Feature Request: Use local LLMs

[reneleonhardt]

Hello, interesting project!

Would it be possible to support local LLMs (i.e. serving with Ollama or LM Studio)?

A developer could label issues first and then run actions locally with https://github.com/nektos/act retrieving all labeled issues, creating and pushing PRs with a local LLM of their choice like Qwen2.5-Coder-32B-Instruct-MLX or Mistral-Small-3.1-24B-Instruct-2503.

Even better would be the possibility to create each PR with multiple local LLMs to give the developer the chance to compare / change / select the best Draft before pushing to GitHub, or even analyze and compare them with automatic tools like linters, but that's music of the future I guess 😅

[HavenDV]

Hi, thank you for the idea

I didn't know about this tool, I think I need to grok it to understand your idea more deeply and implement it.

I also actively use local models, and I think this can be applied

[reneleonhardt]

Cool, thank you for considering it!
It's not without problems on Apple Silicon, but that's more Docker's fault for not implementing cross-platform support 😉