nanoAI is a privacy-first Android app that brings powerful AI capabilities directly to your device. Chat with AI, generate images, process audio, get coding help, and translate languages – all while keeping your data private and secure.
- 🔒 Privacy by Design – Your conversations, personal data, and AI models stay on your device.
- ⚡ Works Offline – No internet required for most features once models are downloaded.
- 🎯 Multimodal AI – Chat, image generation, audio processing, code assistance, and translation all in one app.
- 🎨 Beautiful & Accessible – Clean, intuitive interface that works great on any Android device, with full TalkBack support and Material 3 design.
- 🔄 Flexible & Extensible – Add cloud AI providers (OpenAI, Gemini, custom endpoints) or use local models as you prefer. Persona system for different AI styles.
- 🛡️ Responsible AI – No automated content filters; users are responsible for generated content. First-launch disclaimer explains this clearly.
- 📱 Local-First Architecture – Runs small LLMs on-device for privacy and speed, with cloud fallbacks when needed or for larger models.
- Download nanoAI from github release. (coming soon)
- Launch the app and accept the privacy notice
- Download a model from the built-in library
- Start chatting – you're ready to go!
# Clone and build the app
git clone https://github.com/vjaykrsna/nanoAI.git
cd nanoAI
./gradlew installDebug
# Launch on your device or emulator- Multiple personas – Switch between helpful assistant, coding expert, creative writer, and more. Create custom personas with different prompts and model preferences.
- Threaded conversations – Keep different chats organized and easily accessible. Sidebar history with search and archive options.
- Smart suggestions – Context-aware responses that understand your conversation history.
- Local vs Cloud – Toggle between on-device models (private) and cloud APIs (OpenAI, Gemini, custom endpoints).
- On-device creation – Generate images without sending data to external servers (planned for future release)
- Multiple styles – Choose from various artistic styles and formats (planned)
- Privacy-first – Your prompts and generated images stay completely private
- Voice interaction – Speech-to-Speech and text-to-speech capabilities (planned for future release)
- Audio processing – Transcribe, translate, and analyze audio content (planned)
- Accessibility focus – Full screen reader support for visually impaired users
- Model Library – Browse, download, pause/resume models with progress tracking and size requirements.
- Settings & Export – Configure APIs, export/import personas and settings (JSON format, unencrypted with warnings).
nanoAI follows clean architecture principles with Kotlin-first design:
- UI Layer: Jetpack Compose with Material 3, accessible components
- Domain Layer: Use cases for business logic (chat, downloads, personas)
- Data Layer: Room database for local storage, Retrofit for cloud APIs, WorkManager for background tasks
- Runtime Layer: MediaPipe Generative for on-device inference, extensible to TensorFlow Lite/MLC LLM
Key technologies: Kotlin 2.2.x, Jetpack Compose, Hilt DI, Room, DataStore, WorkManager, Coroutines.
- Comprehensive Test Suite: Unit tests (ViewModels, repositories), instrumentation tests (Compose UI), macrobenchmarks
- Coverage Goals: Targeting 75% ViewModel, 65% UI, 70% Data layer coverage
- Quality Gates: ktlint, Detekt, Android Lint, automated CI checks
- Current Status: Foundation tests in place, working to close coverage gaps
- 🎯 Close test coverage gaps to meet quality thresholds
- 🎯 Implement text generation and polished chat UI
- 🎯 Import/export improvements for personas and settings
- 🖼️ Image generation support (on-device and cloud)
- 🎵 Audio input/output (voice chat, transcription, TTS)
- 🔄 Advanced persona workflows and multi-model orchestration
- 🌐 Translation and summarization modes
nanoAI isn't just an app—it's evolving into your personal AI ecosystem. Imagine:
- 🔄 Multi-API Load Balancing: Seamlessly switch between OpenAI, Gemini, Anthropic, and custom endpoints with intelligent routing. Configure multiple API keys for cost optimization—never get rate-limited again!
