Ask an existing LLM harness like Claude Code. "Install this and run the demo program" or "Set this up with llama-server" or "Give me an Open WebUI page for this model" will work.