Methodology

This map's data is refreshed monthly via an automated pipeline. Source records are fetched from primary sources, structured by Claude (Anthropic's large language model) against a typed schema, and stored as individual objects in a Cloudflare R2 bucket. A GitHub Actions workflow then assembles the records into vector tiles using tippecanoe, uploads the resulting PMTiles file to R2 in place, and triggers a Cloudflare Pages rebuild.

Each record retains a provenance trail in its properties: the source URL, the extraction timestamp, and the model used for structuring.

Refresh cadence

Data is refreshed on the 1st of each month at 03:00 UTC. Trigger a manual refresh via the GitHub Actions tab; the ingest-cron.yml workflow supports workflow_dispatch.