTTPException(
status_code=status.HTTP_401_UNAUTHORIZED,
@ -88,12 +116,12 @@ async def update_group_by_id(
id: str, form_data: GroupUpdateForm, user=Depends(get_admin_user)
):
try:
if form_data.user_ids:
form_data.user_ids = Users.get_valid_user_ids(form_data.user_ids)
group = Groups.update_group_by_id(id, form_data)
if group:
return group
return GroupResponse(
**group.model_dump(),
member_count=Groups.get_group_member_count_by_id(group.id),
)
else:
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
@ -107,6 +135,62 @@ async def update_group_by_id(
)
############################
# AddUserToGroupByUserIdAndGroupId
############################
@router.post("/id/{id}/users/add", response_model=Optional[GroupResponse])
async def add_user_to_group(
id: str, form_data: UserIdsForm, user=Depends(get_admin_user)
):
try:
if form_data.user_ids:
form_data.user_ids = Users.get_valid_user_ids(form_data.user_ids)
group = Groups.add_users_to_group(id, form_data.user_ids)
if group:
return GroupResponse(
**group.model_dump(),
member_count=Groups.get_group_member_count_by_id(group.id),
)
else:
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.DEFAULT("Error adding users to group"),
)
except Exception as e:
log.exception(f"Error adding users to group {id}: {e}")
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.DEFAULT(e),
)
@router.post("/id/{id}/users/remove", response_model=Optional[GroupResponse])
async def remove_users_from_group(
id: str, form_data: UserIdsForm, user=Depends(get_admin_user)
):
try:
group = Groups.remove_users_from_group(id, form_data.user_ids)
if group:
return GroupResponse(
**group.model_dump(),
member_count=Groups.get_group_member_count_by_id(group.id),
)
else:
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.DEFAULT("Error removing users from group"),
)
except Exception as e:
log.exception(f"Error removing users from group {id}: {e}")
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.DEFAULT(e),
)
############################
# DeleteGroupById
############################

975
backend/open_webui/routers/images.py
View file

File diff suppressed because it is too large Load diff

62
backend/open_webui/routers/knowledge.py
View file

@ -1,6 +1,7 @@
from typing import List, Optional
from pydantic import BaseModel
from fastapi import APIRouter, Depends, HTTPException, status, Request
from fastapi import APIRouter, Depends, HTTPException, status, Request, Query
from fastapi.concurrency import run_in_threadpool
import logging
from open_webui.models.knowledge import (
@ -25,6 +26,7 @@ from open_webui.utils.access_control import has_access, has_permission
from open_webui.env import SRC_LOG_LEVELS
from open_webui.config import BYPASS_ADMIN_ACCESS_CONTROL
from open_webui.models.models import Models, ModelForm
@ -42,7 +44,7 @@ router = APIRouter()
async def get_knowledge(user=Depends(get_verified_user)):
knowledge_bases = []
if user.role == "admin":
if user.role == "admin" and BYPASS_ADMIN_ACCESS_CONTROL:
knowledge_bases = Knowledges.get_knowledge_bases()
else:
knowledge_bases = Knowledges.get_knowledge_bases_by_user_id(user.id, "read")
@ -90,7 +92,7 @@ async def get_knowledge(user=Depends(get_verified_user)):
async def get_knowledge_list(user=Depends(get_verified_user)):
knowledge_bases = []
if user.role == "admin":
if user.role == "admin" and BYPASS_ADMIN_ACCESS_CONTROL:
knowledge_bases = Knowledges.get_knowledge_bases()
else:
knowledge_bases = Knowledges.get_knowledge_bases_by_user_id(user.id, "write")
@ -150,6 +152,18 @@ async def create_new_knowledge(
detail=ERROR_MESSAGES.UNAUTHORIZED,
)
# Check if user can share publicly
if (
user.role != "admin"
and form_data.access_control == None
and not has_permission(
user.id,
"sharing.public_knowledge",
request.app.state.config.USER_PERMISSIONS,
)
):
form_data.access_control = {}
knowledge = Knowledges.insert_new_knowledge(user.id, form_data)
if knowledge:
@ -210,7 +224,8 @@ async def reindex_knowledge_files(request: Request, user=Depends(get_verified_us
failed_files = []
for file in files:
try:
process_file(
await run_in_threadpool(
process_file,
request,
ProcessFileForm(
file_id=file.id, collection_name=knowledge_base.id
@ -284,6 +299,7 @@ async def get_knowledge_by_id(id: str, user=Depends(get_verified_user)):
@router.post("/{id}/update", response_model=Optional[KnowledgeFilesResponse])
async def update_knowledge_by_id(
request: Request,
id: str,
form_data: KnowledgeForm,
user=Depends(get_verified_user),
@ -305,10 +321,22 @@ async def update_knowledge_by_id(
detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
)
# Check if user can share publicly
if (
user.role != "admin"
and form_data.access_control == None
and not has_permission(
user.id,
"sharing.public_knowledge",
request.app.state.config.USER_PERMISSIONS,
)
):
form_data.access_control = {}
knowledge = Knowledges.update_knowledge_by_id(id=id, form_data=form_data)
if knowledge:
file_ids = knowledge.data.get("file_ids", []) if knowledge.data else []
files = Files.get_files_by_ids(file_ids)
files = Files.get_file_metadatas_by_ids(file_ids)
return KnowledgeFilesResponse(
**knowledge.model_dump(),
@ -491,6 +519,7 @@ def update_file_from_knowledge_by_id(
def remove_file_from_knowledge_by_id(
id: str,
form_data: KnowledgeFileIdForm,
delete_file: bool = Query(True),
user=Depends(get_verified_user),
):
knowledge = Knowledges.get_knowledge_by_id(id=id)
@ -527,18 +556,19 @@ def remove_file_from_knowledge_by_id(
log.debug(e)
pass
try:
# Remove the file's collection from vector database
file_collection = f"file-{form_data.file_id}"
if VECTOR_DB_CLIENT.has_collection(collection_name=file_collection):
VECTOR_DB_CLIENT.delete_collection(collection_name=file_collection)
except Exception as e:
log.debug("This was most likely caused by bypassing embedding processing")
log.debug(e)
pass
if delete_file:
try:
# Remove the file's collection from vector database
file_collection = f"file-{form_data.file_id}"
if VECTOR_DB_CLIENT.has_collection(collection_name=file_collection):
VECTOR_DB_CLIENT.delete_collection(collection_name=file_collection)
except Exception as e:
log.debug("This was most likely caused by bypassing embedding processing")
log.debug(e)
pass
# Delete file from database
Files.delete_file_by_id(form_data.file_id)
# Delete file from database
Files.delete_file_by_id(form_data.file_id)
if knowledge:
data = knowledge.data or {}

