How h1ppox99/hackathon_mit_mcp-or works
Project type and purpose
• This is a full-stack, data-driven web application (“Optimal Freight Route Planner”)
• Backend (Python + FastAPI via uvicorn)
• Solver engine (gurobipy LP model)
• Data preprocessing modules (CSV, GeoPackage, Java JAR for searoute)
• Frontend (HTML/CSS/JS, Node.js/npm)
• External services (Claude via MCP, searoute CLI)
File structure
• Top level
• main.py: FastAPI entrypoint
• solver.py: encapsulates Gurobi model
• normalize_routes.py, calculate_means.py, csv_editor.py, etc.: utility scripts
• data/: raw & intermediary CSV/GeoJSON/GeoPackage files
• src/data_preprocess/: ETL pipelines, add_city, maritime/road utilities, searoute JAR & scripts
• index.html, page2.html, script.js, styles.css: frontend UI
• package.json, node_modules: frontend dependencies
• send_to_claude(.py/.sh): integration with Claude via MCP server
Key architecture insights from README
• Backend serves REST API to frontend and handles on-the-fly city additions
• Solver module solves an LP based on user preferences (time/price/CO₂)
• Frontend visualizes routes on a map, accepts dynamic user inputs
• Claude integration for real-time data ingestion (traffic, weather)
• Data pipeline generates and normalizes multimodal graph (road/air/sea)