google-gemini/robotics-pointing-sample ? reverse-engineered prompt

Build me a Python tool for the SO 101 robot arm that uses a USB camera and Gemini to point at things I name in plain English.

I want it to first calibrate the workspace with a ChArUco board, save that calibration so I do not have to do it every time, then open an interactive loop where I can type something like blue block or pen and the robot finds it in the camera view, converts the image position into real world coordinates, and moves the arm to point a little above the object. It should take the robot serial port, camera index, robot id, API key, and optional calibration folder from the command line, and it should let me quit cleanly when I type q.

If you need current setup details for Gemini, robot control, or camera handling, look up the latest docs online and wire it up in a way that actually runs.