A versatile voice assistant system written in Rust that provides voice interaction capabilities through various configurable services.
This project is part of a diploma thesis at HTL St. Pölten.
- Voice Recording (Local and Remote)
- Speech-to-Text (Local Whisper and Deepgram)
- Natural Language Understanding (Pattern Matching and Rasa)
- Text-to-Speech (ElevenLabs and Piper)
- LLM Integration (DeepSeek and Ollama)
- Weather Information (OpenWeatherMap)
- System Control
- Volume Management
- Window Management (KWin)
- Timer Functionality
- WebSocket-based Communication
- Configurable Service Implementations
The system uses a TOML-based configuration system that allows for easy customization of service implementations and parameters. The configuration file is automatically created at first run in the XDG config directory.
A graphical frontend is being developed in parallel at voice-frontend, which provides a user-friendly interface for configuring and using the voice assistant.
MIT - see LICENSE