Run ONNX models on Linux.

Know what the runtime is doing.

LinML began as a Trackdub engineering need: our take on bringing the WinML developer experience to Linux. It is an independent implementation on ONNX Runtime, not a Microsoft port. The project separates shared contracts (`LinML.Contracts`), selection rules (`LinML.Policy`), plugin staging (`LinML.Catalog`), and native ONNX Runtime sessions (`LinML.Runtime.Onnx`). That gives developers a place to inspect device readiness and make provider choices before inference, then examine runtime diagnostics when execution differs from expectation.

• List available execution providers and diagnose readiness for the current machine.
• Select providers through explicit policies and inspect runtime or fallback diagnostics.
• Use the CLI, Python client, or .NET libraries to inspect models, run inference, benchmark, and serve persistent sessions.
• Stage plugin execution providers through LinML.Catalog with SHA-256 integrity checks and runtime registration.