facebookresearch/segment-anything ? reverse-engineered prompt

Build me a simple app and codebase for Segment Anything that lets me load an image, click on it with a few points or draw a box, and get clean object masks back right away. I also want an automatic mode that finds and shows masks for everything in the image, plus a command line way to run it on one image or a whole folder and save the results.

Please include a couple of easy notebook examples so I can try it step by step, and make it possible to use a downloaded model checkpoint. If it helps, set it up so the mask part can run in the browser too with an exported model. Keep the experience straightforward, with clear outputs, simple instructions, and a small demo page I can open locally to test segmentation quickly. If you need to check current docs for the export or browser setup, go ahead and look them up online.