What is ai-robot?

This project turns a small Rockchip RK3506 development board into a voice controlled AI companion. You press a physical button, ask a question out loud, and a few seconds later the AI's answer appears as text on an attached LCD screen. The whole flow is button press, voice recording, WiFi upload to a remote AI service, and then a display of the reply. The README lists several everyday scenarios this could serve: a child asking simple questions like how dinosaurs went extinct, someone in the kitchen looking up a recipe while their hands are messy, an older person who prefers speaking over using a smartphone, or someone at a desk wanting a quick answer about exchange rates or the weather. The target is for the reply to appear within about three seconds after the recording finishes. Under the hood, the whole thing is written in plain C with no dynamic memory allocation at runtime, meaning all memory is set aside ahead of time rather than requested while the program runs. It works as a single threaded loop that polls hardware every 50 milliseconds. A state machine moves the device through five states: idle while waiting for a button press, recording while capturing up to 60 seconds of audio, processing while it checks WiFi and uploads the recording, then either a result state showing the AI's reply or an error state if something failed, such as an audio, WiFi, HTTP, timeout, or server problem. To build and run it you need a Rockchip RK3506 board running Linux, a compatible microphone, an SPI or RGB LCD screen, a WiFi module, and a physical button and LED wired to GPIO pins. Setup involves installing several C libraries, cross compiling for the ARM processor with CMake, downloading a Chinese font file for on screen text, editing a configuration file with your WiFi credentials, and then copying the resulting binary onto the board to run. The project is released under the MIT license, based on the badge shown at the top of the README.