Update

Coordinated concurrent local AI inference jobs

Built an Ollama gateway that queued jobs by priority and model, exposed their lifecycle over HTTP, and reported GPU and model state.