Rajamourya661/Ai-Kavach-Pro — reverse-engineered prompt

Build me a working MVP of KAVACH AI Pro, a web dashboard that helps someone check suspicious videos, audio clips, and links for security risks. I want users to sign up, log in, and see a clean security dashboard with recent scans, risk scores, alerts, and basic analytics.

The app should let a user upload a video for deepfake detection, upload an audio file for voice spoofing detection, and paste a URL for phishing detection. Show the result in plain English with a confidence score, reasons if available, and a way to download a PDF report. Include admin and normal user roles if it’s reasonable.

Use a React frontend with a modern Tailwind style, and a FastAPI backend with PostgreSQL and Redis. Use placeholder or simple ML logic where real models are not available, but structure it so real PyTorch, OpenCV, Librosa, XGBoost, and scikit learn models can be plugged in later. Add Docker Compose so I can run everything locally.