CookXpert chef fine-tune MLC bundle for on-device WebGPU inference.

Weights are mirrored from mlc-ai/SmolLM2-1.7B-Instruct-q4f16_1-MLC until custom LoRA MLC artifacts are uploaded. The app applies CookXpert chef system prompts on top of these weights.

When mlc-chat-config.json and wasm weights in this repo match your build, CookXpert shows Weights live.