QuickVoiceBot

QuickVoiceBot is an AI-powered conversational voice bot that utilizes streaming speech-to-text, text-to-speech, and a large language model to engage in fast, fluid conversations. It provides a seamless voice interface for natural interactions.

Features

• Real-time speech recognition using Deepgram's streaming API
• Natural-sounding text-to-speech powered by Deepgram
• Intelligent responses generated by a powerful language model (Groq or OpenAI)
• Streaming capabilities for fast, low-latency interactions
• Simple, easy-to-use conversation flow

Setup

1. Clone the repository:

   git clone https://github.com/yourusername/QuickVoiceBot.git
   

2. Install the required dependencies:

   pip install -r requirements.txt
   

3. Set up your API keys:

   • Sign up for a Deepgram account and obtain an API key
   • (Optional) Sign up for an OpenAI account and obtain an API key if using OpenAI's models
   • Create a .env file in the project root and add your API keys:

     DEEPGRAM_API_KEY=your_deepgram_api_key
     OPENAI_API_KEY=your_openai_api_key
     

4. Run the bot:

   python main.py
   

Usage

1. Start the bot by running main.py
2. Speak to the bot and it will respond with voice
3. Say "goodbye" to end the conversation

Customization

• Adjust the language model settings in main.py:
  • Choose between Groq and OpenAI models by uncommenting the appropriate lines
  • Modify the model names and parameters as needed
• Customize the system prompt in system_prompt.txt to define the bot's persona and behavior
• Experiment with different Deepgram models for speech recognition and synthesis

File Structure

• main.py: Main entry point of the application
• system_prompt.txt: File containing the system prompt for the language model
• requirements.txt: List of required Python packages
• building_blocks/: Directory containing isolated components for inspection
  • speech_to_text_streaming.py: Standalone example of streaming speech recognition
  • text_to_speech.py: Standalone example of text-to-speech
  • llm.py: Standalone example of using the language model

Contributing

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

License

This project is licensed under the MIT License.

Acknowledgements

• Deepgram for providing the speech recognition and synthesis APIs
• Groq for the powerful language model
• OpenAI for the alternative language model option