~950 line, minimal, extensible LLM inference engine built from scratch.
By chatting or signing in you agree to the Terms and chat-message logging (revocable in History).