38
backend/open_webui/routers/memories.py
View file

@ -1,6 +1,7 @@
from fastapi import APIRouter, Depends, HTTPException, Request
from pydantic import BaseModel
import logging
import asyncio
from typing import Optional
from open_webui.models.memories import Memories, MemoryModel
@ -17,7 +18,7 @@ router = APIRouter()
@router.get("/ef")
async def get_embeddings(request: Request):
return {"result": request.app.state.EMBEDDING_FUNCTION("hello world")}
return {"result": await request.app.state.EMBEDDING_FUNCTION("hello world")}
############################
@ -51,15 +52,15 @@ async def add_memory(
):
memory = Memories.insert_new_memory(user.id, form_data.content)
vector = await request.app.state.EMBEDDING_FUNCTION(memory.content, user=user)
VECTOR_DB_CLIENT.upsert(
collection_name=f"user-memory-{user.id}",
items=[
{
"id": memory.id,
"text": memory.content,
"vector": request.app.state.EMBEDDING_FUNCTION(
memory.content, user=user
),
"vector": vector,
"metadata": {"created_at": memory.created_at},
}
],
@ -82,9 +83,15 @@ class QueryMemoryForm(BaseModel):
async def query_memory(
request: Request, form_data: QueryMemoryForm, user=Depends(get_verified_user)
):
memories = Memories.get_memories_by_user_id(user.id)
if not memories:
raise HTTPException(status_code=404, detail="No memories found for user")
vector = await request.app.state.EMBEDDING_FUNCTION(form_data.content, user=user)
results = VECTOR_DB_CLIENT.search(
collection_name=f"user-memory-{user.id}",
vectors=[request.app.state.EMBEDDING_FUNCTION(form_data.content, user=user)],
vectors=[vector],
limit=form_data.k,
)
@ -101,21 +108,28 @@ async def reset_memory_from_vector_db(
VECTOR_DB_CLIENT.delete_collection(f"user-memory-{user.id}")
memories = Memories.get_memories_by_user_id(user.id)
# Generate vectors in parallel
vectors = await asyncio.gather(
*[
request.app.state.EMBEDDING_FUNCTION(memory.content, user=user)
for memory in memories
]
)
VECTOR_DB_CLIENT.upsert(
collection_name=f"user-memory-{user.id}",
items=[
{
"id": memory.id,
"text": memory.content,
"vector": request.app.state.EMBEDDING_FUNCTION(
memory.content, user=user
),
"vector": vectors[idx],
"metadata": {
"created_at": memory.created_at,
"updated_at": memory.updated_at,
},
}
for memory in memories
for idx, memory in enumerate(memories)
],
)
@ -160,15 +174,15 @@ async def update_memory_by_id(
raise HTTPException(status_code=404, detail="Memory not found")
if form_data.content is not None:
vector = await request.app.state.EMBEDDING_FUNCTION(memory.content, user=user)
VECTOR_DB_CLIENT.upsert(
collection_name=f"user-memory-{user.id}",
items=[
{
"id": memory.id,
"text": memory.content,
"vector": request.app.state.EMBEDDING_FUNCTION(
memory.content, user=user
),
"vector": vector,
"metadata": {
"created_at": memory.created_at,
"updated_at": memory.updated_at,

245
backend/open_webui/routers/models.py
View file

@ -1,34 +1,86 @@
from typing import Optional
import io
import base64
import json
import asyncio
import logging
from open_webui.models.models import (
ModelForm,
ModelModel,
ModelResponse,
ModelUserResponse,
ModelListResponse,
Models,
)
from pydantic import BaseModel
from open_webui.constants import ERROR_MESSAGES
from fastapi import APIRouter, Depends, HTTPException, Request, status
from fastapi import (
APIRouter,
Depends,
HTTPException,
Request,
status,
Response,
)
from fastapi.responses import FileResponse, StreamingResponse
from open_webui.utils.auth import get_admin_user, get_verified_user
from open_webui.utils.access_control import has_access, has_permission
from open_webui.config import BYPASS_ADMIN_ACCESS_CONTROL, STATIC_DIR
log = logging.getLogger(__name__)
router = APIRouter()
def is_valid_model_id(model_id: str) -> bool:
return model_id and len(model_id) <= 256
###########################
# GetModels
###########################
@router.get("/", response_model=list[ModelUserResponse])
async def get_models(id: Optional[str] = None, user=Depends(get_verified_user)):
if user.role == "admin":
return Models.get_models()
else:
return Models.get_models_by_user_id(user.id)
PAGE_ITEM_COUNT = 30
@router.get(
"/list", response_model=ModelListResponse
) # do NOT use "/" as path, conflicts with main.py
async def get_models(
query: Optional[str] = None,
view_option: Optional[str] = None,
tag: Optional[str] = None,
order_by: Optional[str] = None,
direction: Optional[str] = None,
page: Optional[int] = 1,
user=Depends(get_verified_user),
):
limit = PAGE_ITEM_COUNT
page = max(1, page)
skip = (page - 1) * limit
filter = {}
if query:
filter["query"] = query
if view_option:
filter["view_option"] = view_option
if tag:
filter["tag"] = tag
if order_by:
filter["order_by"] = order_by
if direction:
filter["direction"] = direction
if not user.role == "admin" or not BYPASS_ADMIN_ACCESS_CONTROL:
filter["user_id"] = user.id
return Models.search_models(user.id, filter=filter, skip=skip, limit=limit)
###########################
@ -41,6 +93,30 @@ async def get_base_models(user=Depends(get_admin_user)):
return Models.get_base_models()
###########################
# GetModelTags
###########################
@router.get("/tags", response_model=list[str])
async def get_model_tags(user=Depends(get_verified_user)):
if user.role == "admin" and BYPASS_ADMIN_ACCESS_CONTROL:
models = Models.get_models()
else:
models = Models.get_models_by_user_id(user.id)
tags_set = set()
for model in models:
if model.meta:
meta = model.meta.model_dump()
for tag in meta.get("tags", []):
tags_set.add((tag.get("name")))
tags = [tag for tag in tags_set]
tags.sort()
return tags
############################
# CreateNewModel
############################
@ -67,6 +143,12 @@ async def create_new_model(
detail=ERROR_MESSAGES.MODEL_ID_TAKEN,
)
if not is_valid_model_id(form_data.id):
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.MODEL_ID_TOO_LONG,
)
else:
model = Models.insert_new_model(form_data, user.id)
if model:
@ -78,18 +160,113 @@ async def create_new_model(
)
############################
# ExportModels
############################
@router.get("/export", response_model=list[ModelModel])
async def export_models(request: Request, user=Depends(get_verified_user)):
if user.role != "admin" and not has_permission(
user.id, "workspace.models_export", request.app.state.config.USER_PERMISSIONS
):
raise HTTPException(
status_code=status.HTTP_401_UNAUTHORIZED,
detail=ERROR_MESSAGES.UNAUTHORIZED,
)
if user.role == "admin" and BYPASS_ADMIN_ACCESS_CONTROL:
return Models.get_models()
else:
return Models.get_models_by_user_id(user.id)
############################
# ImportModels
############################
class ModelsImportForm(BaseModel):
models: list[dict]
@router.post("/import", response_model=bool)
async def import_models(
request: Request,
user=Depends(get_verified_user),
form_data: ModelsImportForm = (...),
):
if user.role != "admin" and not has_permission(
user.id, "workspace.models_import", request.app.state.config.USER_PERMISSIONS
):
raise HTTPException(
status_code=status.HTTP_401_UNAUTHORIZED,
detail=ERROR_MESSAGES.UNAUTHORIZED,
)
try:
data = form_data.models
if isinstance(data, list):
for model_data in data:
# Here, you can add logic to validate model_data if needed
model_id = model_data.get("id")
if model_id and is_valid_model_id(model_id):
existing_model = Models.get_model_by_id(model_id)
if existing_model:
# Update existing model
model_data["meta"] = model_data.get("meta", {})
model_data["params"] = model_data.get("params", {})
updated_model = ModelForm(
**{**existing_model.model_dump(), **model_data}
)
Models.update_model_by_id(model_id, updated_model)
else:
# Insert new model
model_data["meta"] = model_data.get("meta", {})
model_data["params"] = model_data.get("params", {})
new_model = ModelForm(**model_data)
Models.insert_new_model(user_id=user.id, form_data=new_model)
return True
else:
raise HTTPException(status_code=400, detail="Invalid JSON format")
except Exception as e:
log.exception(e)
raise HTTPException(status_code=500, detail=str(e))
############################
# SyncModels
############################
class SyncModelsForm(BaseModel):
models: list[ModelModel] = []
@router.post("/sync", response_model=list[ModelModel])
async def sync_models(
request: Request, form_data: SyncModelsForm, user=Depends(get_admin_user)
):
return Models.sync_models(user.id, form_data.models)
###########################
# GetModelById
###########################
class ModelIdForm(BaseModel):
id: str
# Note: We're not using the typical url path param here, but instead using a query parameter to allow '/' in the id
@router.get("/model", response_model=Optional[ModelResponse])
async def get_model_by_id(id: str, user=Depends(get_verified_user)):
model = Models.get_model_by_id(id)
if model:
if (
user.role == "admin"
(user.role == "admin" and BYPASS_ADMIN_ACCESS_CONTROL)
or model.user_id == user.id
or has_access(user.id, "read", model.access_control)
):
@ -101,8 +278,42 @@ async def get_model_by_id(id: str, user=Depends(get_verified_user)):
)
###########################
# GetModelById
###########################
@router.get("/model/profile/image")
async def get_model_profile_image(id: str, user=Depends(get_verified_user)):
model = Models.get_model_by_id(id)
if model:
if model.meta.profile_image_url:
if model.meta.profile_image_url.startswith("http"):
return Response(
status_code=status.HTTP_302_FOUND,
headers={"Location": model.meta.profile_image_url},
)
elif model.meta.profile_image_url.startswith("data:image"):
try:
header, base64_data = model.meta.profile_image_url.split(",", 1)
image_data = base64.b64decode(base64_data)
image_buffer = io.BytesIO(image_data)
return StreamingResponse(
image_buffer,
media_type="image/png",
headers={"Content-Disposition": "inline; filename=image.png"},
)
except Exception as e:
pass
return FileResponse(f"{STATIC_DIR}/favicon.png")
else:
return FileResponse(f"{STATIC_DIR}/favicon.png")
############################
# ToggelModelById
# ToggleModelById
############################
@ -143,12 +354,10 @@ async def toggle_model_by_id(id: str, user=Depends(get_verified_user)):
@router.post("/model/update", response_model=Optional[ModelModel])
async def update_model_by_id(
id: str,
form_data: ModelForm,
user=Depends(get_verified_user),
):
model = Models.get_model_by_id(id)
model = Models.get_model_by_id(form_data.id)
if not model:
raise HTTPException(
status_code=status.HTTP_401_UNAUTHORIZED,
@ -165,7 +374,7 @@ async def update_model_by_id(
detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
)
model = Models.update_model_by_id(id, form_data)
model = Models.update_model_by_id(form_data.id, ModelForm(**form_data.model_dump()))
return model
@ -174,9 +383,9 @@ async def update_model_by_id(
############################
@router.delete("/model/delete", response_model=bool)
async def delete_model_by_id(id: str, user=Depends(get_verified_user)):
model = Models.get_model_by_id(id)
@router.post("/model/delete", response_model=bool)
async def delete_model_by_id(form_data: ModelIdForm, user=Depends(get_verified_user)):
model = Models.get_model_by_id(form_data.id)
if not model:
raise HTTPException(
status_code=status.HTTP_401_UNAUTHORIZED,
@ -193,7 +402,7 @@ async def delete_model_by_id(id: str, user=Depends(get_verified_user)):
detail=ERROR_MESSAGES.UNAUTHORIZED,
)
result = Models.delete_model_by_id(id)
result = Models.delete_model_by_id(form_data.id)
return result

