OLV-PYQT (Open-LLM-VTuber PyQt)

OLV-PYQT is a Live2D-powered PySide6 desktop application that integrates ASR (Automatic Speech Recognition), TTS (Text-to-Speech), and LLM backends for an interactive VTuber-style chat experience. This repository includes the Qt GUI (OQWindows), backend controllers (OQController), and configuration system (OQConfig) for a responsive AI character with pluggable speech, voice, and chat providers.

Highlights

1. Live2D-powered interface — Qt windows render a Live2D mesh while WSController orchestrates transcription, synthesis, and expression updates.
2. Extensive audio stack — Local and cloud-first voices plus ASR (faster-whisper, Sherpa ONNX, Azure, etc.) keep streaming interactions smooth.
3. LLM-first backend — Flexible config-driven agent templates let you point to Claude, DeepSeek, OpenAI, Groq, or local models such as llama-cpp-python.

Prerequisites

• Python 3.12+ (project requires modern async features and typed libs)
• A machine with OpenGL support for the Qt/Live2D renderer
• Optional: CUDA-enabled GPU if you plan to run faster-whisper, torch, or llama-cpp-python locally

Setup

1. Create and activate a virtual environment next to this repo (python -m venv .venv).

2. Upgrade pip then install dependencies:

   python -m pip install --upgrade pip
   pip install -r requirements.txt

3. Install the correct Live2D runtime wheel that matches your interpreter (wheels are already provided for CPython 3.11/3.12).

Configuration

• The canonical config lives in OQConfig/config.yaml. Before you start the app, update the llm_configs.*.llm_api_key, TTS key placeholders, and any Azure/Groq tokens with your own credentials.
• OQConfig/current_selection.json defaults to Edge TTS, Sherpa ASR, and a DeepSeek-compatible LLM. Adjust the paths (e.g., character_path) to point to Resources/v3/<model>.model3.json for your Live2D model.
• Runtime logs and histories are written out to logs/ and chat_history/; those folders are ignored by Git and are cleared when appropriate.

Running

python main.py

• main.py initializes the Qt loop, starts the Live2D canvas, and boots the WebSocket-backed controller.
• If you are building your own backend, point the controller to ws://127.0.0.1:12393/client-ws (see OQController/ws_controller.py).

Testing

python test_config_sync.py

• This script verifies that the config sync service can push prompts to OQConfig/config.yaml and mirrors the same values back out.

Development Notes

• Use requirements.txt to capture dependency versions. pyproject.toml already lists a subset of packages for publishing.
• Clean temporary artifacts with git clean -fdx only after verifying there are no unsaved changes.
• The .gitignore now filters logs/, chat_history/, .venv/, and other generated files.

Contributing

• Fork this repo, branch from main, and open a pull request. Document new providers in OQConfig/config.yaml.
• Run python -m pytest (or add additional tests) before submitting.

License

Add your preferred open-source license here (e.g., MIT, Apache 2.0, etc.).