NYC 311 JEV Heatmaps

It downloads a bounded slice of NYC 311 reports, evaluates complaint descriptions with JEV, and maps the resulting signals on a projected geographic grid. It caches repeated descriptions, validates model responses, and produces interactive HTML maps and print-ready PNGs.

Across all boroughs on a 500-meter grid with JEV jev-1.13.0: 9,903 reports downloaded, 9,698 mapped, 2,043 occupied cells, and 634 distinct descriptions. The run excluded 205 reports with missing or invalid coordinates and completed 634 API calls without retries, reporting 539,979 input tokens.

Complaint count shows reports per occupied cell; scoring coverage shows the fraction with a model score; severity is the mean normalized position on the JEV rubric; safety concern is the mean probability of the defined hazard proposition; ambiguity is a per-report proxy; the composite index combines severity, concern, and volume illustratively.

No. The README's research context says the maps show reported complaints and model-estimated potential impact, not verified hazards, actual harm probabilities, or population-adjusted safety risk, and blank cells are not evidence of no problem.

Download the NYC borough boundary GeoJSON, then pass --background-geojson data/nyc_borough_boundaries.geojson when generating heatmaps. Serve results over localhost with python -m http.server 8765 --bind 127.0.0.1 --directory results rather than opening files directly, so tile requests carry a web referrer.

The README reports usage rather than prices: 634 API calls and 539,979 input tokens for the September 1 run, with repeated descriptions sharing a single evaluation. Reusing the cache with --max-api-calls 0 avoids new scoring calls. A JEV API key is required for live runs.

It is a research demonstration by CompleteTech LLC AI Research, published with methodology notes and a TESTING.md verification page. The README gives no license, and the caveats above are the authors' own framing, so treat it as an explorable research artifact rather than an operational risk system.