50
backend/open_webui/routers/notes.py
View file

@ -6,6 +6,9 @@ from typing import Optional
from fastapi import APIRouter, Depends, HTTPException, Request, status, BackgroundTasks
from pydantic import BaseModel
from open_webui.socket.main import sio
from open_webui.models.users import Users, UserResponse
from open_webui.models.notes import Notes, NoteModel, NoteForm, NoteUserResponse
@ -45,15 +48,23 @@ async def get_notes(request: Request, user=Depends(get_verified_user)):
"user": UserResponse(**Users.get_user_by_id(note.user_id).model_dump()),
}
)
for note in Notes.get_notes_by_user_id(user.id, "write")
for note in Notes.get_notes_by_permission(user.id, "write")
]
return notes
@router.get("/list", response_model=list[NoteUserResponse])
async def get_note_list(request: Request, user=Depends(get_verified_user)):
class NoteTitleIdResponse(BaseModel):
id: str
title: str
updated_at: int
created_at: int
@router.get("/list", response_model=list[NoteTitleIdResponse])
async def get_note_list(
request: Request, page: Optional[int] = None, user=Depends(get_verified_user)
):
if user.role != "admin" and not has_permission(
user.id, "features.notes", request.app.state.config.USER_PERMISSIONS
):
@ -62,14 +73,17 @@ async def get_note_list(request: Request, user=Depends(get_verified_user)):
detail=ERROR_MESSAGES.UNAUTHORIZED,
)
limit = None
skip = None
if page is not None:
limit = 60
skip = (page - 1) * limit
notes = [
NoteUserResponse(
**{
**note.model_dump(),
"user": UserResponse(**Users.get_user_by_id(note.user_id).model_dump()),
}
NoteTitleIdResponse(**note.model_dump())
for note in Notes.get_notes_by_permission(
user.id, "write", skip=skip, limit=limit
)
for note in Notes.get_notes_by_user_id(user.id, "read")
]
return notes
@ -166,8 +180,26 @@ async def update_note_by_id(
status_code=status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.DEFAULT()
)
# Check if user can share publicly
if (
user.role != "admin"
and form_data.access_control == None
and not has_permission(
user.id,
"sharing.public_notes",
request.app.state.config.USER_PERMISSIONS,
)
):
form_data.access_control = {}
try:
note = Notes.update_note_by_id(id, form_data)
await sio.emit(
"note-events",
note.model_dump(),
to=f"note:{note.id}",
)
return note
except Exception as e:
log.exception(e)