- 🏠 Local Network AI Hub: Turn your device into a local AI server. Host your own load balancer and API switcher accessible by other apps on your network. Share AI capabilities with your smart home, other devices.
- 🌍 Marketplace: Community-driven model library with user-contributed models and persona, with feedback and ratings. Earn rewards for contributing high-quality models.
- 🤖 AI Agent: Build and deploy custom AI agents that work across your devices and services, creating a truly intelligent personal assistant network.
- 💻 Vibe Coding with Linux Backend: Integrated Termux environment for seamless coding experiences. Write, run, and test code directly within the app using a full Linux backend, with AI assistance for code generation, debugging, and project management.
The future is limitless—nanoAI will be the central hub connecting you to the world's AI capabilities, all while keeping your data private and under your control. Join us on this exciting journey!
We welcome contributions! The project uses a structured spec-driven development process:
- Check
specs/for feature specifications and current plans - Review
docs/for architecture, testing, and API documentation - Follow the testing guide for adding comprehensive test coverage
- Run
./gradlew checkto ensure quality gates pass
git clone https://github.com/vjaykrsna/nanoAI.git
cd nanoAI
./gradlew build # Run full build with tests
./gradlew installDebug # Install on deviceSee docs/development/TESTING.md for detailed testing instructions and docs/architecture/ARCHITECTURE.md for technical deep-dive.
Key guides for contributors:
docs/development/BUILD_WORKFLOW.md– build logic conventions and automationdocs/development/QUALITY_GATES.md– required static analysis and coverage thresholdsdocs/development/TESTING.md– testing strategy, tooling, and coverage workflows
[Add license information here]
Built with modern Android technologies and a focus on user privacy and responsible AI use.
- Programming help – Get explanations, debugging help, and code suggestions
- Multiple languages – Support for popular programming languages
- Context aware – Understands your project structure and coding patterns
- Real-time translation – Translate between multiple languages instantly
- Conversation mode – Maintain context across multiple exchanges
- Offline support – Works without internet for downloaded language models
- Local model library – Download and manage AI models optimized for your device
- Cloud integration – Optional connection to services like OpenAI or Gemini for enhanced capabilities
- Automatic updates – Keep your models current with the latest improvements
- Theme selection – Light, dark, or system-following themes
- Layout preferences – Compact or comfortable spacing based on your preference
- Accessibility options – High contrast, large text, and screen reader optimization
- Export conversations – Backup your chat history and persona settings
- Privacy settings – Control what data (if any) is shared for app improvement
- Secure storage – All personal data encrypted and stored locally
nanoAI is built with transparency and user control in mind:
- Open source – Code available for security review and community contribution
- Privacy-first architecture – Designed to minimize data collection and external dependencies
- Quality gates – Automated testing ensures reliability across all features
- Accessibility compliance – Works with screen readers and meets WCAG guidelines
# Run tests and quality checks
./gradlew check
# Install on connected device
./gradlew installDebug
# View test coverage reports
./gradlew jacocoFullReport
# Reports available at app/build/reports/jacoco/full/index.htmlWe welcome contributions! Please:
- Start with tests – Add tests for new features before implementation
- Follow our quality gates – Ensure all tests pass and coverage thresholds are met
- Respect privacy – Any new features must maintain our privacy-first approach
- Document changes – Update user-facing documentation for any new capabilities
See our Testing Guide, Quality Gates, and Architecture Overview for development details.
- Enhanced multimodal – Better image generation and audio processing capabilities
- Expanded model support – More local AI models and cloud provider options
- Improved accessibility – Even better support for users with different abilities
- Performance optimization – Faster responses and smoother interactions
MIT License – you're free to use, modify, and distribute this software.
Made with ❤️ for people who value their privacy and want AI that works for them, not the other way around.
Have questions? Found a bug? Want to contribute? We'd love to hear from you!