Lugia is a high-performance local model inference system designed for macOS, built on top of MLX (Machine Learning X). It aims to provide a seamless experience for deploying and running machine learning models locally with minimal resource consumption.
- Fast Inference: Optimized for local inference with MLX, ensuring low latency and high throughput.
- Customizable Sampling: Advanced sampling techniques including adaptive sampling to generate diverse outputs from language models.
- Modular Tooling: Built in a modular way to allow easy extension and integration of additional functionalities.
- User Interface: Text-based user interface (TUI) for easy interaction.
- Text-to-Speech (TTS): Integration with TTS systems for voice-based interactions.
- Speech-to-Text (STT): Integration with STT systems for voice recognition.
- Sentiment Analysis (SST): Basic sentiment analysis capabilities.
- Conversational assistant: Mimics a conversational assistant using language models.
- Jarvis: Emulate a personal assistant akin to Siri or Alexa but running entirely locally.
- Avoid Bloat: Minimize dependencies and keep the implementation lightweight.
- Efficiency: Avoid excessive computational overhead.
- API-based: Focus on command-line and TUI interactions rather than serving APIs.
-
Install dependencies:
pip install -r requirements.txt
-
Clone the repository:
git clone https://github.com/your-repo/lugia.git cd lugia
python talk.pyThis script records audio input, transcribes it, sends the text to a local language model, and then plays the response as audio.
You can use tool.py to perform various tasks such as transcribing audio, sampling from models, etc.
python tool.py --task transcribe --audio_path path_to_audio.wav