332
backend/open_webui/routers/ollama.py
View file

@ -16,8 +16,8 @@ from urllib.parse import urlparse
import aiohttp
from aiocache import cached
import requests
from urllib.parse import quote
from open_webui.utils.headers import include_user_info_headers
from open_webui.models.chats import Chats
from open_webui.models.users import UserModel
@ -47,7 +47,7 @@ from open_webui.utils.misc import (
from open_webui.utils.payload import (
apply_model_params_to_body_ollama,
apply_model_params_to_body_openai,
apply_model_system_prompt_to_body,
apply_system_prompt_to_body,
)
from open_webui.utils.auth import get_admin_user, get_verified_user
from open_webui.utils.access_control import has_access
@ -59,6 +59,7 @@ from open_webui.config import (
from open_webui.env import (
ENV,
SRC_LOG_LEVELS,
MODELS_CACHE_TTL,
AIOHTTP_CLIENT_SESSION_SSL,
AIOHTTP_CLIENT_TIMEOUT,
AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST,
@ -81,22 +82,17 @@ async def send_get_request(url, key=None, user: UserModel = None):
timeout = aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST)
try:
async with aiohttp.ClientSession(timeout=timeout, trust_env=True) as session:
headers = {
"Content-Type": "application/json",
**({"Authorization": f"Bearer {key}"} if key else {}),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user:
headers = include_user_info_headers(headers, user)
async with session.get(
url,
headers={
"Content-Type": "application/json",
**({"Authorization": f"Bearer {key}"} if key else {}),
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user
else {}
),
},
headers=headers,
ssl=AIOHTTP_CLIENT_SESSION_SSL,
) as response:
return await response.json()
@ -123,6 +119,7 @@ async def send_post_request(
key: Optional[str] = None,
content_type: Optional[str] = None,
user: UserModel = None,
metadata: Optional[dict] = None,
):
r = None
@ -131,23 +128,20 @@ async def send_post_request(
trust_env=True, timeout=aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT)
)
headers = {
"Content-Type": "application/json",
**({"Authorization": f"Bearer {key}"} if key else {}),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user:
headers = include_user_info_headers(headers, user)
if metadata and metadata.get("chat_id"):
headers["X-OpenWebUI-Chat-Id"] = metadata.get("chat_id")
r = await session.post(
url,
data=payload,
headers={
"Content-Type": "application/json",
**({"Authorization": f"Bearer {key}"} if key else {}),
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user
else {}
),
},
headers=headers,
ssl=AIOHTTP_CLIENT_SESSION_SSL,
)
@ -183,7 +177,6 @@ async def send_post_request(
)
else:
res = await r.json()
await cleanup_response(r, session)
return res
except HTTPException as e:
@ -195,6 +188,9 @@ async def send_post_request(
status_code=r.status if r else 500,
detail=detail if e else "Open WebUI: Server Connection Error",
)
finally:
if not stream:
await cleanup_response(r, session)
def get_api_key(idx, url, configs):
@ -237,21 +233,16 @@ async def verify_connection(
timeout=aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST),
) as session:
try:
headers = {
**({"Authorization": f"Bearer {key}"} if key else {}),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user:
headers = include_user_info_headers(headers, user)
async with session.get(
f"{url}/api/version",
headers={
**({"Authorization": f"Bearer {key}"} if key else {}),
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user
else {}
),
},
headers=headers,
ssl=AIOHTTP_CLIENT_SESSION_SSL,
) as r:
if r.status != 200:
@ -320,17 +311,21 @@ def merge_ollama_models_lists(model_lists):
for idx, model_list in enumerate(model_lists):
if model_list is not None:
for model in model_list:
id = model["model"]
if id not in merged_models:
model["urls"] = [idx]
merged_models[id] = model
else:
merged_models[id]["urls"].append(idx)
id = model.get("model")
if id is not None:
if id not in merged_models:
model["urls"] = [idx]
merged_models[id] = model
else:
merged_models[id]["urls"].append(idx)
return list(merged_models.values())
@cached(ttl=1)
@cached(
ttl=MODELS_CACHE_TTL,
key=lambda _, user: f"ollama_all_models_{user.id}" if user else "ollama_all_models",
)
async def get_all_models(request: Request, user: UserModel = None):
log.info("get_all_models()")
if request.app.state.config.ENABLE_OLLAMA_API:
@ -406,15 +401,15 @@ async def get_all_models(request: Request, user: UserModel = None):
try:
loaded_models = await get_ollama_loaded_models(request, user=user)
expires_map = {
m["name"]: m["expires_at"]
m["model"]: m["expires_at"]
for m in loaded_models["models"]
if "expires_at" in m
}
for m in models["models"]:
if m["name"] in expires_map:
if m["model"] in expires_map:
# Parse ISO8601 datetime with offset, get unix timestamp as int
dt = datetime.fromisoformat(expires_map[m["name"]])
dt = datetime.fromisoformat(expires_map[m["model"]])
m["expires_at"] = int(dt.timestamp())
except Exception as e:
log.debug(f"Failed to get loaded models: {e}")
@ -456,22 +451,17 @@ async def get_ollama_tags(
r = None
try:
headers = {
**({"Authorization": f"Bearer {key}"} if key else {}),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user:
headers = include_user_info_headers(headers, user)
r = requests.request(
method="GET",
url=f"{url}/api/tags",
headers={
**({"Authorization": f"Bearer {key}"} if key else {}),
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user
else {}
),
},
headers=headers,
)
r.raise_for_status()
@ -635,7 +625,10 @@ async def get_ollama_versions(request: Request, url_idx: Optional[int] = None):
class ModelNameForm(BaseModel):
name: str
model: Optional[str] = None
model_config = ConfigDict(
extra="allow",
)
@router.post("/api/unload")
@ -644,10 +637,12 @@ async def unload_model(
form_data: ModelNameForm,
user=Depends(get_admin_user),
):
model_name = form_data.name
form_data = form_data.model_dump(exclude_none=True)
model_name = form_data.get("model", form_data.get("name"))
if not model_name:
raise HTTPException(
status_code=400, detail="Missing 'name' of model to unload."
status_code=400, detail="Missing name of the model to unload."
)
# Refresh/load models if needed, get mapping from name to URLs
@ -710,11 +705,14 @@ async def pull_model(
url_idx: int = 0,
user=Depends(get_admin_user),
):
form_data = form_data.model_dump(exclude_none=True)
form_data["model"] = form_data.get("model", form_data.get("name"))
url = request.app.state.config.OLLAMA_BASE_URLS[url_idx]
log.info(f"url: {url}")
# Admin should be able to pull models from any source
payload = {**form_data.model_dump(exclude_none=True), "insecure": True}
payload = {**form_data, "insecure": True}
return await send_post_request(
url=f"{url}/api/pull",
@ -725,7 +723,7 @@ async def pull_model(
class PushModelForm(BaseModel):
name: str
model: str
insecure: Optional[bool] = None
stream: Optional[bool] = None
@ -742,12 +740,12 @@ async def push_model(
await get_all_models(request, user=user)
models = request.app.state.OLLAMA_MODELS
if form_data.name in models:
url_idx = models[form_data.name]["urls"][0]
if form_data.model in models:
url_idx = models[form_data.model]["urls"][0]
else:
raise HTTPException(
status_code=400,
detail=ERROR_MESSAGES.MODEL_NOT_FOUND(form_data.name),
detail=ERROR_MESSAGES.MODEL_NOT_FOUND(form_data.model),
)
url = request.app.state.config.OLLAMA_BASE_URLS[url_idx]
@ -817,23 +815,18 @@ async def copy_model(
key = get_api_key(url_idx, url, request.app.state.config.OLLAMA_API_CONFIGS)
try:
headers = {
"Content-Type": "application/json",
