djpapzin/malawi-rag-sql-chatbot

# Dziwani - Infrastructure Transparency Chatbot A specialized chatbot for querying and exploring Malawi's infrastructure projects database. Named "Dziwani" (meaning "what you should know" in Chichewa), this tool provides a natural language interface to access detailed information about infrastructure projects, their status, and related statistics. ## Features - Natural language querying of infrastructure projects - Interactive chat interface with guidance tiles for: - Finding projects by sector (Infrastructure, Water, Energy, etc.) - Finding projects by location (districts) - Finding specific project details - Detailed project information including: - Project Name - District - Project Sector - Project Status - Budget (in MWK) - Completion Percentage - Start and Completion Dates - Real-time chat responses - Loading states and error handling - Responsive design for all devices ## Technical Architecture ### Frontend Components The frontend of Dziwani uses a lightweight, server-side rendered approach: - **Technology Stack**: - HTML with Jinja2 Templates - Vanilla JavaScript (no framework) - Plain CSS for styling - No build process required - **Key Files**: - `frontend/templates/index.html` - Main chat interface - `frontend/static/js/main.js` - Core application logic - `frontend/static/css/styles.css` - Main styling - `frontend/static/css/loading.css` - Loading animations - **Features**: - Responsive design for mobile and desktop - Client-side form validation - Interactive chat history - Expandable query details for technical users - Loading indicators and error handling ### Backend Components The backend is built on FastAPI with LangChain for RAG-based SQL generation: - **Technology Stack**: - FastAPI web framework - SQLite database - LangChain for RAG-SQL capabilities - Python 3.11+ - Jinja2 for templating - **Key Components**: - `app/main.py` - Entry point and API configuration - `app/database/langchain_sql.py` - SQL generation with LangChain - `app/model …