Neural Network Runtime Protocol
NNRP is a domain-level application protocol for AI model calls, streaming results, cache bindings, and session migration. It moves runtime cooperation semantics into the protocol layer so SDKs, conformance suites, and cloud runtime integrations can evolve around the same contract.
Start from the background, boundaries, core objects, and version entry points.
ConformanceShared Test SuiteRead capability manifests, CI integration, and version-selection rules.
SDKLanguage SDKsOpen quick starts and API references for Python, C#, Rust, and more.