**({"Authorization": f"Bearer {key}"} if key else {}),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user:
headers = include_user_info_headers(headers, user)
r = requests.request(
method="POST",
url=f"{url}/api/copy",
headers={
"Content-Type": "application/json",
**({"Authorization": f"Bearer {key}"} if key else {}),
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user
else {}
),
},
headers=headers,
data=form_data.model_dump_json(exclude_none=True).encode(),
)
r.raise_for_status()
@ -866,40 +859,40 @@ async def delete_model(
url_idx: Optional[int] = None,
user=Depends(get_admin_user),
):
form_data = form_data.model_dump(exclude_none=True)
form_data["model"] = form_data.get("model", form_data.get("name"))
model = form_data.get("model")
if url_idx is None:
await get_all_models(request, user=user)
models = request.app.state.OLLAMA_MODELS
if form_data.name in models:
url_idx = models[form_data.name]["urls"][0]
if model in models:
url_idx = models[model]["urls"][0]
else:
raise HTTPException(
status_code=400,
detail=ERROR_MESSAGES.MODEL_NOT_FOUND(form_data.name),
detail=ERROR_MESSAGES.MODEL_NOT_FOUND(model),
)
url = request.app.state.config.OLLAMA_BASE_URLS[url_idx]
key = get_api_key(url_idx, url, request.app.state.config.OLLAMA_API_CONFIGS)
try:
headers = {
"Content-Type": "application/json",
**({"Authorization": f"Bearer {key}"} if key else {}),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user:
headers = include_user_info_headers(headers, user)
r = requests.request(
method="DELETE",
url=f"{url}/api/delete",
headers=headers,
data=form_data.model_dump_json(exclude_none=True).encode(),
headers={
"Content-Type": "application/json",
**({"Authorization": f"Bearer {key}"} if key else {}),
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user
else {}
),
},
)
r.raise_for_status()
@ -927,38 +920,38 @@ async def delete_model(
async def show_model_info(
request: Request, form_data: ModelNameForm, user=Depends(get_verified_user)
):
form_data = form_data.model_dump(exclude_none=True)
form_data["model"] = form_data.get("model", form_data.get("name"))
await get_all_models(request, user=user)
models = request.app.state.OLLAMA_MODELS
if form_data.name not in models:
model = form_data.get("model")
if model not in models:
raise HTTPException(
status_code=400,
detail=ERROR_MESSAGES.MODEL_NOT_FOUND(form_data.name),
detail=ERROR_MESSAGES.MODEL_NOT_FOUND(model),
)
url_idx = random.choice(models[form_data.name]["urls"])
url_idx = random.choice(models[model]["urls"])
url = request.app.state.config.OLLAMA_BASE_URLS[url_idx]
key = get_api_key(url_idx, url, request.app.state.config.OLLAMA_API_CONFIGS)
try:
headers = {
"Content-Type": "application/json",
**({"Authorization": f"Bearer {key}"} if key else {}),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user:
headers = include_user_info_headers(headers, user)
r = requests.request(
method="POST",
url=f"{url}/api/show",
headers={
"Content-Type": "application/json",
**({"Authorization": f"Bearer {key}"} if key else {}),
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user
else {}
),
},
headers=headers,
data=form_data.model_dump_json(exclude_none=True).encode(),
)
r.raise_for_status()
@ -989,6 +982,10 @@ class GenerateEmbedForm(BaseModel):
options: Optional[dict] = None
keep_alive: Optional[Union[int, str]] = None
model_config = ConfigDict(
extra="allow",
)
@router.post("/api/embed")
@router.post("/api/embed/{url_idx}")
@ -1029,23 +1026,18 @@ async def embed(
form_data.model = form_data.model.replace(f"{prefix_id}.", "")
try:
headers = {
"Content-Type": "application/json",
**({"Authorization": f"Bearer {key}"} if key else {}),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user:
headers = include_user_info_headers(headers, user)
r = requests.request(
method="POST",
url=f"{url}/api/embed",
headers={
"Content-Type": "application/json",
**({"Authorization": f"Bearer {key}"} if key else {}),
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user
else {}
),
},
headers=headers,
data=form_data.model_dump_json(exclude_none=True).encode(),
)
r.raise_for_status()
@ -1116,23 +1108,18 @@ async def embeddings(
form_data.model = form_data.model.replace(f"{prefix_id}.", "")
try:
headers = {
"Content-Type": "application/json",
**({"Authorization": f"Bearer {key}"} if key else {}),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user:
headers = include_user_info_headers(headers, user)
r = requests.request(
method="POST",
url=f"{url}/api/embeddings",
headers={
"Content-Type": "application/json",
**({"Authorization": f"Bearer {key}"} if key else {}),
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user
else {}
),
},
headers=headers,
data=form_data.model_dump_json(exclude_none=True).encode(),
)
r.raise_for_status()
@ -1303,7 +1290,7 @@ async def generate_chat_completion(
system = params.pop("system", None)
payload = apply_model_params_to_body_ollama(params, payload)
payload = apply_model_system_prompt_to_body(system, payload, metadata, user)
payload = apply_system_prompt_to_body(system, payload, metadata, user)
# Check if user has access to the model
if not bypass_filter and user.role == "user":
@ -1344,6 +1331,7 @@ async def generate_chat_completion(
key=get_api_key(url_idx, url, request.app.state.config.OLLAMA_API_CONFIGS),
content_type="application/x-ndjson",
user=user,
metadata=metadata,
)
@ -1382,6 +1370,8 @@ async def generate_openai_completion(
url_idx: Optional[int] = None,
user=Depends(get_verified_user),
):
metadata = form_data.pop("metadata", None)
try:
form_data = OpenAICompletionForm(**form_data)
except Exception as e:
@ -1447,6 +1437,7 @@ async def generate_openai_completion(
stream=payload.get("stream", False),
key=get_api_key(url_idx, url, request.app.state.config.OLLAMA_API_CONFIGS),
user=user,
metadata=metadata,
)
@ -1488,7 +1479,7 @@ async def generate_openai_chat_completion(
system = params.pop("system", None)
payload = apply_model_params_to_body_openai(params, payload)
payload = apply_model_system_prompt_to_body(system, payload, metadata, user)
payload = apply_system_prompt_to_body(system, payload, metadata, user)
# Check if user has access to the model
if user.role == "user":
@ -1528,6 +1519,7 @@ async def generate_openai_chat_completion(
stream=payload.get("stream", False),
key=get_api_key(url_idx, url, request.app.state.config.OLLAMA_API_CONFIGS),
user=user,
metadata=metadata,
)
@ -1658,25 +1650,27 @@ async def download_file_stream(
yield f'data: {{"progress": {progress}, "completed": {current_size}, "total": {total_size}}}\n\n'
if done:
file.seek(0)
chunk_size = 1024 * 1024 * 2
hashed = calculate_sha256(file, chunk_size)
file.seek(0)
file.close()
url = f"{ollama_url}/api/blobs/sha256:{hashed}"
response = requests.post(url, data=file)
with open(file_path, "rb") as file:
chunk_size = 1024 * 1024 * 2
hashed = calculate_sha256(file, chunk_size)
if response.ok:
res = {
"done": done,
"blob": f"sha256:{hashed}",
"name": file_name,
}
os.remove(file_path)
url = f"{ollama_url}/api/blobs/sha256:{hashed}"
with requests.Session() as session:
response = session.post(url, data=file, timeout=30)
yield f"data: {json.dumps(res)}\n\n"
else:
raise "Ollama: Could not create blob, Please try again."
if response.ok:
res = {
"done": done,
"blob": f"sha256:{hashed}",
"name": file_name,
}
os.remove(file_path)
yield f"data: {json.dumps(res)}\n\n"
else:
raise "Ollama: Could not create blob, Please try again."
# url = "https://huggingface.co/TheBloke/stablelm-zephyr-3b-GGUF/resolve/main/stablelm-zephyr-3b.Q2_K.gguf"

