Really wish both completion and chat could be run against the same LLM end point. Most of these seem to be doing their own incompatible docker thing which is annoying