Ishita-dutta12/Universal-Image-Forensic-Analysis-Framework ? reverse-engineered prompt

Build me a Python app that can analyze an image and tell me whether it looks authentic or likely AI generated, edited, or forged.

I want it to accept either an uploaded image or an image URL, run a full forensic check, and return a clear JSON report with the final verdict, confidence scores, likely manipulation type, metadata findings, and any visual explanation like a heatmap or highlighted regions. It should feel like a modular forensic pipeline, so the analysis can combine image features, metadata analysis, attention based reasoning, localization, and an evidence summary into one result.

Please make it work as a FastAPI service with a simple API and Swagger docs, and also include a script or entry point I can use locally for inference and another one for training if needed. Keep the code organized and easy to extend, since I may want to add more detection types later like deepfakes, copy move forgery, splicing, object removal, and screenshot detection. If you need current library details, look up the latest docs online.