Tektra - AI Voice Assistant

🚀 A voice-interactive AI assistant powered by Google's Gemma-3n model running locally on your machine

Features

✅ Local AI - Runs Gemma-3n model locally with Metal acceleration on Apple Silicon
✅ Voice Interaction - Native audio recording with future speech-to-text support
✅ Smart Conversations - Context-aware responses with chat history
✅ Beautiful UI - Modern, gradient-based interface with real-time feedback
✅ Self-Contained - Automatic model downloading and caching
✅ Privacy-First - Everything runs locally on your machine

Architecture

• Frontend: React + TypeScript with Vite
• Backend: Rust with Tauri framework
• AI Engine: GGUF models for efficient inference (Gemma-3n E2B)
• Audio: Native audio recording with future STT/TTS support
• Model: Google Gemma-3n E2B (2.79GB) - automatically downloaded on first run

Installation

Pre-built Binaries (Recommended)

Download the latest pre-built binary for your platform from the GitHub Releases page.

From Source

Note: Tektra is a desktop application that requires both Rust backend and React frontend. cargo install alone will not work properly.

# Clone the repository
git clone https://github.com/dirvine/tektra
cd tektra

# Install frontend dependencies
npm install

# Build the complete application
npm run tauri:build

# The built application will be in src-tauri/target/release/bundle/

Cargo Install (Limited - Backend Only)

⚠️ Not recommended for end users - This only installs the Rust backend without the frontend UI:

cargo install tektra

Development Setup

# Clone the repository
git clone https://github.com/dirvine/tektra
cd tektra

# Install frontend dependencies
npm install

# Run in development mode
tektra dev
# OR
npm run tauri dev

Usage

Running the Application

Simply run:

tektra

CLI Commands

tektra --help     # Show help
tektra --version  # Show version information
tektra dev        # Run in development mode with hot reload

In-App Features

1. Text Chat: Type messages and get intelligent responses from Gemma-3n
2. Voice Input: Click the microphone button to record audio (STT coming soon)
3. Voice Output: Enable auto-speech in settings for spoken responses
4. Progress Tracking: Visual progress bar shows model download status
5. Settings: Customize preferences and view model information

Project Structure

tektra/
├── src/
│   └── main.rs          # Rust backend with built-in AI
├── icons/               # Application icons
├── index.html           # Main frontend interface
├── main.js              # Frontend JavaScript
├── package.json         # Node.js dependencies
├── Cargo.toml           # Rust dependencies
├── tauri.conf.json      # Tauri configuration
├── vite.config.js       # Build configuration
├── entitlements.plist   # macOS permissions
└── CLAUDE.md            # Development guidelines

Building from Source

Prerequisites

• Rust 1.70+ (install from rustup.rs)
• Node.js 16+ and npm
• macOS 11+ (for Metal acceleration, other platforms supported)

Build Steps

1. Clone and enter the repository:

git clone https://github.com/dirvine/tektra
cd tektra

2. Install frontend dependencies:

npm install

3. Build for release:

./build-release.sh

4. Install locally:

cargo install --path src-tauri

Models

Tektra uses the Gemma-3n E2B model (2.79GB) from Google, which is automatically downloaded on first run. The model is cached in ~/.cache/huggingface/hub/ for subsequent uses.

Model Details

• Name: Gemma-3n E2B (2 billion parameters)
• Size: 2.79GB (4-bit quantized GGUF format)
• Performance: Optimized for Apple Silicon with Metal acceleration
• Source: Automatically downloaded from HuggingFace Hub

Development

This project follows specification-driven development:

1. Read CLAUDE.md for detailed development guidelines
2. Use UV for Python dependencies (if needed)
3. Test thoroughly before committing changes
4. Follow Rust best practices - no unwrap() in production

Migration from Previous Versions

This is a complete rewrite of Project Tektra as a native desktop application:

• Previous: Python FastAPI backend + Next.js frontend
• Current: Rust Tauri desktop application
• Benefits: Self-contained, offline, native performance, simplified deployment

Contributing

1. Fork the repository
2. Create a feature branch
3. Follow the development guidelines in CLAUDE.md
4. Submit a pull request

Publishing to Crates.io

To publish this package to crates.io:

1. Make sure you're logged in:

cargo login

2. From the src-tauri directory:

cd src-tauri
cargo publish

Note: The frontend assets are bundled with the binary during the build process.

License

This project is licensed under either of:

• Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
• MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)

at your option.

Credits

Project Tektra - Built with ❤️ using Rust, Tauri, and modern web technologies.

Release History

For previous releases and development history, see the RELEASE_NOTES files in this repository.