I started playing with an alternative ui/ux for llm profiles @breezy-kilobyte-39430. It is a small pass but it makes default LLM more visible with an explanation if it conflicts with agent profile, adds provider connections as a collapsable group, makes bulk add relative to the provider connection instead of a standalone button. What do you think?

I feel this is a bit difficult problem... Still a lot of nouns per inch, so to speak 😅 Like this, I know why "models as profiles", but idk

Yea there is a lot to cover in one place. I made an issue to try and get some feedback and included a few of the other mockups I generated github.com/OpenHands/OpenHands/issues/18886

Thank you. I don’t want to promise right now because I’m very caught, but I want to say this: it’s great that we’re thinking about this and I really appreciate it. I also think that as long as we’re moving fast, we can try things and see what people feel/report/ask. Everything on those UI pages should be, I think, easy to rip out and rebuild. 🙏 I want to “officially” propose that we use feature flags more, and try more stuff under it, and we can move faster that way. I’ll get to that in its own topic. 😅

Thoughts on a single beta flag users can toggle vs env or startup flags to enable a feature?

Ah, yes, that makes sense to me, too; and “graduate” any of them