TransitFlow is a backend API built with FastAPI that reads GO Transit GTFS files (CSV format) and provides clean, developer-friendly endpoints for:

• Listing and searching routes
• Listing and searching stops
• Predicting crowd levels (prototype model for now)

For now, all data is read directly from CSV files so it’s quick and easy to test without setting up a database.

• Python 3.11+
• FastAPI – high-performance API framework
• Uvicorn – ASGI server
• Pydantic – data validation & serialization
• GTFS CSV – public transit data format (routes.txt, stops.txt, etc.)

Right now, the predict endpoint uses a placeholder. The goal is to replace it with a trained model that considers:

1. Historical ridership patterns from CSV/GTFS + public reports
2. Time-based trends – weekday rush hours, weekend lows
3. Stop-specific patterns – since some stops are always busier (based on real rider experience)
4. Special events & anomalies – game days, holidays, weather

Example model approach:

• Start with rule-based logic (e.g., “If weekday 7–9 AM and stop in downtown core → high crowd”)
• Transition to a machine learning model once enough data is collected
• Use personal ride knowledge as seed training data to bootstrap predictions before large-scale datasets are integrated