Mintplex-Labs/anything-llm
1
0
Fork 0
mirror of https://github.com/Mintplex-Labs/anything-llm.git synced 2026-07-19 14:13:35 -04:00
Code Issues 328 Packages Projects Releases 10 Wiki Activity

[GH-ISSUE #4593] [BUG]: Changing Vector Database Setting to "Accuracy Optimized" causes Connection error on second prompt #2920

New Issue
Closed
opened 2026-02-22 18:31:51 -05:00 by yindo · 5 comments
yindo commented 2026-02-22 18:31:51 -05:00
Owner
Copy Link
Copy Source

Originally created by @aghoras on GitHub (Oct 29, 2025).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4593

How are you running AnythingLLM?

AnythingLLM desktop app

What happened?

If I change my workspace Vector Database setting to Accuracy Optimized, AnythingLLM gives me an error message saying "Could not respond to message. An error occurred while streaming response. network error"

After that, the whole Desktop Application becomes unresponsive. Running version 1.9.0-r2.

I'm using LanceDB with using the AnythingLLM embedder "all-MiniLM-L6-v2". Chuck size = 250. Overlap 20
Also using LocalAI LLM.

Everything works fine and I get meaningful results if I change the Search Preference to default.

Are there known steps to reproduce?

  • Create embedded documents in workspace
  • Change Search Preference on workspace under Vector Database to "Accuracy Optimized"
  • Send two prompts (could be the same) to the LLM.
  • Network error shows on after the second prompt and the program becomes unresponsive (i.e. spinning indicator)
Originally created by @aghoras on GitHub (Oct 29, 2025). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4593 ### How are you running AnythingLLM? AnythingLLM desktop app ### What happened? If I change my workspace Vector Database setting to Accuracy Optimized, AnythingLLM gives me an error message saying _"Could not respond to message. An error occurred while streaming response. network error"_ After that, the whole Desktop Application becomes unresponsive. Running version 1.9.0-r2. I'm using LanceDB with using the AnythingLLM embedder "all-MiniLM-L6-v2". Chuck size = 250. Overlap 20 Also using LocalAI LLM. Everything works fine and I get meaningful results if I change the Search Preference to default. ### Are there known steps to reproduce? - Create embedded documents in workspace - Change Search Preference on workspace under Vector Database to "Accuracy Optimized" - Send two prompts (could be the same) to the LLM. - Network error shows on after the second prompt and the program becomes unresponsive (i.e. spinning indicator)
yindo added the possible bugneeds info / can't replicateinvestigating labels 2026-02-22 18:31:51 -05:00
yindo closed this issue 2026-02-22 18:31:51 -05:00
yindo commented 2026-02-22 18:31:51 -05:00
Author
Owner
Copy Link
Copy Source

@jpwoods92 commented on GitHub (Nov 14, 2025):

I am also seeing this issue

@jpwoods92 commented on GitHub (Nov 14, 2025): I am also seeing this issue
yindo commented 2026-02-22 18:31:52 -05:00
Author
Owner
Copy Link
Copy Source

@timothycarambat commented on GitHub (Nov 21, 2025):

The frontend error is not useful enough here - we will need the logs from the backend. You can find them here in the logs folder.

Without that, there isn't more we can do since I cannot replicate this on my machine(s).

Some other questions that might help more:

  • If you open the workspace settings > Vector Database - How many vectors are in there?
  • What are the specs of the machine you are on? Reranking can be quite intensive on a large corpus
@timothycarambat commented on GitHub (Nov 21, 2025): The frontend error is not useful enough here - we will need the logs from the backend. You can find them [here](https://docs.anythingllm.com/installation-desktop/storage#where-is-my-data-located) in the logs folder. Without that, there isn't more we can do since I cannot replicate this on my machine(s). Some other questions that might help more: - If you open the workspace settings > Vector Database - How many vectors are in there? - What are the specs of the machine you are on? Reranking can be quite intensive on a large corpus
yindo commented 2026-02-22 18:31:52 -05:00
Author
Owner
Copy Link
Copy Source

@aghoras commented on GitHub (Dec 2, 2025):

