🛡️ VeriNewsAI - AI-Powered News Verification System

VeriNewsAI is a sophisticated, production-ready system designed to combat misinformation. By leveraging hybrid AI models, it analyzes news articles in real-time to determine their credibility.

---

✨ Key Features

• 🔍 Intelligent URL Extraction: Seamlessly extracts core article content from any URL using the newspaper3k library.
• 🤖 Hybrid Multi-Model Scoring:
  • AI Core (70%): Powered by Hugging Face's BERT-based fake news detection models.
  • Suspicion Engine (30%): Heuristic keyword-based analysis to catch common misinformation patterns.
• 📝 LLM Explainer: Provides human-readable, context-aware explanations for every score using Google Gemini 1.5 Flash.
• ⚡ High Performance:
  • In-memory caching for repetitive requests.
  • Predictive model warm-up on server startup.
• 🛡️ Robust Design: Structured logging, request timeouts (10s), and a reliable fallback mechanism when AI services are unreachable.

---

🛠️ Tech Stack

Backend

• Core Framework: FastAPI (Asynchronous, High-performance)
• AI Models: Hugging Face Inference API (BERT), Google Gemini 1.5 Flash
• Web Scraping: Newspaper3k
• Data Validation: Pydantic v2
• Utilities: Python-dotenv, Uvicorn, Requests

Frontend

• Interface: Clean, responsive HTML/JS dashboard
• Architecture: Statically served from the API backend

---

🚀 Getting Started

Prerequisites

• Python 3.9 or higher
• API Keys for Hugging Face and Google Gemini

Installation

1. Clone the Repository

   git clone <repo-url>
   cd VeriNewsAI

2. Setup Virtual Environment

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install Dependencies

   pip install -r requirements.txt

4. Configuration Create a .env file in the root directory:

   HF_API_KEY=your_huggingface_inference_api_key
   GEMINI_API_KEY=your_google_gemini_api_key
   PORT=8000

5. Fire it up!

   python main.py

   Server will start at http://localhost:8000.

---

🔌 API Reference

Health Check

GET /health

• Description: Returns the system status and model readiness.

Analyze News

POST /analyze

• Body:

  {
    "url": "https://example-news.com/article-123",
    "detailed": true
  }

• Sample Response:

  {
    "result": "Fake",
    "score": 88,
    "confidence": "High",
    "explanation": "The article uses highly sensationalized language and lacks citations from verified sources...",
    "source": "url",
    "processing_time_ms": 1150
  }

---

📂 Project Structure

VeriNewsAI/
├── frontend/          # Web dashboard (HTML/JS/CSS)
├── routes/            # FastAPI API routers
├── services/          # Business logic (Scraper, AI Detector, Explainer)
├── utils/             # Logging, Caching, and Text Processing
├── main.py            # Application entry point & configuration
├── requirements.txt   # Backend dependencies
└── render.yaml        # Deployment configuration

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Developed with ❤️ for a more informed world.