RunAnywhere AI — React Native Example

A cross-platform reference app for the RunAnywhere React Native SDK. One TypeScript codebase demonstrating on-device LLM chat, speech, vision, RAG, voice agents, and model management on iOS and Android.

Requirements

This project uses nodeLinker: node-modules (not Plug'n'Play).

Setup

Important: This sample consumes local file: workspace packages. A clean clone needs JavaScript dependencies, staged native binaries (Android JNI + iOS XCFrameworks), and CocoaPods codegen before either platform will build.

1. Clone and install JavaScript dependencies

git clone https://github.com/RunanywhereAI/runanywhere-sdks.git
cd runanywhere-sdks/examples/react-native/RunAnywhereAI

corepack enable
yarn install --ignore-scripts

Use --ignore-scripts on first install so postinstall hooks do not mask missing natives. Run yarn postinstall manually only after reviewing local patches.

2. Build and stage native artifacts (repo root)

cd ../../..
./scripts/build/build-core-android.sh arm64-v8a
./sdk/runanywhere-swift/scripts/build-core-xcframework.sh
cd examples/react-native/RunAnywhereAI

One run stages every Apple artifact the app can link: RACommons, RABackendLLAMACPP, RABackendONNX, RABackendSherpa, onnxruntime, onnx, RABackendCoreML, and the three MLX frameworks (RABackendMLX, RunAnywhereMLXRuntime, RunAnywhereMLXMetal). Any of them missing means this step was skipped.

3. Install iOS pods and verify builds

yarn pod-install

cd android && ./gradlew :app:assembleDebug && cd ..

xcodebuild \
  -project ios/RunAnywhereAI.xcodeproj \
  -scheme RunAnywhereAI \
  -configuration Debug \
  -sdk iphonesimulator \
  -destination 'generic/platform=iOS Simulator' \
  build

Or run the combined gate:

./scripts/verify.sh        # Android by default; RUN_IOS=1 for iOS too

4. Run the app

yarn start          # Metro bundler
yarn ios            # iOS simulator or device
yarn android        # Android emulator or device

After modifying the SDK

Features

On Android, QHexRT NPU acceleration is available on supported Snapdragon devices when the native backend is staged. See NPU / QHexRT below.

NPU / QHexRT (Android)

The @runanywhere/qhexrt package registers on supported Hexagon hardware. Its Gradle project is included only when the native library is present at node_modules/@runanywhere/qhexrt/android/src/main/jniLibs/arm64-v8a/librac_backend_qhexrt.so.

To test private HNPU model bundles:

1. Open Settings → Downloads.
2. Save a Hugging Face token.
3. Download and load an HNPU model from the model UI.
4. Tap Clear to return to public downloads.

The token is passed via RunAnywhere.setHfToken(...) at runtime—not stored in source or logs.

Project structure

RunAnywhereAI/
├── App.tsx                     # SDK init, backend registration, navigation
├── src/
│   ├── screens/                # Chat, STT, TTS, Voice, RAG, Vision, …
│   ├── components/             # Chat bubbles, model picker, banners
│   ├── stores/                 # Zustand conversation persistence
│   └── theme/system/           # Material-3-style design tokens (#FF6900)
├── ios/                        # Xcode project, Podfile, native audio module
├── android/                    # Gradle project, manual SDK module wiring
├── scripts/
│   ├── verify.sh
│   └── smoke.sh
└── package.json                # Workspace deps → sdk/runanywhere-react-native/packages/*

SDK packages use NitroModules for JSI bridging. Hermes does not support for await...of on Nitro async iterables—use manual iterator.next() loops (see chat and download flows in the app).

Troubleshooting

Quick static check:

./scripts/smoke.sh
yarn typecheck

Related links

License

This project is licensed under the RunAnywhere License (Apache 2.0 based, with additional commercial-use terms). See LICENSE for details.