Hi Timothy,

  • I have attached the backend and collector logs
  • The vector count is 4265
  • I'm running on a DXG A100 Station 80G
  • It's worth mentioning I'm running in Query Mode and not Chat Mode
  • I created a fresh configuration folder and repeated the above steps (after initial setup)

Thanks

backend-2025-12-02.log
collector-2025-12-02.log

@aghoras commented on GitHub (Dec 2, 2025): Hi Timothy, - I have attached the backend and collector logs - The vector count is 4265 - I'm running on a DXG A100 Station 80G - It's worth mentioning I'm running in Query Mode and not Chat Mode - I created a fresh configuration folder and repeated the above steps (after initial setup) Thanks [backend-2025-12-02.log](https://github.com/user-attachments/files/23888971/backend-2025-12-02.log) [collector-2025-12-02.log](https://github.com/user-attachments/files/23888970/collector-2025-12-02.log)
yindo changed title from [BUG]: Changing Vector Database Setting to "Accuracy Optimized" causes Connection error on second prompt to [GH-ISSUE #4593] [BUG]: Changing Vector Database Setting to "Accuracy Optimized" causes Connection error on second prompt 2026-06-05 14:49:17 -04:00
yindo commented 2026-06-05 14:49:18 -04:00
Author
Owner
Copy Link
Copy Source

@bunopus commented on GitHub (Feb 24, 2026):

Doubling that, same issue, same reproduction steps

<!-- gh-comment-id:3953915366 --> @bunopus commented on GitHub (Feb 24, 2026): Doubling that, same issue, same reproduction steps
yindo commented 2026-06-05 14:49:18 -04:00
Author
Owner
Copy Link
Copy Source

@timothycarambat commented on GitHub (Feb 25, 2026):

"level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Initialized","service":"backend"}
{"level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Loading reranker suite...","service":"backend"}
{"level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Loaded model Xenova/ms-marco-MiniLM-L-6-v2","service":"backend"}
{"level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Loaded tokenizer Xenova/ms-marco-MiniLM-L-6-v2","service":"backend"}
{"level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Reranking 50 documents...","service":"backend"}
{"level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Reranking 50 documents to top 4 took 1300ms","service":"backend"}

So this run works, but then

{"level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Initialized","service":"backend"}
{"level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Reranker suite already fully initialized - reusing.","service":"backend"}

Seems to be reusing the module so it does not need to be reload, but nothing happens after this - which seems like it is hanging or crashing. Are you running the LLM as well on the same exact machine? Wondering is there is some memory thing going on where it is kicking out the mlocked reranking model and so when it tries to run again it just hangs

<!-- gh-comment-id:3960343429 --> @timothycarambat commented on GitHub (Feb 25, 2026): ``` "level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Initialized","service":"backend"} {"level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Loading reranker suite...","service":"backend"} {"level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Loaded model Xenova/ms-marco-MiniLM-L-6-v2","service":"backend"} {"level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Loaded tokenizer Xenova/ms-marco-MiniLM-L-6-v2","service":"backend"} {"level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Reranking 50 documents...","service":"backend"} {"level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Reranking 50 documents to top 4 took 1300ms","service":"backend"} ``` So this run works, but then ``` {"level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Initialized","service":"backend"} {"level":"info","message":"\u001b[36m[NativeEmbeddingReranker]\u001b[0m Reranker suite already fully initialized - reusing.","service":"backend"} ``` Seems to be reusing the module so it does not need to be reload, but nothing happens after this - which seems like it is hanging or crashing. Are you running the LLM as well on the same exact machine? Wondering is there is some memory thing going on where it is kicking out the mlocked reranking model and so when it tries to run again it just hangs
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
Desktop
Docker
Integration Request
Integration Request
OS: Linux
OS: Mobile
OS: Windows
UI/UX
blocked
bug
bug
core-team-only
documentation
duplicate
embed-widget
enhancement
feature request
github_actions
good first issue
investigating
needs info / can't replicate
possible bug
pull-request
Mirrored from GitHub Pull Request
 question
stage: specifications
wontfix
No Label investigating needs info / can't replicate possible bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Mintplex-Labs/anything-llm#2920
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?