OpenBMB/MiniCPM-OCR ? reverse-engineered prompt

Build me a clean web demo for MiniCPM OCR that lets a user upload an image or PDF and get back readable text, tables, formulas, and a simple review summary showing which parts look reliable and which parts might need checking.

I want it to feel easy for someone who is not technical, so the page should have a clear upload area, a progress state while it works, and a results view that shows the extracted content in a nice readable format. If it helps, include language switching for Chinese and English, and make sure the output can be copied or downloaded.

Please make it work well for scans, photos of documents, skewed pages, and messy inputs. If you need current docs or examples, look them up online. Keep the UI polished and practical, and make sure the app is ready to run locally and easy to self host.