what is esp32-bluetooth-voice-agent fr?

This is experimental firmware that turns a cheap ESP32 microcontroller board into a Bluetooth handset for talking to an AI voice agent, at a hardware cost of under $20 excluding the phone and a bridge computer. The board pairs with a phone or tablet the same way a Bluetooth headset would, then carries the conversation audio to a voice AI service instead of a human on the other end of a call. The project has two parts. The ESP32 board advertises itself as a Bluetooth hands-free device and exchanges audio with the phone. That audio is sent over the local Wi-Fi network using WebRTC, a technology for real-time audio and video, to a companion program called the bridge that runs on a laptop or similar computer. The bridge, written in Go, manages the connection to the voice AI service, converts the audio into the right format, and can log activity to a small local database. The included demo by default sends live microphone audio to a free voice API hosted by a company in China, which the author flags as a possible privacy concern. Users can instead run their own MiniCPM-o 4.5 model or connect the bridge to a different real-time voice API, including ones from providers based in other regions. Setup involves specific hardware, the AI-Thinker ESP32 Audio Kit V2.2 board with an antenna and an optional small screen and battery, plus a defined toolchain including PlatformIO, ESP-IDF, and Go. The README includes detailed wiring notes, power warnings for the battery and OLED display, and configuration steps for Wi-Fi credentials and the bridge's network settings. This first version automatically answers every incoming phone call and has no caller allow-list or physical on/off control yet, so the author recommends using it only for controlled testing rather than everyday use.