578
backend/open_webui/routers/openai.py
View file

@ -2,17 +2,21 @@ import asyncio
import hashlib
import json
import logging
from pathlib import Path
from typing import Literal, Optional, overload
from typing import Optional
import aiohttp
from aiocache import cached
import requests
from urllib.parse import quote
from fastapi import Depends, FastAPI, HTTPException, Request, APIRouter
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import FileResponse, StreamingResponse
from azure.identity import DefaultAzureCredential, get_bearer_token_provider
from fastapi import Depends, HTTPException, Request, APIRouter
from fastapi.responses import (
FileResponse,
StreamingResponse,
JSONResponse,
PlainTextResponse,
)
from pydantic import BaseModel
from starlette.background import BackgroundTask
@ -21,6 +25,7 @@ from open_webui.config import (
CACHE_DIR,
)
from open_webui.env import (
MODELS_CACHE_TTL,
AIOHTTP_CLIENT_SESSION_SSL,
AIOHTTP_CLIENT_TIMEOUT,
AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST,
@ -30,19 +35,21 @@ from open_webui.env import (
from open_webui.models.users import UserModel
from open_webui.constants import ERROR_MESSAGES
from open_webui.env import ENV, SRC_LOG_LEVELS
from open_webui.env import SRC_LOG_LEVELS
from open_webui.utils.payload import (
apply_model_params_to_body_openai,
apply_model_system_prompt_to_body,
apply_system_prompt_to_body,
)
from open_webui.utils.misc import (
convert_logit_bias_input_to_json,
stream_chunks_handler,
)
from open_webui.utils.auth import get_admin_user, get_verified_user
from open_webui.utils.access_control import has_access
from open_webui.utils.headers import include_user_info_headers
log = logging.getLogger(__name__)
@ -60,21 +67,16 @@ async def send_get_request(url, key=None, user: UserModel = None):
timeout = aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST)
try:
async with aiohttp.ClientSession(timeout=timeout, trust_env=True) as session:
headers = {
**({"Authorization": f"Bearer {key}"} if key else {}),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user:
headers = include_user_info_headers(headers, user)
async with session.get(
url,
headers={
**({"Authorization": f"Bearer {key}"} if key else {}),
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user
else {}
),
},
headers=headers,
ssl=AIOHTTP_CLIENT_SESSION_SSL,
) as response:
return await response.json()
@ -94,12 +96,12 @@ async def cleanup_response(
await session.close()
def openai_o_series_handler(payload):
def openai_reasoning_model_handler(payload):
"""
Handle "o" series specific parameters
Handle reasoning model specific parameters
"""
if "max_tokens" in payload:
# Convert "max_tokens" to "max_completion_tokens" for all o-series models
# Convert "max_tokens" to "max_completion_tokens" for all reasoning models
payload["max_completion_tokens"] = payload["max_tokens"]
del payload["max_tokens"]
@ -115,6 +117,86 @@ def openai_o_series_handler(payload):
return payload
async def get_headers_and_cookies(
request: Request,
url,
key=None,
config=None,
metadata: Optional[dict] = None,
user: UserModel = None,
):
cookies = {}
headers = {
"Content-Type": "application/json",
**(
{
"HTTP-Referer": "https://openwebui.com/",
"X-Title": "Open WebUI",
}
if "openrouter.ai" in url
else {}
),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user:
headers = include_user_info_headers(headers, user)
if metadata and metadata.get("chat_id"):
headers["X-OpenWebUI-Chat-Id"] = metadata.get("chat_id")
token = None
auth_type = config.get("auth_type")
if auth_type == "bearer" or auth_type is None:
# Default to bearer if not specified
token = f"{key}"
elif auth_type == "none":
token = None
elif auth_type == "session":
cookies = request.cookies
token = request.state.token.credentials
elif auth_type == "system_oauth":
cookies = request.cookies
oauth_token = None
try:
if request.cookies.get("oauth_session_id", None):
oauth_token = await request.app.state.oauth_manager.get_oauth_token(
user.id,
request.cookies.get("oauth_session_id", None),
)
except Exception as e:
log.error(f"Error getting OAuth token: {e}")
if oauth_token:
token = f"{oauth_token.get('access_token', '')}"
elif auth_type in ("azure_ad", "microsoft_entra_id"):
token = get_microsoft_entra_id_access_token()
if token:
headers["Authorization"] = f"Bearer {token}"
if config.get("headers") and isinstance(config.get("headers"), dict):
headers = {**headers, **config.get("headers")}
return headers, cookies
def get_microsoft_entra_id_access_token():
"""
Get Microsoft Entra ID access token using DefaultAzureCredential for Azure OpenAI.
Returns the token string or None if authentication fails.
"""
try:
token_provider = get_bearer_token_provider(
DefaultAzureCredential(), "https://cognitiveservices.azure.com/.default"
)
return token_provider()
except Exception as e:
log.error(f"Error getting Microsoft Entra ID access token: {e}")
return None
##########################################
#
# API routes
@ -206,34 +288,23 @@ async def speech(request: Request, user=Depends(get_verified_user)):
return FileResponse(file_path)
url = request.app.state.config.OPENAI_API_BASE_URLS[idx]
key = request.app.state.config.OPENAI_API_KEYS[idx]
api_config = request.app.state.config.OPENAI_API_CONFIGS.get(
str(idx),
request.app.state.config.OPENAI_API_CONFIGS.get(url, {}), # Legacy support
)
headers, cookies = await get_headers_and_cookies(
request, url, key, api_config, user=user
)
r = None
try:
r = requests.post(
url=f"{url}/audio/speech",
data=body,
headers={
"Content-Type": "application/json",
"Authorization": f"Bearer {request.app.state.config.OPENAI_API_KEYS[idx]}",
**(
{
"HTTP-Referer": "https://openwebui.com/",
"X-Title": "Open WebUI",
}
if "openrouter.ai" in url
else {}
),
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS
else {}
),
},
headers=headers,
cookies=cookies,
stream=True,
)
@ -357,11 +428,22 @@ async def get_all_models_responses(request: Request, user: UserModel) -> list:
prefix_id = api_config.get("prefix_id", None)
tags = api_config.get("tags", [])
for model in (
model_list = (
response if isinstance(response, list) else response.get("data", [])
):
)
if not isinstance(model_list, list):
# Catch non-list responses
model_list = []
for model in model_list:
# Remove name key if its value is None #16689
if "name" in model and model["name"] is None:
del model["name"]
if prefix_id:
model["id"] = f"{prefix_id}.{model['id']}"
model["id"] = (
f"{prefix_id}.{model.get('id', model.get('name', ''))}"
)
if tags:
model["tags"] = tags
@ -386,7 +468,10 @@ async def get_filtered_models(models, user):
return filtered_models
@cached(ttl=1)
@cached(
ttl=MODELS_CACHE_TTL,
key=lambda _, user: f"openai_all_models_{user.id}" if user else "openai_all_models",
)
async def get_all_models(request: Request, user: UserModel) -> dict[str, list]:
log.info("get_all_models()")
@ -402,50 +487,55 @@ async def get_all_models(request: Request, user: UserModel) -> dict[str, list]:
return response
return None
def merge_models_lists(model_lists):
def is_supported_openai_models(model_id):
if any(
name in model_id
for name in [
"babbage",
"dall-e",
"davinci",
"embedding",
"tts",
"whisper",
]
):
return False
return True
def get_merged_models(model_lists):
log.debug(f"merge_models_lists {model_lists}")
merged_list = []
models = {}
