OpenAI compatible endpoint for AI Refine

[ewencodes]

Hi, first of all : this project is sooo useful, i use it everyday! Love it!

The local model for AI Refine (Qwen3.5) seems not as good as like Qwen3.6 27b. I get mistakes, even sometime informations are missing after the refine.

My simple question is : Is it possible to enable other methods rather than only Ollama and local ? It would be great to have the possibility to use OpenAI-compatible endpoints because in my case i host open-weight models behind a litellm proxy.

Thanks!!

[zachswift615]

Hey thanks for the feature request, and glad you're getting good use out of Speak2!

I'll look into this - yeah the default local model sucks tbh. I'll put this on my backlog.