Commit graph

419 commits

Author SHA1 Message Date
Timothy Jaeryang Baek
e000494e48 refac 2025-10-07 11:53:30 -05:00
Timothy Jaeryang Baek
b98d8aa8ec refac 2025-10-07 07:31:06 -05:00
Timothy Jaeryang Baek
a2a2bafdf6 enh/refac: url input handling 2025-10-04 02:02:26 -05:00
Timothy Jaeryang Baek
01a5b97415 refac/fix: do not process xlsx files with azure doc intelligence 2025-09-29 23:05:24 -05:00
Timothy Jaeryang Baek
e7fa86aa26 chore: format 2025-09-29 00:58:21 -05:00
Tim Jaeryang Baek
2d94b8e905
Merge pull request #17837 from Classic298/milvus-multitenancy
feat: Impelement Milvus multitenancy // breaking: set milvus multitenancy as standard option (just like Qdrant already is)
2025-09-29 00:29:35 -05:00
Timothy Jaeryang Baek
118549caf3 enh/fix: filter content metadata 2025-09-28 20:17:27 -05:00
Classic298
b1e63639cd
ADD FAT WARNING - QDRANT 2025-09-28 21:17:07 +02:00
Classic298
0e99c43495
ADD FAT WARNING 2025-09-28 21:16:02 +02:00
Classic298
01d4a8ab7a
Update factory.py 2025-09-28 11:06:29 +02:00
Classic298
8dc43f9e3a
Create milvus_multitenancy.py 2025-09-28 11:05:15 +02:00
Tim Jaeryang Baek
f8a3ed2d18
Merge pull request #17770 from Classic298/feat-milvus-diskann-support
feat: Add DISKANN index type support for Milvus
2025-09-26 14:23:53 -05:00
Classic298
9e3d5407ae
Merge branch 'open-webui:main' into feat-milvus-diskann-support 2025-09-26 10:43:01 +02:00
Classic298
b550d78905
Merge branch 'open-webui:main' into fix-milvus-limit-error 2025-09-26 10:42:53 +02:00
google-labs-jules[bot]
123dbf152e feat: Add DISKANN index type support for Milvus
This commit introduces support for the DISKANN index type in the Milvus vector database integration.

Changes include:
- Added `MILVUS_DISKANN_MAX_DEGREE` and `MILVUS_DISKANN_SEARCH_LIST_SIZE` configuration variables.
- Updated the Milvus client to recognize and configure the DISKANN index type during collection creation.
2025-09-26 06:54:06 +00:00
google-labs-jules[bot]
e7ccaf6e78 Fix: milvus error because the limit set to None by default
The pymilvus library expects -1 for unlimited queries, but the code was passing None, which caused a TypeError. This commit changes the default value of the limit parameter in the query method from None to -1. It also updates the call site in the get method to pass -1 instead of None and updates the type hint and a comment to reflect this change.
2025-09-26 06:39:54 +00:00
Timothy Jaeryang Baek
7f411dd5cc feat/enh: perplexity search support 2025-09-25 14:02:46 -05:00
Timothy Jaeryang Baek
fe65fe0b97 refac: ollama cloud web search count support 2025-09-24 15:58:56 -05:00
Timothy Jaeryang Baek
e06489d92b enh: search_ollama_cloud 2025-09-24 15:19:05 -05:00
Timothy Jaeryang Baek
6e4a2f18e1 refac 2025-09-21 00:14:43 -04:00
Timothy Jaeryang Baek
a51f0c30ec refac/fix: knowledge permission 2025-09-15 11:40:31 -05:00
Timothy Jaeryang Baek
e61e7434a0 refac 2025-09-14 10:46:49 +02:00
Timothy Jaeryang Baek
1ef8204359 refac 2025-09-14 10:45:52 +02:00
Timothy Jaeryang Baek
58d7ca35e3 refac 2025-09-14 10:27:07 +02:00
Timothy Jaeryang Baek
aa8ab349ed feat: ref chat 2025-09-14 10:26:46 +02:00
Timothy Jaeryang Baek
210197fd43 refac/fix: web/youtube file attachment handling 2025-09-13 00:02:48 +04:00
Timothy Jaeryang Baek
2185fc61c0 refac 2025-09-11 21:29:56 +04:00
Timothy Jaeryang Baek
485392fe63 chore: format 2025-09-09 18:19:31 +04:00
Tim Jaeryang Baek
71fd483fba
Merge pull request #17276 from Elettrotecnica/extend-docling-configuration
feat: Extend docling configuration options
2025-09-09 18:04:30 +04:00
Timothy Jaeryang Baek
0214c1e66c refac 2025-09-09 16:48:59 +04:00
Timothy Jaeryang Baek
5f0d262c59 fix: yt embed 2025-09-09 16:00:42 +04:00
Antonio Pisano
daa2a036f8 Extend docling configuration options to include:
* do_ocr
* force_ocr
* pdf_backend
* table_mode
* pipeline

