Local AI toolkit that runs on any hardware \u2014 llama.cpp + ONNX + TensorRT-LLM engines with unified OpenAI API.