KingBobJoeIV/cubed-core ? reverse-engineered prompt

Build me a local web app for reconstructing a 3x3 cube solve from 120 fps phone video.

I want a small workbench where I can watch a demo run, upload my own recording, sample the cube colors from clear frames, and then run the solve reconstruction. It should also keep a history of decode runs so I can reopen past results and inspect the video, overlays, face reads, move sequence, and timeline without rerunning anything. If a recording already has a known starting scramble, let me import that too.

It’s okay if the live decode part needs a CUDA GPU, but the rest should work locally on a normal machine. Please make it easy to run on macOS, Linux, or WSL2, and if possible include a Docker option. Also add a simple labeling tool for people making training data. Use the existing dataset and model files if they’re already part of the project, and look up current docs online if you need to.