Open Edge AI

Local-first AI chat for mobile. Built with React Native on Android, native SwiftUI on iOS, and on-device AI modules that keep private context close to the user.

Overview

Open Edge AI is an open-source mobile AI app for building a modern chat
experience around local models, multimodal inputs, and device-side retrieval.
Android uses a React Native shell; iOS uses a separate native SwiftUI shell.
The app currently includes:

a polished iOS-inspired chat interface;
persistent chat sessions and work folders;
an Android React Native bridge for native AI calls;
Android Kotlin modules for model status, model loading, routing, embeddings,
vector search, and background indexing;
a web preview target for fast UI iteration with react-native-web.

The project is under active development. APIs and native model integration
details may change before the first stable release.

Features

Local-first chat UI with model selection and session management.
Work folder tree with rename, settings, pin, remove, and delete actions.
Persisted chat sessions, folders, selected model, and display settings.
FontAwesome/SVG icon pipeline for consistent cross-platform rendering.
Native AI bridge contract in TypeScript.
Android native AI engine scaffolding and implementation under com.openedgeai.
Web preview for frontend-only development.

Project Status

Area	Status
Android React Native UI	Active
Android native bridge	Active
Local model lifecycle	Active
Vector DB / RAG pipeline	Active
iOS native app	Active
Public release process	In progress

Repository Structure

.
├── App.tsx                     # React Native root component
├── src/                        # Cross-platform UI, bridge, theme, and web shims
│   ├── assets/                 # App logo and static assets
│   ├── components/             # Reusable React Native components
│   ├── native/                 # TypeScript wrapper around NativeModules.AIEngine
│   ├── screens/                # Chat and settings screens
│   ├── theme/                  # Tokens, icons, and display scaling
│   └── web/                    # React Native Web compatibility shims
├── android/                    # Android app and Kotlin on-device AI core
│   └── app/src/main/java/com/openedgeai/
│       ├── bridge/             # React Native <-> Kotlin bridge
│       ├── core/               # Model runtime, routing, embeddings, vision
│       ├── db/                 # SQLite vector store helpers
│       └── workers/            # Background indexing workers
├── ios/                        # Native SwiftUI iOS app and Swift AI runtime
├── docs/                       # Architecture, development, and structure docs
├── .github/                    # Issue templates, PR template, CI, Dependabot
├── COMMERCIAL_SUPPORT.md       # Voluntary commercial support notice
└── LICENSE                     # MIT License

For a more detailed tree, see docs/PROJECT_STRUCTURE.md.

Getting Started

Requirements

Node.js >= 22.11.0
npm
OpenJDK 17 for Android builds
Android Studio and Android SDK for Android development
Xcode and CocoaPods for iOS development on macOS
iOS 26.2 or newer for native Apple Foundation Models and Gemma runtime testing

Install

npm install

Run

# Metro for Android React Native workflows
npm start

# Same Metro port, kept for Android-specific workflows
npm run start:android

# Android app. This installs debug with Metro port 8082,
# applies adb reverse for 8082, and starts com.openedgeai/.MainActivity.
npm run android
npm run android:8082

# Native iOS app. Requires Xcode and CocoaPods; Metro is not required.
# React Native CLI iOS discovery is disabled in react-native.config.js.
npm run ios:pods
npm run ios

# Browser preview
npm run web

Model Assets

Large local model binaries are intentionally not committed. Development builds
can use placeholder files under android/app/src/main/assets/, while real model
files should be downloaded on device or stored locally outside Git.

Expected development model path:

android/app/src/main/assets/models/gemma-4-E2B-it.litertlm

See docs/MODEL_ASSETS.md for model asset policy and
runtime notes.

Development

Common quality checks:

npm run lint
npx tsc --noEmit
npm test -- --runInBand

More setup and troubleshooting notes are available in
docs/DEVELOPMENT.md.

Branding

Brand-facing names and assets are centralized through
docs/BRANDING.md. Use npm run branding:apply with a local
.env file to update runtime branding, launcher labels, bundle identifiers,
logos, and prepared app icon sets.

Architecture

Open Edge AI uses a React Native shell on Android, a native SwiftUI shell on
iOS, and native model execution modules on each platform. The React Native CLI
is configured to ignore the iOS project so RN tooling cannot build or mutate
the SwiftUI app path.

Android React Native UI
  -> src/native/AIEngine.ts
  -> Android Kotlin bridge

iOS SwiftUI UI
  -> AIEngineFoundationModelClient / AIEngineGemmaModelClient
  -> Apple Foundation Models / LiteRT-LM runtime

See docs/ARCHITECTURE.md for details.

Contributing

Contributions are welcome. Please read:

Commercial Support

Open Edge AI is MIT-licensed and can be used commercially under the license
terms. Organizations that rely on the project commercially are encouraged to
support maintenance. See COMMERCIAL_SUPPORT.md.

License

Open Edge AI is released under the MIT License.

core-app