bharath-0814/Kavach.ai ? reverse-engineered prompt

Build me a small web app that can spot scammy SMS messages in Indian languages and Hinglish, especially messages that try to trick people with fake KYC updates, OTP requests, account blocks, urgent threats, or suspicious links.

I want a simple page where someone can paste an SMS and get back a clear result like safe, suspicious, or high risk, along with a short explanation of why it was flagged. It should handle obfuscated text too, like leetspeak, spaced out words, and weird spellings, not just normal English.

Please also make an API endpoint for classifying messages and another one for showing the most recent flagged messages. Use Google Gemini for the AI part, and store recent results in a database so the logs can be viewed later. Make it work locally and for deployment on Vercel, and if you need current docs for anything, look them up online.