Continuously triages NASA FIRMS detections, GOES-19 imagery, and follow-up investigations without waiting for a human to initiate each step.

Each cycle returns both an assessment and the next-cycle decision, while incidents, observations, predictions, and agent decisions are logged separately.

Correlates satellite thermal imagery with geographic and environmental data to analyze spatial patterns and temporal evolution.

Uses data from NWS, USGS, LANDFIRE, Census, OpenStreetMap, and grounded web context to evaluate fire behavior, nearby exposure, and likely persistence. Model-driven analysis is used where the problem stops being clean rules: deciding which weak signals are worth another pass, what context to pull next, and how multiple noisy sources combine into a structured assessment.

Persists raw FIRMS detections, interpreted GOES observations, weather inputs, predictions, and official cross-checks in structured tables.

Assessments, tool calls, incident updates, and prediction outcomes stay visible in the live feed instead of being hidden behind one final answer.