About This Project

What is this system?

This is a fully autonomous AI agent designed to act as a digital detective for bKash. Its sole purpose is to automatically patrol the internet, identify websites that are potentially misusing the bKash name to promote betting or gambling scams, and present these findings in a clean, focused, and user-friendly web dashboard.

It is a zero-maintenance system; once deployed, it runs entirely on its own without needing a dedicated computer or any manual intervention.

How It Works: The Waterfall Method

The agent is designed for maximum reliability using a "waterfall" fallback system. If one tool fails, it automatically tries the next.

Scheduled Hunt: Every morning, the powerful scheduler at GitHub Actions wakes up two separate micro-agents (one for English, one for Bangla).
Multi-Tool Search: The agent first tries to get search results from SerpApi. If that fails, it automatically falls back to using ScrapingBee to ensure the hunt is always successful.
Multi-Tool Investigation: For each URL found, the agent tries to read its content using ScraperAPI. If that fails, it falls back to ScrapingBee, and finally to a direct request. This ensures that even difficult-to-reach sites are investigated.
The AI Verdict: The extracted text is sent to Google's Gemini AI, which makes the final decision on whether the site is a potential threat.
Intelligent Reporting: Only sites flagged by the AI as genuine threats are ever saved to the database, keeping the final dashboard clean, focused, and actionable.

Technology & Design

The system was designed for reliability, focus, and zero maintenance. By using a "Serverless Architecture" with multiple, redundant tools for each step, the agent is robust, scalable, and requires no ongoing management.

Key Technologies: Vercel (Dashboard Hosting), GitHub Actions (Scheduling & Automation), Supabase (Database), Google Gemini (AI Model), SerpApi, ScraperAPI, & ScrapingBee (Web Data), Python (Agent Logic), and HTML/CSS/JavaScript (Dashboard).