Developer hshatti has released TONNXRuntime, a native Delphi wrapper for Microsoft ONNX Runtime that brings high-performance machine learning inference, local AI execution, and hardware-accelerated neural networks directly into Delphi applications.
TONNXRuntime enables Delphi developers to load and execute ONNX models using CPU, CUDA, DirectML, and TensorRT backends without requiring Python runtimes or external AI services. The library is designed for deploying production-grade local AI workflows directly inside native Delphi applications.
The framework includes support for:
- ONNX Runtime integration for Delphi
- Local AI and offline inference
- GPU acceleration with CUDA and DirectML
- TensorRT execution providers
- Image classification and computer vision
- Native tensor processing
- High-performance inference pipelines
The project includes support for Windows, Linux, and macOS deployments through ONNX Runtime while exposing a native Object Pascal interface for integrating modern machine learning models into Delphi applications.