Voice Assistant with Mistral AI

A powerful voice assistant that uses Mistral AI for natural language processing, Whisper for speech recognition, and text-to-speech capabilities. The assistant can understand voice commands, search the web, and provide natural conversational responses.

Features

• 🎙️ Voice activation with wake word detection ("Hey Mistral")
• 🗣️ Natural speech recognition using Whisper
• 🤖 AI-powered responses using Mistral
• 🌐 Web search integration for real-time information
• 🔊 High-quality text-to-speech synthesis
• ⚡ Interrupt capability during responses
• 🎯 Context-aware conversations

Prerequisites

• Python 3.8 or higher
• Mistral AI running locally (via Ollama)
• Audio input device (microphone)
• Audio output device (speakers)

Installation

1. Clone this repository:

git clone [your-repository-url]
cd voice_assistant

2. Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows, use: venv\Scripts\activate

3. Install the required packages:

pip install -r requirements.txt

4. Make sure you have Ollama installed and the Mistral model running locally:

ollama run mistral

Usage

1. Activate your virtual environment if not already activated:

source venv/bin/activate  # On Windows, use: venv\Scripts\activate

2. Run the voice assistant:

python main.py

3. Wake up the assistant by saying "Hey Mistral" or any of the following wake words:

   • "Hi Mistral"
   • "Hello Mistral"
   • "Mistral"
   • "Arise"

4. Ask your question or give a command after the assistant acknowledges you

5. To exit, simply say "exit" or press Ctrl+C

Key Components

• main.py: Core application logic and voice assistant implementation
• stt.py: Speech-to-text functionality
• requirements.txt: Project dependencies

Features in Detail

Voice Activation

The assistant listens for wake words and becomes active only when called upon, preserving system resources and privacy.

Intelligent Responses

• Uses Mistral AI for generating contextual and intelligent responses
• Integrates web search capabilities for real-time information
• Streams responses sentence by sentence for natural conversation flow

Interruption Handling

• Users can interrupt the assistant's response by speaking
• Smart detection of user input during response playback
• Graceful handling of interruptions and conversation flow

Web Search Integration

• Automatic detection of queries requiring current information
• Integration with DuckDuckGo for web searches
• Contextual responses incorporating web search results

Dependencies

• sounddevice==0.4.6: Audio input/output handling
• numpy==1.22.0: Numerical operations and audio processing
• requests==2.31.0: HTTP requests for web integration
• faster-whisper==0.10.0: Speech recognition
• TTS==0.22.0: Text-to-speech synthesis

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

[Your chosen license]

Acknowledgments

• Mistral AI for the language model
• Whisper for speech recognition
• Mozilla TTS for speech synthesis
• DuckDuckGo for web search capabilities