This is a voice assistant application built using LiveKit, a real-time communication platform. The assistant leverages various plugins and technologies to enable voice-based interaction and provide intelligent responses.

• Voice-based user interface
• Speech-to-text using Deepgram
• Text-to-speech using ElevenLabs
• Language model powered by OpenAI's GPT
• Voice activity detection using Silero
• Minimal and function-based assistant examples

• Docker
• Python 3.11.6

1. Clone the repository:

   git clone https://github.com/avonx/live2d-voice-assistant.git
   

2. Change to the project directory:

   cd voice-assistant
   

3. Build the Docker image:

   docker build -t voice-assistant .
   

4. Create a .env file in the project root directory and provide the necessary environment variables:

   OPENAI_API_KEY=your_openai_api_key
   DEEPGRAM_API_KEY=your_deepgram_api_key
   ELEVENLABS_API_KEY=your_elevenlabs_api_key
   

5. Run the Docker container:

   docker run --env-file ../.env --env WORKER_HOST=0.0.0.0 --env WORKER_PORT=8081 --net=host voice-assistant
   

The voice assistant provides two examples:

1. Minimal Assistant: A basic voice assistant that listens to user input, processes it using the language model, and responds with voice output.

2. Function-based Assistant: An advanced voice assistant that demonstrates function calling capabilities, such as toggling lights in different rooms and handling user interruptions.

To run the desired assistant, uncomment the corresponding entrypoint in the Dockerfile and rebuild the Docker image.

You can customize the voice assistant by modifying the code in the voice-assistant directory. The main files of interest are:

• minimal_assistant.py: Contains the code for the minimal voice assistant example.
• function_calling.py: Contains the code for the function-based voice assistant example.

Feel free to explore and extend the functionality of the voice assistant based on your requirements.

Contributions are welcome! If you find any issues or have suggestions for improvements, please open an issue or submit a pull request.

This project is licensed under the MIT License.

• LiveKit - Real-time communication platform
• OpenAI - Language model provider
• Deepgram - Speech-to-text provider
• ElevenLabs - Text-to-speech provider
• Silero - Voice activity detection library

• LiveKit Agents