The project enables users to speak to various large language models through Xiaomi AI speakers. It supports multiple LLM backends including ChatGPT, New Bing, ChatGLM, Gemini, Doubao, Moonshot, 01 AI, Llama3 via Groq, and Alibaba’s Tongyi Qianwen. Users interact by speaking to the speaker, which queries the selected LLM and returns the response as audio. The system requires a Xiaomi account to obtain a device ID and pipes voice input to LLM APIs while converting responses back to speech.