is-project-4th-year/AgriScanDetection — reverse-engineered prompt

I want an Android app called AgriScan that helps farmers diagnose leaf diseases without needing internet. The app should let someone take a photo or pick one from their gallery, then run the diagnosis right on the phone for tomato, potato, and bell pepper leaves. Show the top 3 likely diseases with confidence scores, then open a simple advice screen with symptoms, causes, spread, and practical next steps from a local knowledge file.

Please make the flow feel like capture, diagnose, explain, and save. I also want users to organize observations by field or plot, and have a library screen where they can review past scans, images, predictions, and notes. Keep everything stored locally on the device, including images and history, so it stays private and works in low connectivity areas.

Make the interface clean and farmer friendly, and wire it up to the included model, labels, and local knowledge base. If anything is unclear, look up current Android docs online.