BonziAssist: Your New (Old) Assistant!

This project uses the TETYYS generator. Also see: BonziBuddy-TTS.

⚠️ Do Not Abuse the API ⚠️

BonziAssist revives the classic BonziBuddy assistant in a modern, voice-activated form, allowing you to interact with an AI through spoken commands. Here’s a bit more about what makes BonziAssist special and how to use it.

Features

• Voice Activation: Simply say "Bonzi" and once you hear Bonzi confirm he's listening, you're ready to go!
• Text-to-Speech: Utilizes the nostalgic BonziBuddy voice for responses, thanks to the TETYYS SAPI4 API.
• Real-Time Interaction: Communicate in real-time, with responses generated through a lightweight LLM.

Installation

1. Clone this repository.
2. Ensure you have Python installed.
   • Windows users may also need Visual C++ Build Tools.
   • macOS users may need PortAudio first: brew install portaudio.
3. Install the Python dependencies:

   pip install -r requirements.txt

4. Set up your environment variables:

   cp helpers/.env.example helpers/.env

   Then edit helpers/.env and set LLM_PROVIDER and GROQ_API_KEY.
5. Confirm the local setup:

   python main.py --check

Usage

To start BonziAssist:

python main.py

You will be prompted to configure your microphone settings upon the first run or if you choose to be prompted every time by the configuration settings. After startup, say "Bonzi" and wait for the confirmation sound before speaking your command.

TODO

• Add LiteLLM: Enable Bonzi to talk back to you
• HIGH PRIO: Improve STT Accuracy: Potentially use a bigger VOSK model
• Improve STT Speed: Potentially use a different package
• Improve TTS Speed: Find a way to run locally for quicker responses
• Offline Functionality: Use fully offline APIs such as Ollama for those with beefier computers
• Add Voice Volume Control
• Add simple windows functions: Such as Media control, opening apps, etc.
• Add interruption support: Potentially via a hotkey (press esc to interrupt)
• Add fancy splash screen

Continuous goals

• When VOSK misunderstands, add to the list of phrases Bonzi will respond to
• Update system prompt to give more personality and functionality
• Optimize Performance

Contributors are welcome to help with these tasks. If you are interested, please fork the repository, make your changes, and submit a pull request. If you have any questions, feel free to open an issue and ask away!

How it Works

1. Listening: The system starts by listening for the activation word "Bonzi". Once heard, it's ready to receive commands.
2. Processing: Speech is converted to text and processed by an LLM to understand the context and generate an appropriate response.
3. Responding: The response is then converted to speech using the SAPI4 API, maintaining the classic voice of BonziBuddy.

Contributing

First of all, I appreciate anyone being excited about this project. That being said, contributions are welcome! For major changes, please open an issue first to discuss what you would like to change.

Acknowledgments

• SAPI4: The Text-to-Speech engine that allows BonziAssist to speak with the classic BonziBuddy voice. Visit TETYYS's online SAPI4.
• BonziBuddy-TTS GitHub Project: Inspired the integration of the BonziBuddy voice into this project. View the repository.
• Vosk Speech Recognition: Provides the voice recognition capabilities required to interpret user commands. Learn more about Vosk.
• LitteLLM: For processing natural language and generating responses. Their LLM services form the backbone of BonziAssist's AI capabilities.
• SimpleAudio: Used for playing back audio responses generated by the system.
• PyAudio: Handles microphone input and audio stream management.
• Python-dotenv: For managing environment variables in a .env file, which helps secure API keys and other sensitive settings.
• GitHub Community: For all the open-source contributors who have developed libraries and tools that made this project feasible.