as per https://github.com/docling-project/docling-serve/blob/main/docs/usage.md

See https://github.com/open-webui/open-webui/issues/17148
2025-09-08 18:51:33 +02:00
Timothy Jaeryang Baek
4f2e426fc7 refac 2025-09-01 14:27:20 +04:00
Timothy Jaeryang Baek
609a6a3721 refac 2025-09-01 14:22:02 +04:00
Timothy Jaeryang Baek
85153afda8 refac 2025-09-01 14:21:17 +04:00
Timothy Jaeryang Baek
487979859a fix: web/youtube attachements 2025-09-01 01:22:50 +04:00
Timothy Jaeryang Baek
ac0243e8b7 refac 2025-09-01 00:57:13 +04:00
Tim Jaeryang Baek
719d115d49
Merge pull request #17049 from rgaricano/dev-FIX_lex-sem
FIX: Hybrid Search
2025-09-01 00:00:25 +04:00
Tim Jaeryang Baek
4e7b0ea4b4
Merge pull request #17013 from athoik/fix-17000
fix: handle unicode filenames in external document loader
2025-08-31 23:58:52 +04:00
Timothy Jaeryang Baek
c2b4976c82 enh: PGVECTOR_CREATE_EXTENSION env var 2025-08-31 23:58:18 +04:00
_00_
647e38f701
Revert bypass hybrid search when BM25_weight=0
Revert PR https://github.com/open-webui/open-webui/commit/74b1c801
2025-08-30 10:45:35 +02:00
Athanasios Oikonomou
d735b036fe fix: handle unicode filenames in external document loader
Files with special characters in their names (e.g., ü.pdf) caused issues since HTTP headers only allow Latin-1 characters.
This change URL-encodes `X-Filename` before adding it to request headers, preventing failures when uploading or processing such files.

Fixes: #17000
2025-08-28 22:19:50 +03:00
Timothy Jaeryang Baek
2bb6063dcb refac/fix: marker 2025-08-28 03:03:31 +04:00
Timothy Jaeryang Baek
23a9731899 refac/fix: hybrid search 2025-08-26 15:04:46 +04:00
Tim Jaeryang Baek
4267e22d4a
Merge pull request #16826 from selenecodes/feat/azure-document-intelligence-azure-entra-auth
feat: Authenticate Azure Document Intelligence using DefaultAzureCredential
2025-08-26 14:32:04 +04:00
_00_
093af754e7
FIX: Playwright Timeout (ms) interpreted as seconds
Fix for Playwright Timeout (ms) interpreted as seconds.

To address https://github.com/open-webui/open-webui/issues/16801

In Frontend Playwright Timeout is setted as (ms), but in backend is interpreted as (s) doing a time conversion for playwright_timeout var (that have to be in ms).

& as  _Originally posted by @rawbby in [#16801](https://github.com/open-webui/open-webui/issues/16801#issuecomment-3216782565)_

> I personally think milliseconds are a reasonable choice for the timeout. Maybe the conversion should be fixed, not the label.
> This would further not break existing configurations from users that rely on their current config.
>
2025-08-23 14:15:00 +02:00
Selene Blok
5051bfe7ab feat(document retrieval): Authenticate Azure Document Intelligence using AzureDefaultCredential if API key is not provided 2025-08-22 16:15:43 +02:00
Timothy Jaeryang Baek
fbff4e19de fix: reranking 2025-08-22 16:47:05 +04:00
Timothy Jaeryang Baek
60b8cfb9fa refac 2025-08-21 21:48:21 +04:00
Timothy Jaeryang Baek
02479425a5 refac 2025-08-21 12:51:41 +04:00