0.9.2

[GamerC0der]

Max Mode & Deepseek R1 0528

[vercel]

@GamerC0der is attempting to deploy a commit to the Karo Roghzai's projects Team on Vercel.

A member of the Team first needs to authorize it.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
libre-assistant	Ready	Preview, Comment	Feb 26, 2026 1:05pm

[GamerC0der]

@Mostlime12195

[Mostlime12195]

Hi. I reviewed the PR. You can actually check a preview here, though I'm not sure if you have permissions to see it.

First off, this is a very big PR. Too big to just be a patch. If this were to be merged, it'd be a minor update, not patch (it would be 0.10.0, not 0.9.2).

Second, this fails a lot of checks:

1. The 'Max' model selector doesn't work with mobile. It should properly integrate into the current model selector we have now without overcomplicating or duplicating the code.
2. You should be able to de-select models in the Max mode.
3. The Max toggle probably shouldn't be in the message form, it likely fits better in the top bar.
4. Max mode doesn't work. I mean, it functions, but it generally doesn't work very well. The thinking UI doesn't integrate well into our current UI (aside from not working), we do not get nice interactive visible streaming, and the general functionality doesn't seem to work as intended (refined prompt engineering I suspect may fix that)
5. The UI is a little messed up.
   Example 1 (sidebar):

Example 2 (settings): 6. You changed the image rate limit from 12 to 8 when it should remain 12. I don't get why?

At the moment, I think those are all of the issues you may need to address IF you want to keep this PR as large as it is. I'll happily merge smaller PRs like, say, only adding a couple new models or adding that usage indicator UI (which I really love!)

I really like the ideas in this PR though. It's clear you're vibecoding it, just please make sure to test every main feature of your PR (or clearly address any issues in the PR itself) thoroughly before commiting/making a PR so that we don't run into issues such as the Max mode barely being functional or not working on mobile screens.

[Mostlime12195]

P.S. if you make another commit, please add these models from HCAI:

• liquid/lfm-2-24b-a2b link
• google/gemini-3.1-flash-image-preview link (this one especially is important, I'm sure people'll love getting to play around with the latest and greatest image generation model.)