for idx, models in enumerate(model_lists):
if models is not None and "error" not in models:
for idx, model_list in enumerate(model_lists):
if model_list is not None and "error" not in model_list:
for model in model_list:
model_id = model.get("id") or model.get("name")
merged_list.extend(
[
{
if (
"api.openai.com"
in request.app.state.config.OPENAI_API_BASE_URLS[idx]
and not is_supported_openai_models(model_id)
):
# Skip unwanted OpenAI models
continue
if model_id and model_id not in models:
models[model_id] = {
**model,
"name": model.get("name", model["id"]),
"name": model.get("name", model_id),
"owned_by": "openai",
"openai": model,
"connection_type": model.get("connection_type", "external"),
"urlIdx": idx,
}
for model in models
if (model.get("id") or model.get("name"))
and (
"api.openai.com"
not in request.app.state.config.OPENAI_API_BASE_URLS[idx]
or not any(
name in model["id"]
for name in [
"babbage",
"dall-e",
"davinci",
"embedding",
"tts",
"whisper",
]
)
)
]
)
return merged_list
return models
models = {"data": merge_models_lists(map(extract_data, responses))}
models = get_merged_models(map(extract_data, responses))
log.debug(f"models: {models}")
request.app.state.OPENAI_MODELS = {model["id"]: model for model in models["data"]}
return models
request.app.state.OPENAI_MODELS = models
return {"data": list(models.values())}
@router.get("/models")
@ -474,19 +564,9 @@ async def get_models(
timeout=aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST),
) as session:
try:
headers = {
"Content-Type": "application/json",
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS
else {}
),
}
headers, cookies = await get_headers_and_cookies(
request, url, key, api_config, user=user
)
if api_config.get("azure", False):
models = {
@ -494,11 +574,10 @@ async def get_models(
"object": "list",
}
else:
headers["Authorization"] = f"Bearer {key}"
async with session.get(
f"{url}/models",
headers=headers,
cookies=cookies,
ssl=AIOHTTP_CLIENT_SESSION_SSL,
) as r:
if r.status != 200:
@ -557,7 +636,9 @@ class ConnectionVerificationForm(BaseModel):
@router.post("/verify")
async def verify_connection(
form_data: ConnectionVerificationForm, user=Depends(get_admin_user)
request: Request,
form_data: ConnectionVerificationForm,
user=Depends(get_admin_user),
):
url = form_data.url
key = form_data.key
@ -569,56 +650,61 @@ async def verify_connection(
timeout=aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST),
) as session:
try:
headers = {
"Content-Type": "application/json",
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS
else {}
),
}
headers, cookies = await get_headers_and_cookies(
request, url, key, api_config, user=user
)
if api_config.get("azure", False):
headers["api-key"] = key
api_version = api_config.get("api_version", "") or "2023-03-15-preview"
# Only set api-key header if not using Azure Entra ID authentication
auth_type = api_config.get("auth_type", "bearer")
if auth_type not in ("azure_ad", "microsoft_entra_id"):
headers["api-key"] = key
api_version = api_config.get("api_version", "") or "2023-03-15-preview"
async with session.get(
url=f"{url}/openai/models?api-version={api_version}",
headers=headers,
cookies=cookies,
ssl=AIOHTTP_CLIENT_SESSION_SSL,
) as r:
if r.status != 200:
# Extract response error details if available
error_detail = f"HTTP Error: {r.status}"
res = await r.json()
if "error" in res:
error_detail = f"External Error: {res['error']}"
raise Exception(error_detail)
try:
response_data = await r.json()
except Exception:
response_data = await r.text()
if r.status != 200:
if isinstance(response_data, (dict, list)):
return JSONResponse(
status_code=r.status, content=response_data
)
else:
return PlainTextResponse(
status_code=r.status, content=response_data
)
response_data = await r.json()
return response_data
else:
headers["Authorization"] = f"Bearer {key}"
async with session.get(
f"{url}/models",
headers=headers,
cookies=cookies,
ssl=AIOHTTP_CLIENT_SESSION_SSL,
) as r:
if r.status != 200:
# Extract response error details if available
error_detail = f"HTTP Error: {r.status}"
res = await r.json()
if "error" in res:
error_detail = f"External Error: {res['error']}"
raise Exception(error_detail)
try:
response_data = await r.json()
except Exception:
response_data = await r.text()
if r.status != 200:
if isinstance(response_data, (dict, list)):
return JSONResponse(
status_code=r.status, content=response_data
)
else:
return PlainTextResponse(
status_code=r.status, content=response_data
)
response_data = await r.json()
return response_data
except aiohttp.ClientError as e:
@ -629,17 +715,12 @@ async def verify_connection(
)
except Exception as e:
log.exception(f"Unexpected error: {e}")
error_detail = f"Unexpected error: {str(e)}"
raise HTTPException(status_code=500, detail=error_detail)
raise HTTPException(
status_code=500, detail="Open WebUI: Server Connection Error"
)
def convert_to_azure_payload(
url,
payload: dict,
):
model = payload.get("model", "")
# Filter allowed parameters based on Azure OpenAI API
def get_azure_allowed_params(api_version: str) -> set[str]:
allowed_params = {
"messages",
"temperature",
@ -667,10 +748,32 @@ def convert_to_azure_payload(
"response_format",
"seed",
"max_completion_tokens",
"reasoning_effort",
}
try:
if api_version >= "2024-09-01-preview":
allowed_params.add("stream_options")
except ValueError:
log.debug(
f"Invalid API version {api_version} for Azure OpenAI. Defaulting to allowed parameters."
)
return allowed_params
def is_openai_reasoning_model(model: str) -> bool:
return model.lower().startswith(("o1", "o3", "o4", "gpt-5"))
def convert_to_azure_payload(url, payload: dict, api_version: str):
model = payload.get("model", "")
# Filter allowed parameters based on Azure OpenAI API
allowed_params = get_azure_allowed_params(api_version)
# Special handling for o-series models
if model.startswith("o") and model.endswith("-mini"):
if is_openai_reasoning_model(model):
# Convert max_tokens to max_completion_tokens for o-series models
if "max_tokens" in payload:
payload["max_completion_tokens"] = payload["max_tokens"]
@ -720,7 +823,7 @@ async def generate_chat_completion(
system = params.pop("system", None)
payload = apply_model_params_to_body_openai(params, payload)
payload = apply_model_system_prompt_to_body(system, payload, metadata, user)
payload = apply_system_prompt_to_body(system, payload, metadata, user)
# Check if user has access to the model
if not bypass_filter and user.role == "user":
@ -775,10 +878,9 @@ async def generate_chat_completion(
url = request.app.state.config.OPENAI_API_BASE_URLS[idx]
key = request.app.state.config.OPENAI_API_KEYS[idx]
# Check if model is from "o" series
is_o_series = payload["model"].lower().startswith(("o1", "o3", "o4"))
if is_o_series:
payload = openai_o_series_handler(payload)
# Check if model is a reasoning model that needs special handling
if is_openai_reasoning_model(payload["model"]):
payload = openai_reasoning_model_handler(payload)
elif "api.openai.com" not in url:
# Remove "max_completion_tokens" from the payload for backward compatibility
if "max_completion_tokens" in payload:
@ -794,37 +896,23 @@ async def generate_chat_completion(
convert_logit_bias_input_to_json(payload["logit_bias"])
)
headers = {
"Content-Type": "application/json",
**(
{
"HTTP-Referer": "https://openwebui.com/",
