remi2257/sudoku-solver — reverse-engineered prompt

Build me a simple Python app that can solve Sudoku from either a live camera view or a photo in the gallery. I want it to feel like the demo in the README, where the app detects the grid, reads the digits, solves the puzzle, and draws the answer back onto the image so it looks like an overlay on the original puzzle.

Please include two main screens. One should be for live solving from the camera, with both a full solve mode and a hint mode. The other should let me pick an image from disk and solve the Sudoku in that picture. If there are multiple grids in one image, handle that too if the existing code supports it.

Keep the interface clean and easy to run, and make sure the project actually starts from the main app file. If anything is missing, especially model files or setup details, fill in the gaps as reasonably as you can and look up current docs online if needed.