coreai-onnx

coreai-onnx is the missing ONNX path into Apple’s Core AI stack. It converts ONNX models — exported from PyTorch, TensorFlow, scikit-learn, or any other ONNX-compatible framework — directly into the .aimodel format consumed by Core AI, enabling on-device inference on Apple silicon without any manual model surgery.

Contents

• Quickstart
• CLI Reference
• coreai-onnx — Agent Guide
• MCP server
• Supported op lowerings
• Custom Lowerings
• Automatic repair