"X-Title": "Open WebUI",
}
if "openrouter.ai" in url
else {}
),
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS
else {}
),
}
headers, cookies = await get_headers_and_cookies(
request, url, key, api_config, metadata, user=user
)
if api_config.get("azure", False):
request_url, payload = convert_to_azure_payload(url, payload)
api_version = api_config.get("api_version", "") or "2023-03-15-preview"
headers["api-key"] = key
api_version = api_config.get("api_version", "2023-03-15-preview")
request_url, payload = convert_to_azure_payload(url, payload, api_version)
# Only set api-key header if not using Azure Entra ID authentication
auth_type = api_config.get("auth_type", "bearer")
if auth_type not in ("azure_ad", "microsoft_entra_id"):
headers["api-key"] = key
headers["api-version"] = api_version
request_url = f"{request_url}/chat/completions?api-version={api_version}"
else:
request_url = f"{url}/chat/completions"
headers["Authorization"] = f"Bearer {key}"
payload = json.dumps(payload)
@ -843,6 +931,7 @@ async def generate_chat_completion(
url=request_url,
data=payload,
headers=headers,
cookies=cookies,
ssl=AIOHTTP_CLIENT_SESSION_SSL,
)
@ -850,7 +939,7 @@ async def generate_chat_completion(
if "text/event-stream" in r.headers.get("Content-Type", ""):
streaming = True
return StreamingResponse(
r.content,
stream_chunks_handler(r.content),
status_code=r.status,
headers=dict(r.headers),
background=BackgroundTask(
@ -864,27 +953,23 @@ async def generate_chat_completion(
log.error(e)
response = await r.text()
r.raise_for_status()
if r.status >= 400:
if isinstance(response, (dict, list)):
return JSONResponse(status_code=r.status, content=response)
else:
return PlainTextResponse(status_code=r.status, content=response)
return response
except Exception as e:
log.exception(e)
detail = None
if isinstance(response, dict):
if "error" in response:
detail = f"{response['error']['message'] if 'message' in response['error'] else response['error']}"
elif isinstance(response, str):
detail = response
raise HTTPException(
status_code=r.status if r else 500,
detail=detail if detail else "Open WebUI: Server Connection Error",
detail="Open WebUI: Server Connection Error",
)
finally:
if not streaming and session:
if r:
r.close()
await session.close()
if not streaming:
await cleanup_response(r, session)
async def embeddings(request: Request, form_data: dict, user):
@ -908,33 +993,31 @@ async def embeddings(request: Request, form_data: dict, user):
models = request.app.state.OPENAI_MODELS
if model_id in models:
idx = models[model_id]["urlIdx"]
url = request.app.state.config.OPENAI_API_BASE_URLS[idx]
key = request.app.state.config.OPENAI_API_KEYS[idx]
api_config = request.app.state.config.OPENAI_API_CONFIGS.get(
str(idx),
request.app.state.config.OPENAI_API_CONFIGS.get(url, {}), # Legacy support
)
r = None
session = None
streaming = False
headers, cookies = await get_headers_and_cookies(
request, url, key, api_config, user=user
)
try:
session = aiohttp.ClientSession(trust_env=True)
r = await session.request(
method="POST",
url=f"{url}/embeddings",
data=body,
headers={
"Authorization": f"Bearer {key}",
"Content-Type": "application/json",
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user
else {}
),
},
headers=headers,
cookies=cookies,
)
r.raise_for_status()
if "text/event-stream" in r.headers.get("Content-Type", ""):
streaming = True
return StreamingResponse(
@ -946,27 +1029,29 @@ async def embeddings(request: Request, form_data: dict, user):
),
)
else:
response_data = await r.json()
try:
response_data = await r.json()
except Exception:
response_data = await r.text()
if r.status >= 400:
if isinstance(response_data, (dict, list)):
return JSONResponse(status_code=r.status, content=response_data)
else:
return PlainTextResponse(
status_code=r.status, content=response_data
)
return response_data
except Exception as e:
log.exception(e)
detail = None
if r is not None:
try:
res = await r.json()
if "error" in res:
detail = f"External: {res['error']['message'] if 'message' in res['error'] else res['error']}"
except Exception:
detail = f"External: {e}"
raise HTTPException(
status_code=r.status if r else 500,
detail=detail if detail else "Open WebUI: Server Connection Error",
detail="Open WebUI: Server Connection Error",
)
finally:
if not streaming and session:
if r:
r.close()
await session.close()
if not streaming:
await cleanup_response(r, session)
@router.api_route("/{path:path}", methods=["GET", "POST", "PUT", "DELETE"])
@ -992,33 +1077,26 @@ async def proxy(path: str, request: Request, user=Depends(get_verified_user)):
streaming = False
try:
headers = {
"Content-Type": "application/json",
**(
{
"X-OpenWebUI-User-Name": quote(user.name),
"X-OpenWebUI-User-Id": quote(user.id),
"X-OpenWebUI-User-Email": quote(user.email),
"X-OpenWebUI-User-Role": quote(user.role),
}
if ENABLE_FORWARD_USER_INFO_HEADERS
else {}
),
}
headers, cookies = await get_headers_and_cookies(
request, url, key, api_config, user=user
)
if api_config.get("azure", False):
headers["api-key"] = key
headers["api-version"] = (
api_config.get("api_version", "") or "2023-03-15-preview"
)
api_version = api_config.get("api_version", "2023-03-15-preview")
# Only set api-key header if not using Azure Entra ID authentication
auth_type = api_config.get("auth_type", "bearer")
if auth_type not in ("azure_ad", "microsoft_entra_id"):
headers["api-key"] = key
headers["api-version"] = api_version
payload = json.loads(body)
url, payload = convert_to_azure_payload(url, payload)
url, payload = convert_to_azure_payload(url, payload, api_version)
body = json.dumps(payload).encode()
request_url = f"{url}/{path}?api-version={api_config.get('api_version', '2023-03-15-preview')}"
request_url = f"{url}/{path}?api-version={api_version}"
else:
headers["Authorization"] = f"Bearer {key}"
request_url = f"{url}/{path}"
session = aiohttp.ClientSession(trust_env=True)
@ -1027,9 +1105,9 @@ async def proxy(path: str, request: Request, user=Depends(get_verified_user)):
url=request_url,
data=body,
headers=headers,
cookies=cookies,
ssl=AIOHTTP_CLIENT_SESSION_SSL,
)
r.raise_for_status()
# Check if response is SSE
if "text/event-stream" in r.headers.get("Content-Type", ""):
@ -1043,27 +1121,27 @@ async def proxy(path: str, request: Request, user=Depends(get_verified_user)):
),
)
else:
response_data = await r.json()
try:
response_data = await r.json()
except Exception:
response_data = await r.text()
if r.status >= 400:
if isinstance(response_data, (dict, list)):
return JSONResponse(status_code=r.status, content=response_data)
else:
return PlainTextResponse(
status_code=r.status, content=response_data
)
return response_data
except Exception as e:
log.exception(e)
detail = None
if r is not None:
try:
res = await r.json()
log.error(res)
if "error" in res:
detail = f"External: {res['error']['message'] if 'message' in res['error'] else res['error']}"
except Exception:
detail = f"External: {e}"
raise HTTPException(
status_code=r.status if r else 500,
detail=detail if detail else "Open WebUI: Server Connection Error",
detail="Open WebUI: Server Connection Error",
)
finally:
if not streaming and session:
if r:
r.close()
await session.close()
if not streaming:
await cleanup_response(r, session)

Some files were not shown because too many files have changed in this diff Show more