TransitGPT

TransitGPT is a specialized chatbot that helps transit enthusiasts retrieve transit information and analyze General Transit Feed Specification (GTFS) data via natural language instructions. It bridges the gap between complex transit data and user-friendly interaction by generating and executing code to answer queries.

Key Features

• Interactive Chat Interface — Query GTFS data using natural language in a conversational interface.
• Code Generation & Execution — Automatically generates and runs Python code for GTFS analysis in a safe environment.
• Multi-LLM Support — Works with models like Claude 3.5 Sonnet, GPT-4o, GPT-4o-mini, and Llama 3.1 8B Instant.
• Visualization Capabilities — Creates static/interactive maps, plots, and tables using Matplotlib, Plotly, and Folium.
• Multi-Feed Support — Analyzes multiple GTFS feeds from different transit agencies.
• Feedback System — Collects user feedback on responses to improve the system.

Architecture

The system follows a four-step workflow: query moderation, main LLM code generation, safe code execution with retries, and summarization of results into chat responses.

Philosophy

TransitGPT aims to make transit data analysis accessible by allowing users to interact with complex GTFS datasets through simple natural language, eliminating the need for specialized programming skills.