API LLM Support / GPU-Free support

[camAtGitHub]

Hey very cool project.
What's needed to be done to replace the Ollama instance with any myriad of LLM API endpoints?
I don't have a workable GPU, but with deepseek-flash being beyond cheap it's very workable cost wise.

Are you able to provide high level direction as to how someone might go adding such functionality?

Thanks again.