fiery-rumor

by46_SE-C_Prathamesh Vishwakarma

Role: Lead Full-Stack Engineer & AI Architect. Project Name: VerifyAlert — An AI-driven crisis rumor verification engine. Tech Stack Requirements: Frontend: HTML5, CSS3 (modern, responsive UI), and Vanilla JavaScript. Backend: Python 3.11+ with Flask. Database: SQLite for storing request history, results, and confidence scores. Core AI Stack: Google Gemini (google-genai): For Stage 1 (Translation and Entity Extraction). Use gemini-3-flash-preview. NewsData.io API: For Stage 2 (Live News Retrieval). Groq (groq SDK): For Stage 3 (Final Verification using Llama 3 70B/8B). The 3-Stage Pipeline Logic: Stage 1 (Gemini - Process): Accept user input. Use Gemini to: Detect the source language. Translate to English if needed. Extract structured JSON: {"event": str, "location": str, "date": str, "keywords": list}. Stage 2 (NewsData.io - Retrieve): Use the extracted event and location as query parameters for the newsdata.io/api/1/news endpoint to fetch the top 3 related articles. Stage 3 (Groq - Analyze): Provide Groq with the translated rumor + the text snippets from the news articles. Instruct Groq to return: Verdict: (TRUE / FAKE / UNVERIFIED) Confidence: (0-100%) AI Summary: A 2-sentence explanation of the logic. UI/UX Branding: Navbar: "VerifyAlert" with a clean, high-tech logo. Input Area: Centered "Paste Rumor" box. Dashboard View: Results page must show the "Transparency List" (clickable news source titles and URLs) below the verdict card. Development Instructions: Please provide the app.py script using the new google.genai Client, the HTML templates (index.html and result.html), and a clear .env example for the three API keys.