wukong-robot

Wukong-robot is an open-source Chinese-language voice assistant framework built in Python. It is designed for hobbyists and makers who want to build their own smart speaker or voice-controlled device, primarily targeting Raspberry Pi and other Linux-based single-board computers, though it also runs on Intel Macs and Windows with the Linux subsystem. As of March 2023 the README reports it had been installed on over 13,000 devices and triggered more than 700,000 wake-word detections.

The framework is modular. Speech recognition, text-to-speech synthesis, and the conversational dialogue engine are each interchangeable plugins. Supported speech recognition providers include Baidu, iFlytek, Alibaba, Tencent, OpenAI Whisper, Apple, and Microsoft Edge. Supported voice synthesis options include VITS voice cloning and several cloud TTS services. For conversation, the project supports a local question-answering engine, a third-party chatbot service, and ChatGPT for multi-turn dialogue.

Wake-word detection works offline using either Porcupine or Snowboy, so the device listens for its trigger phrase without sending audio to a server. The README also mentions support for waking the robot via a Muse brain-computer interface headset, which the authors describe as possibly the first open-source smart speaker project to support that.

Smart home integration is available through MQTT, HomeAssistant, and Xiaomi's Xiao Ai speaker. A built-in web management panel runs at a local address and lets you change settings, view logs, and interact with the robot remotely. The backend also exposes a REST API for building custom integrations.

The README is written in Chinese and provides installation guides, plugin documentation, and configuration instructions. Requires Python 3.7 to 3.9.