TensorRT-Model-Connect

TensorRT-Model-Connect turns one model checkpoint into a native TensorRT bundle. Every supported model is a self-contained vertical slice under families/<family>/: Python builder, runtime DSO, bundle sections, and tests.

The shared core resolves one family, writes or reads the bundle container, and transfers control. It does not contain model logic, a model registry, or a runtime-strategy switch.

• Quick start
• Supported models
• AI-native horizontal scaling architecture
• Add a family