Comment by car

Please make this more generic, using the OpenAI API standard with model discovery. That way it works against a bare Llama.cpp instance out of the box, the local inference server that many others wrap around (like Ollama and LM Studio).