shahu285/clearance-ai-security-engine — reverse-engineered prompt

Build me a working demo called ClearanceAI. I want a simple web app where someone can upload a PDF or TXT document, have Gemini read it and decide what security roles should be allowed to see it, then store the document chunks in a vector database with that security info attached.

The main use case is secure document search. A user should be able to choose or type their current role, ask a question, and only get matching document snippets they’re allowed to see. If they ask about something relevant but their role doesn’t have clearance, the app should clearly say the result was blocked instead of leaking the text.

Please make it in Python with FastAPI, a clean HTML and Tailwind style dashboard, Qdrant running in memory, FastEmbed for local embeddings, and Gemini 1.5 Flash for the document classification. Support PDF and TXT upload. Keep setup simple with a requirements file, a .env for the Gemini key, and instructions to run locally. Look up current docs online if you need to.