[PR #17049] [MERGED] FIX: Hybrid Search #24305

New Issue

GiteaMirror · 2026-04-20T05:20:02-05:00

GiteaMirror commented

2026-04-20 05:20:02 -05:00

📋 Pull Request Information

Original PR: https://github.com/open-webui/open-webui/pull/17049
Author: @rgaricano
Created: 8/29/2025
Status: ✅ Merged
Merged: 8/31/2025
Merged by: @tjbck

Base: dev ← Head: dev-FIX_lex-sem

📝 Commits (2)

292cb62 FIX: Hybrid Search lexical-semantic tags
647e38f Revert bypass hybrid search when BM25_weight=0

📊 Changes

2 files changed (+4 additions, -9 deletions)

View changed files

📝 backend/open_webui/retrieval/utils.py (+2 -7)
📝 src/lib/components/admin/Settings/Documents.svelte (+2 -2)

📄 Description

FIX Error in Hybrid Search

Fix 2 errors in hibryd Search:

lexical-semantic terms are inverted
BM25 weight=1 --> lexical
BM25 weight=0 --> semantic
Errors when Hybrid Search when bm25_weight=0.
As noted in https://github.com/open-webui/open-webui/issues/17046 & https://github.com/open-webui/open-webui/discussions/16957 errors arise in this condition due to return empty list instead of fetching the actual collection data.

I opted for revert the bypass when bm25_weight=0 done in PR https://github.com/open-webui/open-webui/commit/74b1c801 because this proposal sought to save calls if bm25 weight=0, which makes sense, since that setting practically implies the same thing as disabling hybrid search.
However, it was not considered that running hybrid search with bm25 weight=0 can also be used to perform a new reranking with different cutoff values (both in terms of the number of results and the relevance threshold).

Contributor License Agreement

By submitting this pull request, I confirm that I have read and fully agree to the Contributor License Agreement (CLA), and I am providing my contributions under its terms.

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/open-webui/open-webui/pull/17049 **Author:** [@rgaricano](https://github.com/rgaricano) **Created:** 8/29/2025 **Status:** ✅ Merged **Merged:** 8/31/2025 **Merged by:** [@tjbck](https://github.com/tjbck) **Base:** `dev` ← **Head:** `dev-FIX_lex-sem` --- ### 📝 Commits (2) - [`292cb62`](https://github.com/open-webui/open-webui/commit/292cb62d4af2ed34e11c866c8d45e1a25ef2057d) FIX: Hybrid Search lexical-semantic tags - [`647e38f`](https://github.com/open-webui/open-webui/commit/647e38f701ff93bb40ed71ba445a8ba903518306) Revert bypass hybrid search when BM25_weight=0 ### 📊 Changes **2 files changed** (+4 additions, -9 deletions) <details> <summary>View changed files</summary> 📝 `backend/open_webui/retrieval/utils.py` (+2 -7) 📝 `src/lib/components/admin/Settings/Documents.svelte` (+2 -2) </details> ### 📄 Description ### FIX Error in Hybrid Search Fix 2 errors in hibryd Search: - lexical-semantic terms are inverted BM25 weight=1 --> lexical BM25 weight=0 --> semantic - Errors when Hybrid Search when bm25_weight=0. As noted in https://github.com/open-webui/open-webui/issues/17046 & https://github.com/open-webui/open-webui/discussions/16957 errors arise in this condition due to return empty list instead of fetching the actual collection data. I opted for revert the bypass when bm25_weight=0 done in PR https://github.com/open-webui/open-webui/commit/74b1c801 because this proposal sought to save calls if bm25 weight=0, which makes sense, since that setting practically implies the same thing as disabling hybrid search. However, it was not considered that running hybrid search with bm25 weight=0 can also be used to perform a new reranking with different cutoff values (both in terms of the number of results and the relevance threshold). -------------------- ### 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. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

GiteaMirror added the pull-request label 2026-04-20 05:20:02 -05:00

GiteaMirror closed this issue

2026-04-20 05:20:03 -05:00

GiteaMirror referenced this issue

2026-05-06 12:49:07 -05:00

[PR #24305] [CLOSED] fix: merge filter tools with internal tools instead of replacing #66450

GiteaMirror referenced this issue

2026-05-13 16:59:37 -05:00

[PR #24305] [CLOSED] fix: merge filter tools with internal tools instead of replacing #82467

GiteaMirror referenced this issue

2026-05-16 01:27:02 -05:00

[PR #24305] [CLOSED] fix: merge filter tools with internal tools instead of replacing #98632