Mintplex-Labs/anything-llm
1
0
Fork 0
mirror of https://github.com/Mintplex-Labs/anything-llm.git synced 2026-07-21 01:25:28 -04:00
Code Issues 328 Packages Projects Releases 10 Wiki Activity

[GH-ISSUE #5702] [FEAT]: Compare LLM responses / conversation branching #5244

New Issue
Open
opened 2026-06-05 14:52:52 -04:00 by yindo · 1 comment
yindo commented 2026-06-05 14:52:52 -04:00
Owner
Copy Link
Copy Source

Originally created by @tredondo on GitHub (May 25, 2026).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/5702

What would you like to see?

I'm looking for a way to compare side-by-side responses from different models (or even different generations for the same model). This would be very useful especially in the initial exploration phase, when different responses can uncover directions of research that a single response would've missed. At that point, the user may choose to continue the chat with one of the responses, or ideally in parallel with a subset of the responses.

Prior art

Bonus feature: the ability to branch the conversation at any point, vs. only in the beginning.

Originally created by @tredondo on GitHub (May 25, 2026). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/5702 ### What would you like to see? I'm looking for a way to compare side-by-side responses from different models (or even different generations for the same model). This would be very useful especially in the initial exploration phase, when different responses can uncover directions of research that a single response would've missed. At that point, the user may choose to continue the chat with one of the responses, or ideally in parallel with a subset of the responses. ## Prior art - [Beam in big-AGI](https://big-agi.com/blog/beam-multi-model-ai-reasoning) ([screenshot](https://github.com/user-attachments/assets/64f151f8-7fa8-4643-b87d-1edd0a45594f)) - [Split chats in Msty.AI](https://msty.ai/studio/features#split-chats) (though the UI seems to expose the chats in tabs, rather than side-by-side) - [BrowserOS LLM Hub](https://docs.browseros.com/features/llm-chat-hub#llm-hub-clash-of-gpts) Bonus feature: the ability to branch the conversation at any point, vs. only in the beginning.
yindo added the enhancementfeature request labels 2026-06-05 14:52:52 -04:00
yindo commented 2026-06-05 14:52:53 -04:00
Author
Owner
Copy Link
Copy Source

@timothycarambat commented on GitHub (May 27, 2026):

RE: Branching - this exists it is simply in the submenu because it isnt really a "quick tool" since its quite a deliberate action by the user. You should see this under every reply

Image

This will branch the chat from that point into a new thread under the same workspace.

RE: Model vs Model output comparison

So this is something we have closed on before (cant find issue but i know we have talked about it) and the reasoning for why ultimately we dont support it is roughly the following:

We consider that a "playground" feature - if that makes sense? Comparing the outputs of models to try for the best output is more of an evaluation action whereas that isnt what AnythingLLM's core value is. We are focused on the usability, access, not comparing the best models for a given response. This is a better form factor for tools like LMStudio, Msty, or Jan which are more about model runtime optimization not so much the integration experience - even though we share similar UX in many cases

Now outside of that specific UX, there is the "Regenerate" chat action - which reruns the chat but fully overwrites the previous output instead of making a "carousel" to toggle through the answers. I think that is something we could improve.

For the idea of exploration there is the concept of generating followup questions to quickly tap to send, the tradeoff for this is similar to why we don't have LLM-based thread naming - it spends cycles locally running inference that could better be served for main chats! Now all that being said there is something to discuss about the UX of how we would even accomplish the exploration part of this issue, but might help to understand at least why we dont do the model v model stuff at the very least.

<!-- gh-comment-id:4556407322 --> @timothycarambat commented on GitHub (May 27, 2026): RE: Branching - this exists it is simply in the submenu because it isnt really a "quick tool" since its quite a deliberate action by the user. You should see this under every reply <img width="869" height="210" alt="Image" src="https://github.com/user-attachments/assets/222c8ac7-e604-457c-b59a-6f0a2842e485" /> This will branch the chat from that point into a new thread under the same workspace. --- RE: Model vs Model output comparison So this is something we have closed on before (cant find issue but i know we have talked about it) and the reasoning for why ultimately we dont support it is roughly the following: We consider that a "playground" feature - if that makes sense? Comparing the outputs of models to try for the best output is more of an evaluation action whereas that **isnt** what AnythingLLM's core value is. We are focused on the usability, access, not comparing the best models for a given response. This is a better form factor for tools like LMStudio, Msty, or Jan which are more about model runtime optimization not so much the integration experience - even though we share similar UX in many cases Now outside of that specific UX, there is the "Regenerate" chat action - which reruns the chat but fully overwrites the previous output instead of making a "carousel" to toggle through the answers. I think that is something we could improve. For the idea of exploration there is the concept of generating followup questions to quickly tap to send, the tradeoff for this is similar to why we don't have LLM-based thread naming - it spends cycles locally running inference that could better be served for main chats! Now all that being said there is something to discuss about the UX of how we would even accomplish the exploration part of this issue, but might help to understand at least why we dont do the model v model stuff at the very least.
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 enhancement feature request
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#5244
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?