We are actively working on officially supporting this. Because it's open source it is pretty easy to point a coding agent at it now and switch out the model.
Not yet, unfortunately. We only just opened the repo, and there isn't a public issue specifically tracking local or OpenAI-compatible endpoint support.