Conda environment with Open WebUI (main branch, v0.4.8).
Issue is reproducible in Docker as well.
Environment
Open WebUI Version: v0.4.8 (main branch, 2024-12-07)
Operating System: macOS 15.2
Browser (if applicable): Chrome 120.0
Confirmation:
I have read and followed all the instructions provided in the README.md.
I am on the latest version of Open WebUI.
I have included the browser console logs.
I have included the Docker container logs.
I have provided the exact steps to reproduce the bug in the "Steps to Reproduce" section below.
Expected Behavior:
The application should remain responsive while processing a request to the /completions endpoint with ENABLE_RAG_HYBRID_SEARCH enabled.
Actual Behavior:
The application freezes during the execution of the /completions endpoint when ENABLE_RAG_HYBRID_SEARCH is enabled.
The UI becomes unresponsive and no further interactions can be made.
Once the /completions request completes, the app becomes functional again.
Description
Bug Summary:
When ENABLE_RAG_HYBRID_SEARCH is turned on and the reranking modelBAAI/bge-reranker-v2-m3 is used, calling the /completions endpoint causes the application to freeze until the request finishes processing. This issue is consistent and reproducible.
Reproduction Details
Steps to Reproduce:
Install Open WebUI using Conda or Docker.
Enable ENABLE_RAG_HYBRID_SEARCH in the settings.
Set the reranking model to BAAI/bge-reranker-v2-m3.
Trigger a request to the /completions endpoint (e.g., perform a query using HybridSearch).
Observe that:
The application freezes during the request execution.
No interactions can be performed until the /completions endpoint finishes processing.
After the request completes, the app returns to normal.
The issue consistently occurs during /completions calls with HybridSearch enabled.
The reranking model in use is BAAI/bge-reranker-v2-m3.
CPU or memory usage appears normal during the freeze.
This behavior happens both in Conda and Docker installations.
Note
Please let me know if further testing or logs are required to resolve this issue. Thank you for maintaining Open WebUI!
Originally created by @gabriel-ecegi on GitHub (Dec 17, 2024).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/7913
# Bug Report
---
## Installation Method
Conda environment with Open WebUI (`main` branch, v0.4.8).
Issue is reproducible in Docker as well.
## Environment
- **Open WebUI Version:** v0.4.8 (main branch, 2024-12-07)
- **Operating System:** macOS 15.2
- **Browser (if applicable):** Chrome 120.0
**Confirmation:**
- [x] I have read and followed all the instructions provided in the README.md.
- [x] I am on the latest version of Open WebUI.
- [x] I have included the browser console logs.
- [x] I have included the Docker container logs.
- [x] I have provided the exact steps to reproduce the bug in the "Steps to Reproduce" section below.
---
## Expected Behavior:
The application should remain responsive while processing a request to the `/completions` endpoint with `ENABLE_RAG_HYBRID_SEARCH` enabled.
---
## Actual Behavior:
The application freezes during the execution of the `/completions` endpoint when `ENABLE_RAG_HYBRID_SEARCH` is enabled.
- The UI becomes unresponsive and no further interactions can be made.
- Once the `/completions` request completes, the app becomes functional again.
---
## Description
**Bug Summary:**
When `ENABLE_RAG_HYBRID_SEARCH` is turned on and the **reranking model** `BAAI/bge-reranker-v2-m3` is used, calling the `/completions` endpoint causes the application to freeze until the request finishes processing. This issue is consistent and reproducible.
---
## Reproduction Details
**Steps to Reproduce:**
1. Install Open WebUI using Conda or Docker.
2. Enable `ENABLE_RAG_HYBRID_SEARCH` in the settings.
3. Set the reranking model to `BAAI/bge-reranker-v2-m3`.
4. Trigger a request to the `/completions` endpoint (e.g., perform a query using HybridSearch).
5. Observe that:
- The application freezes during the request execution.
- No interactions can be performed until the `/completions` endpoint finishes processing.
6. After the request completes, the app returns to normal.
---
## Logs and Screenshots
**Browser Console Logs:**
N/A (no relevant logs).
**Docker Container Logs:**
N/A (no relevant logs).
**Screenshots/Screen Recordings (if applicable):**
N/A
---
## Additional Information
- The issue consistently occurs during `/completions` calls with HybridSearch enabled.
- The reranking model in use is `BAAI/bge-reranker-v2-m3`.
- CPU or memory usage appears normal during the freeze.
- This behavior happens both in Conda and Docker installations.
---
## Note
Please let me know if further testing or logs are required to resolve this issue. Thank you for maintaining Open WebUI!
What Embedding Model did you set? For information, I use the BAAI/bge-reranker-v2-m3 in hybrid search without problems. It just takes a while to get the answer, but CPU work is high at the time and the UI is responsive. (I also have top-k at 15)
<!-- gh-comment-id:2551340174 -->
@mahenning commented on GitHub (Dec 18, 2024):
What `Embedding Model` did you set? For information, I use the `BAAI/bge-reranker-v2-m3` in hybrid search without problems. It just takes a while to get the answer, but CPU work is high at the time and the UI is responsive. (I also have top-k at 15)
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Originally created by @gabriel-ecegi on GitHub (Dec 17, 2024).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/7913
Bug Report
Installation Method
Conda environment with Open WebUI (
mainbranch, v0.4.8).Issue is reproducible in Docker as well.
Environment
Confirmation:
Expected Behavior:
The application should remain responsive while processing a request to the
/completionsendpoint withENABLE_RAG_HYBRID_SEARCHenabled.Actual Behavior:
The application freezes during the execution of the
/completionsendpoint whenENABLE_RAG_HYBRID_SEARCHis enabled./completionsrequest completes, the app becomes functional again.Description
Bug Summary:
When
ENABLE_RAG_HYBRID_SEARCHis turned on and the reranking modelBAAI/bge-reranker-v2-m3is used, calling the/completionsendpoint causes the application to freeze until the request finishes processing. This issue is consistent and reproducible.Reproduction Details
Steps to Reproduce:
ENABLE_RAG_HYBRID_SEARCHin the settings.BAAI/bge-reranker-v2-m3./completionsendpoint (e.g., perform a query using HybridSearch)./completionsendpoint finishes processing.Logs and Screenshots
Browser Console Logs:
N/A (no relevant logs).
Docker Container Logs:
N/A (no relevant logs).
Screenshots/Screen Recordings (if applicable):
N/A
Additional Information
/completionscalls with HybridSearch enabled.BAAI/bge-reranker-v2-m3.Note
Please let me know if further testing or logs are required to resolve this issue. Thank you for maintaining Open WebUI!
@mahenning commented on GitHub (Dec 18, 2024):
What
Embedding Modeldid you set? For information, I use theBAAI/bge-reranker-v2-m3in hybrid search without problems. It just takes a while to get the answer, but CPU work is high at the time and the UI is responsive. (I also have top-k at 15)@gaboe commented on GitHub (Dec 18, 2024):
I use BAAI/bge-reranker-v2-m3, also.
@tjbck commented on GitHub (Dec 19, 2024):
Closing in favour of #7942