mandeepsingh2007/Multi-Modal-Document-Intelligence-System ? reverse-engineered prompt

Build me a document intelligence app that can take in PDFs and actually understand them, not just read plain text. It should handle normal text, scanned pages, tables, charts, and images, then give me a clear answer summary and confidence score after processing.

I want a simple API where I can upload a PDF, get back extracted insights, and then ask questions about the document later with semantic search over everything that was ingested. If a page is scanned, it should use OCR, and if there are visuals like charts or layouts, it should analyze those too. Please make the workflow feel smart and self correcting, so it can combine visual and text based results and flag anything uncertain.

Use the README as the guide, keep it runnable locally and in Docker, and if you need to check current docs for any libraries, feel free to look them up online.