feat: Full context mode (RAG) per-user, per-chat #4719

New Issue

GiteaMirror · 2025-11-11T16:01:24-06:00

GiteaMirror commented

2025-11-11 16:01:24 -06:00

Originally created by @frenzybiscuit on GitHub (Apr 4, 2025).

Check Existing Issues

I have searched the existing issues and discussions.

Problem Description

Please consider adding an option for individual users to use full context mode when selecting a document in the knowledge base. Specifically, when using the chat interface and using # function.

Perhaps add in a button on the file that shows up to enable full context mode.

I'm aware it can be enabled globally in the admin panel, however I feel like it should be an option in the WebUI for users. There are use cases where you want full context and where you want to use RAG.

In addition, if it is per-file you could load one file with full context in the chat while using RAG on the other file, which may be handy when one file will fit into the context of the model but the other one is too large to.

I can confirm I do not see any open issues about this matter. There may be some closed issues (not sure), but I'd like to bring this up for discussion again if there are.

Desired Solution you'd like

^ above

Alternatives Considered

No response

Additional Context

No response

Originally created by @frenzybiscuit on GitHub (Apr 4, 2025). ### Check Existing Issues - [x] I have searched the existing issues and discussions. ### Problem Description Please consider adding an option for individual users to use full context mode when selecting a document in the knowledge base. Specifically, when using the chat interface and using # function. Perhaps add in a button on the file that shows up to enable full context mode. I'm aware it can be enabled globally in the admin panel, however I feel like it should be an option in the WebUI for users. There are use cases where you want full context and where you want to use RAG. In addition, if it is per-file you could load one file with full context in the chat while using RAG on the other file, which may be handy when one file will fit into the context of the model but the other one is too large to. I can confirm I do not see any open issues about this matter. There may be some closed issues (not sure), but I'd like to bring this up for discussion again if there are. ### Desired Solution you'd like ^ above ### Alternatives Considered _No response_ ### Additional Context _No response_

GiteaMirror commented

2025-11-11 16:01:24 -06:00

@carpmike-oa commented on GitHub (Apr 25, 2025):

Agreed. Or allow this to be configured per knowledge instead of globally, which would mean some knowledge is RAG and some is full context. This would be particularly helpful for building models for different purposes.

@carpmike-oa commented on GitHub (Apr 25, 2025): Agreed. Or allow this to be configured per knowledge instead of globally, which would mean some knowledge is RAG and some is full context. This would be particularly helpful for building models for different purposes.

GiteaMirror commented

2025-11-11 16:01:25 -06:00

@flefevre commented on GitHub (Jun 3, 2025):

I do not know which final solution (button, #, xxx) is better but allowing at user level to switch from pure rag to full context mode is a good idea

@flefevre commented on GitHub (Jun 3, 2025): I do not know which final solution (button, #, xxx) is better but allowing at user level to switch from pure rag to full context mode is a good idea

GiteaMirror commented

2025-11-11 16:01:25 -06:00

@dojoca commented on GitHub (Aug 26, 2025):

Adding some support for this... would be great to do this.

@dojoca commented on GitHub (Aug 26, 2025): Adding some support for this... would be great to do this.

GiteaMirror referenced this issue

2025-11-11 17:51:30 -06:00

[PR #4719] [MERGED] chore(deps): bump faster-whisper from 1.0.2 to 1.0.3 in /backend #8344

Sign in to join this conversation.