Aura Vision

An ONNX Runtime demo that validates the “AI-native” promise: tensor shapes are part of the type, ONNX model IO shapes are extracted, and Z3 enforces shape equality for inference.

Highlights
  • ai.load_model reads the ONNX contract
  • model.infer(x) is verifier-enforced shape-safe
  • Inference callsites are tagged for AI optimization