How it works
Portland Transit AI combines four live signals to predict whether your TriMet arrival will actually be on time.
🚌 Live vehicles
We pull TriMet's live vehicle feed — position, schedule drift, and a "in congestion" flag set by the bus's onboard GPS.
📡 Real-time arrivals
TriMet's own predicted ETAs vs scheduled times tell us how much each trip is already drifting.
🌧️ Weather
Open-Meteo gives us current conditions and the next-hour forecast for Portland. Heavy rain, snow, and wind all hurt reliability.
🕐 Time of day
AM/PM rush hour, weekends, late night — each affects road congestion differently.
The AI layer
A reasoning model (Lovable AI · Gemini) reads all four signals together and produces a structured prediction: a reliability label (on time / likely late / high risk), a confidence score, the predicted delay, and a plain-language explanation. Same engine powers congestion hotspots and the commute risk score.
Limits
- Predictions are estimates, not guarantees.
- Reliability quality depends on TriMet feed freshness.
- This v1 doesn't yet learn from accumulated history — that's next.