WebUI "edit assistant messages"

[3Simplex]

@ngxson
The ability to edit the assistant messages would greatly enhance the experience.

• I have been using this method to steer future responses.
• Users can "fix" mistakes generated by assistants.

Editing any previous message from "user or assistant" without required regeneration.

• I have used this method to refine the context resolving misunderstanding.

[ngxson]

Yes it should be possible to add now, after the addition of "branching" conversation support. I will add it in very near future.

[3Simplex]

Yeah, I love that branching feature!

[sidran]

Yes it should be possible to add now, after the addition of "branching" conversation support. I will add it in very near future.

@ngxson
Has there been any progress with unhindered, complete editing?

While we're on the topic, could we also explore a few impactful improvements to webUI? I’m not suggesting turning it into a tangled mess like Open webUI, just small but meaningful changes:

1. WebUI should apply its default model parameters only if it doesn’t receive any from Llama-server. For example, if we start Llama-server with --temp 0.2, webUI should detect and respect such parameters, reflecting them in its settings. If nothing is received, it can fall back to its current defaults. This makes frequent model switching far smoother.
2. Allow renaming conversations on the left side instead of clipping the first few words as titles.
3. Enable the creation and easy switching of different prompts. This would eliminate the need to store prompts elsewhere and constantly copy/paste. Separating "system instructions" from the prompts themselves could also improve clarity.

WebUI is already beautiful, minimalistic, and highly functional, but these additions would significantly enhance usability.

Would it be possible to implement these improvements?

[sidran]

Anyone?