Scoring model
The scoring logic is simple enough to understand and specific enough to feel trustworthy. It turns weather into a strong call, not a black box.
- Open-Meteo inputs feed a 0-100 score tuned to the sunset window
- High clouds, mid clouds, low-cloud penalties, rain, humidity, and visibility all move the result
- Fallback forecasts keep the experience usable even when the API is down
Visual product design
The product wins on interface as much as on logic. The number only matters because the UI makes it legible, local, and memorable.
- Canvas sky scenes respond to the same data that drives the score
- Forecast cards, reason chips, and labels make the outlook instantly scannable
- The main surface feels like a local consumer app, not a weather console
Explainer loop
The simulator and post-sunset feedback keep the model from feeling like a magic number. Users can see why the call moves and feed reality back into the product.
- Sliders let people see how haze, clouds, and atmosphere change the score
- Scrollytelling explains why certain sunset conditions look better
- Post-sunset feedback is stored locally to calibrate future calls
