The facts are never the language model’s. Questions are matched to what the engine can actually answer — a route, a water depth, a shelter recommendation — and the numbers come out of that result. It cannot invent a road that does not exist.

No model key is configured, so you are reading the engine’s own wording. Add GROQ_API_KEY to .env.local and restart to have answers phrased conversationally. Every number stays the engine’s either way.

Answers are computed against the timeline position shared across the whole app — currently Day 1 · 04:00. Scrub the timeline on the map and ask again; the answer changes.