Example apps

Five focused Expo apps demonstrate the runtime, plus a reference cloud server. They share UI and setup via a private _shared package.

App	Demonstrates
basic-generate	Capability card + ExpoAI.generate with provider/privacy badges and errors
structured-output	ExpoAI.generateObject with JSON-schema validation + repair
streaming-chat	ExpoAI.stream with live tokens and a Stop button
cloud-fallback	Sensitivity gating + explicit, privacy-aware cloud routing
hooks-demo	@stewmore/expo-ai-react hooks + streamObject; wired via the config plugins
server	Reference cloud backend (mock-by-default)

Run an app

These apps include custom native code, so they need a development build (not Expo Go).

# 1) (for cloud fallback / mock cloud) start the reference server
npm run server                       # from the repo root → http://localhost:8787

# 2) align Expo dependency versions for the app you want to run
cd examples/basic-generate
npx expo install --fix               # pins expo-build-properties etc. to this SDK

# 3) build & launch a dev client
npx expo prebuild --clean
npx expo run:ios                     # iOS 26 simulator (Apple Silicon) + Apple Intelligence → on-device
npx expo run:android                 # supported device for on-device; otherwise cloud fallback

What to expect per platform

• iOS 26 simulator with Apple Intelligence: capability card shows apple-foundation-models, results are on-device.
• Older iOS / simulator: capability shows unavailable; generate({ fallback: "cloud" }) routes to the reference server (third-party-cloud).
• Android emulator: AICore is usually unavailable → graceful fallback. On a supported Gemini Nano device, generation runs on-device.

Networking note

_shared points the cloud adapter at http://localhost:8787 (and 10.0.2.2 for the Android emulator). On a physical device, edit _shared/src/setup.ts to use your machine's LAN IP.