Skip to main content

Expo AI Runtime — first cut

Stewart Moreland
Maintainer, Expo AI Runtime

The first cut of the Expo AI Runtime is here: one TypeScript API over Apple Foundation Models, Android Gemini Nano, and an explicit cloud fallback.

What's in this release​

  • @stewmore/expo-ai-core — the ExpoAI API, provider router, capability registry, sessions, structured-output validation/repair, privacy metadata, and normalized errors.
  • @stewmore/expo-ai-apple-foundation-models — on-device iOS provider (iOS 26+).
  • @stewmore/expo-ai-android-aicore — on-device Android provider (Gemini Nano via AICore), with model download.
  • @stewmore/expo-ai-cloud — opt-in cloud fallback client adapter.
  • @stewmore/expo-ai-evals — Node-first eval harness.

Built on Expo SDK 56. See the roadmap for what's next — streaming hardening, task APIs, LiteRT-LM bring-your-own-model, and beyond.