llama.cpp

Use OpenAI-compatible `/v1/models` endpoint by default

#17689

Merged