xiaozhi-esp32-server-java

Why we included this project

Xiaozhi ESP32 Server Java is the backend half of the Xiaozhi smart speaker project, written for people who want to host their own voice assistant instead of leaning on a closed cloud service. The server runs two independent processes: one handles the live audio dialogue with the hardware, the other is an admin panel for monitoring devices, pushing OTA firmware updates, switching conversation personas, and reviewing past chat logs. Speech recognition and synthesis work locally or in the cloud, voice cloning included, and each device can point at any of several LLM backends, OpenAI, Ollama, Zhipu, Dify or Coze, with an MCP tool protocol for extending behavior. Because the two processes scale separately, a team can grow from a single gadget to a small fleet without touching the admin side. It is a practical fit for developers comfortable with Spring Boot and MySQL who want a documented path to running their own fleet.