[GH-ISSUE #1859] [BUG]: In Query mode, it will utilize a LLM to generate answers. #1212

Closed
opened 2026-02-22 18:23:43 -05:00 by yindo · 2 comments
Owner

Originally created by @dkasa on GitHub (Jul 11, 2024).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/1859

How are you running AnythingLLM?

Docker (local)

What happened?

When I use the query mode and ask a question that is not within the search scope of the knowledge base, AnythingLLM responds that it cannot find the answer.
However, after I ask a question that can be searched and found, the mode changes to chat mode, and for the previous questions that couldn't be found, AnythingLLM starts to use the LLM to provide answers instead of staying in query mode.
Is this a bug?

Are there known steps to reproduce?

No response

Originally created by @dkasa on GitHub (Jul 11, 2024). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/1859 ### How are you running AnythingLLM? Docker (local) ### What happened? When I use the query mode and ask a question that is not within the search scope of the knowledge base, AnythingLLM responds that it cannot find the answer. However, after I ask a question that can be searched and found, the mode changes to chat mode, and for the previous questions that couldn't be found, AnythingLLM starts to use the LLM to provide answers instead of staying in query mode. Is this a bug? ### Are there known steps to reproduce? _No response_
yindo added the possible bug label 2026-02-22 18:23:43 -05:00
yindo closed this issue 2026-02-22 18:23:43 -05:00
Author
Owner

@timothycarambat commented on GitHub (Jul 12, 2024):

Since documents/references were previously found in the chat thread those references are used even if the last question sent has no references. It used to work the way you described but that is not longer the case because it limited the user to basically needed to know exactly what existed in the vector db or else they would get a refusal. Even followup questions were refused and it leads to a bad experience

previous state of query:
"Tell me about XYZ"
"Sure, here is information with citations..."
"What about Z?"
-> LLM Refuses since the previous prompt has no result.

currenty state of query:
"Tell me about XYZ"
"Sure, here is information with citations..."
"What about Z?"
"Sure, here is more about Z i can infer" (using the citations from the first answer)

@timothycarambat commented on GitHub (Jul 12, 2024): Since documents/references were previously found in the chat thread _those_ references are used even if the last question sent has no references. It used to work the way you described but that is not longer the case because it limited the user to basically needed to know exactly what existed in the vector db or else they would get a refusal. Even followup questions were refused and it leads to a bad experience previous state of `query`: "Tell me about XYZ" "Sure, here is information with citations..." "What about Z?" -> LLM Refuses since the previous prompt has no result. currenty state of `query`: "Tell me about XYZ" "Sure, here is information with citations..." "What about Z?" "Sure, here is more about Z i can infer" (using the citations from the first answer)
Author
Owner

@dkasa commented on GitHub (Jul 15, 2024):

i see ...

@dkasa commented on GitHub (Jul 15, 2024): i see ...
yindo changed title from [BUG]: In Query mode, it will utilize a LLM to generate answers. to [GH-ISSUE #1859] [BUG]: In Query mode, it will utilize a LLM to generate answers. 2026-06-05 14:39:32 -04:00
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Mintplex-Labs/anything-llm#1212