Routes all on-device ML and speech-to-text work on iOS through a single entry point. Handles CoreML model conversion from PyTorch and TensorFlow, model compression techniques like quantization and pruning, and audio transcription. The mandatory routing approach is opinionated but makes sense if you're working across multiple ML tasks and want consistent patterns. Draws a clear boundary with ios-ai (cloud-based inference) to keep concerns separated. From the axiom collection by charleswiltgen, which has decent traction at 945 stars and 181 installs.
npx skills add https://github.com/charleswiltgen/axiom --skill axiom-ios-ml