VarunB453/Lumidoc — reverse-engineered prompt

Build me Lumidoc, an AI document and media app where people can sign up, upload PDFs, audio, and videos, then ask questions about their own files.

I want a clean web app with login, a dashboard, an upload page, a file library, a chat page, and settings. Uploaded files should be processed in the background. PDFs should have text extracted, audio and video should be transcribed, and each file should get a useful summary plus timestamps or topics when possible. The chat should answer from the uploaded content and show the sources it used so users can trust the answer.

Use the existing idea of a React frontend and Python FastAPI backend, with workers for long processing jobs, MongoDB and Redis, and local file storage for development. Make it runnable with Docker Compose and include sensible environment examples. Please wire up the main flow end to end, even if some AI provider keys are placeholders. Look up current docs online if you need to.