Code has been updated to use the Gemini Live model specifically designed for back-and-forth voice interaction. This allows for smoother responses and more consistent output. Code & docs are at: https://github.com/smitamw/esp32s3-assistant