Auto-detect client hardware capabilities and match against in-browser AI models across WebGPU, WASM, and ONNX Runtime.GH: https://github.com/hemanth/browser-llm-fit