Ask HN: What are you using to host local LLM models that are accessible via API?

I quickly used Oobabooga months ago, but never got llama.cpp working. What are people using for this? I will need to fine tune, and as mentioned in the title, I want to host it on a computer and access it via API. Any thoughts to get me closer is appreciated.