a-dera/biligri

An AI-powered WhatsApp bot for real-time fact-checking, specifically designed for West Africa. This system automatically analyzes the veracity of shared content and returns a clear verdict (true, false, or unverifiable) supported by reliable sources. # FactCheck WhatsApp Bot ## 🌍 Overview An AI-powered WhatsApp bot for real-time fact-checking, specifically designed for West Africa. This system automatically analyzes the veracity of shared content and returns a clear verdict (true, false, or unverifiable) supported by reliable sources. ## 🏗️ Architecture User WhatsApp → Bot → AI Server: - Text extraction from links (scraper or LLM) - Text classification (Real vs False) - Fact verification: - Search in databases (AFP, AfricaCheck...) - Semantic similarity with intelligent search engines (Bing Search API + BERT) - Response generation: - Summary + Verdict + Sources + Confidence (%) For more detailed documentation, see docs/architecture.md. ## 🔧 Technical Stack - **AI/NLP**: HuggingFace Transformers (multilingual BERT or CamemBERT) - **Backend**: FastAPI - **WhatsApp Bot**: Twilio API - **Fact-checking**: Bing Search API + semantic analysis, BeautifulSoup - **Content extraction**: newspaper3k, BeautifulSoup - **Semantic similarity**: sentence-transformers + cosine distance - **User response**: Template + NLP / potentially LLM - **Database**: MongoDB - **Deployment**: Docker, AWS Lightsail ## 🚀 Quick Start ### Prerequisites - Python 3.8+ - Docker and Docker Compose - Twilio account for WhatsApp integration - Bing Search API key - MongoDB instance ### Installation 1. Clone the repository: ```bash git clone github.com cd factcheck-whatsapp ``` 2. Build and run with Docker: ```bash docker-compose up -d ``` 3. Or set up the development environment: **On Linux/macOS:** ```bash chmod +x setup.sh ./setup.sh ``` **On Windows:** ```bash setup.bat ``` 4. Configure environment variables: ```bash # The setup script has already copied .env.example to .env # Modify .env with your API keys and configuration ``` 5. Run the API: ```bash uvicorn app.main:app --reload ``` ## 📖 Documentation - API Documentation: Available at `localhost` locally - Detailed …