tektra 0.8.1

Advanced AI assistant with Microsoft Phi-4 Multimodal integration, voice, vision, and robotics capabilities

Tektra AI Assistant

Tektra AI Assistant is an advanced AI assistant featuring Microsoft Phi-4 Multimodal integration with superior voice, vision, and robotics capabilities. It delivers state-of-the-art speech recognition, intelligent chat completion, and multimodal understanding with persistent conversation management and a beautiful web interface.

✨ Features

🧠 Advanced AI Integration

• Microsoft Phi-4 Multimodal: Primary processor with #1 OpenASR leaderboard performance
• Intelligent Fallback: Automatic fallback to OpenAI Whisper for reliability
• Unified Architecture: Single model handles speech recognition and chat completion
• 8-Language Support: English, Chinese, German, French, Italian, Japanese, Spanish, Portuguese
• Streaming Responses: Real-time token-by-token generation with 128K context
• Model Management: Load, unload, and switch between models with automatic device detection

💬 Rich Chat Interface

• Persistent Conversations: Full conversation history with database storage
• Real-time Chat: WebSocket-based streaming chat
• Conversation Management: Create, search, organize, and delete conversations
• Message Actions: Copy, regenerate, and manage individual messages
• Beautiful UI: Modern, responsive web interface

🎤 Superior Voice & Multimodal Capabilities

• Advanced Speech Recognition: Phi-4 Multimodal with superior accuracy and 8-language support
• Real-time Transcription: WebSocket-based streaming with voice activity detection
• Intelligent TTS: Edge-TTS with 200+ neural voices and automatic language matching
• Noise Cancellation: WebRTC VAD with advanced audio preprocessing
• Multimodal Understanding: Vision, audio, and text processing in unified model
• Camera Integration: Real-time video processing with computer vision

🤖 Robotics & Automation

• Robot Control: Command and control robotic systems
• Avatar System: 3D avatar with expressions and gestures
• Real-time Communication: WebSocket-based robot communication
• Safety Features: Emergency stop and safety monitoring

🚀 Quick Start

Installation

Install Tektra with UV (recommended):

# Install UV if you don't have it
curl -LsSf https://astral.sh/uv/install.sh | sh

# Basic installation
uv tool install tektra

# With all optional features
uv tool install tektra --with tektra[all]

# Alternative: Traditional pip installation also works
pip install tektra

Setup and First Run

Just run Tektra - setup happens automatically!

tektra

That's it! 🎉

• Auto-Setup: Directories, database, and dependencies are created automatically
• Model Loading: Phi-4 Multimodal downloads and loads automatically on first run
• Browser Opens: Web interface opens automatically at http://localhost:8000
• Ready to Use: Start chatting, speaking, or using the camera immediately

🚀 Alternative Commands

tektra start       # Explicit start command
tektra enhance     # Enhanced setup with guided Phi-4 configuration
tektra enable-phi4 # Manually enable Phi-4 if needed

🚀 New in v0.8.0 - Streamlined Experience

🎯 Zero-Setup Experience

• One Command Install: uv tool install tektra → tektra → Done!
• Automatic Setup: No more tektra setup - directories, database, and models initialize automatically
• Instant Start: Just run tektra and everything works immediately
• Smart Defaults: Optimized configuration for immediate productivity

🚀 UV Package Manager Integration

• Modern Installation: UV-first approach with traditional pip fallback
• Faster Dependencies: UV's speed benefits for installation and development
• Better Development: uv sync --all-extras for seamless dev setup
• Tool Integration: Native UV tool installation support

Microsoft Phi-4 Multimodal Integration

• Superior Speech Recognition: #1 performance on OpenASR leaderboard
• Unified Processing: Single model for both STT and chat completion
• 8-Language Audio Support: Enhanced multilingual capabilities
• 128K Context Length: Extended context for better understanding
• Intelligent Fallback: Automatic Whisper fallback for reliability

