libs/inference/README.md

inference

Swappable local inference backends (llama_cpp default, transformers optional extra).

from inference.factory import get_backend

backend = get_backend()
backend.load()
reply = backend.chat([{"role": "user", "content": "Hello!"}])