78/xiaozhi-esp32 ? reverse-engineered prompt

Build me a fun voice powered AI chatbot that runs on ESP32 boards like the M5Stack CoreS3 or Waveshare displays. It wakes up offline when I say a custom word, records what I say, streams it to something like Qwen or DeepSeek for a smart reply, then speaks back naturally with TTS while showing cute emojis and text on the screen.

Add speaker recognition so it knows who is talking, battery monitoring, WiFi connectivity, and the ability to control lights, servos or other gadgets around me. Make the whole thing feel alive and responsive with streaming audio using OPUS. Support a bunch of popular open source hardware out of the box and let me tweak wake words, fonts and backgrounds easily. Set it up in C++ with ESP IDF so I can flash it right away and start chatting.