Enhanced API Endpoints

• /api/v1/audio/phi4/load - Load Phi-4 Multimodal model
• /api/v1/audio/phi4/unload - Unload model to free memory
• /api/v1/audio/phi4/info - Get model status and capabilities
• Enhanced transcription and language detection with Phi-4 primary processing

Performance Improvements

• Faster Processing: Unified model reduces latency
• Better Accuracy: Superior speech recognition and understanding
• Memory Efficiency: Optimized model loading and device detection
• Real-time Streaming: Enhanced WebSocket processing with Phi-4

📖 Usage

Command Line Interface

# Start the server (default command)
tektra

# Start with custom options
tektra start --host 0.0.0.0 --port 8000 --no-browser

# Enhanced setup (server + Phi-4 guidance)
tektra enhance

# Enable Phi-4 Multimodal (superior AI)
tektra enable-phi4

# Check Phi-4 status
tektra phi4-status

# Disable Phi-4 (free memory)
tektra disable-phi4

# Show system information
tektra info

# Show version
tektra version

# Get help
tektra --help

Web Interface

Once started, you can:

1. Chat with AI: Start conversations with intelligent responses
2. Manage Conversations: Browse, search, and organize your chat history
3. Voice Interaction: Use voice input and output
4. Control Systems: Manage robots, avatars, and connected devices

🔧 Configuration

Environment Variables

Create a .env file in your working directory:

# Database
DATABASE_URL=sqlite:///./tektra.db

# Server
HOST=0.0.0.0
PORT=8000
DEBUG=false

# AI Models
MODEL_CACHE_DIR=./models
MAX_MODEL_MEMORY_GB=8

# Audio
AUDIO_SAMPLE_RATE=16000
MAX_AUDIO_DURATION=30

# API Keys (optional)
OPENAI_API_KEY=your_openai_key
ANTHROPIC_API_KEY=your_anthropic_key

🏗️ Architecture

Tektra AI Assistant is built with a modern, modular architecture:

tektra/
├── app/                    # Main application
│   ├── routers/           # API endpoints
│   ├── services/          # Business logic
│   ├── models/            # Database models
│   ├── database.py        # Database management
│   └── config.py          # Configuration
├── frontend/              # Web interface (React/Next.js)
├── cli.py                 # Command-line interface
└── server.py              # Server management

Technology Stack

• Backend: FastAPI, SQLAlchemy, WebSockets, Python 3.9+
• Frontend: Next.js, React, TypeScript, Tailwind CSS
• Database: SQLite (default), PostgreSQL (optional)
• AI/ML: Transformers, PyTorch, MLX (Apple Silicon)
• Audio: PyAudio, LibROSA, SoundFile
• Vision: OpenCV, MediaPipe

📦 Installation Options

Basic Installation

pip install tektra

With Machine Learning

pip install tektra[ml]

With Audio Support

pip install tektra[audio]

With Vision Support

pip install tektra[vision]

With Robotics Support

pip install tektra[robotics]

Everything Included

pip install tektra[all]

Development Installation

git clone https://github.com/tektra/tektra.git
cd tektra
uv sync --all-extras

🛠️ Development

Setting Up Development Environment

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

# Install in development mode with UV
uv sync --all-extras

# Set up pre-commit hooks
uv run pre-commit install

# Run tests
uv run pytest

# Start development server
uv run tektra start --reload --debug

🐛 Troubleshooting

Common Issues

Installation Problems:

# If you get permission errors
pip install --user tektra

# If you have dependency conflicts
pip install tektra --no-deps
pip install -r requirements.txt

Audio Issues:

# On macOS
brew install portaudio
pip install pyaudio

# On Ubuntu/Debian
sudo apt-get install portaudio19-dev
pip install pyaudio

Database Issues:

# Reset database
rm tektra.db
tektra setup

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Made with ❤️ by the Tektra Team