RESEARCH

Legal and geospatial research sources

Open http://127.0.0.1:8903/research/ (or /research/ on the hosted app) for the research panel. After a job finishes, FlexSearch 0.8.212 builds a browser-memory index of that job’s saved JSON artifacts, protected by the workspace and per-job tokens. It does not download or index every historical file at once; use Basin SQL for archive-wide search. FlexSearch is Apache-2.0; its license is in static/FLEXSEARCH-LICENSE.txt.

Requests are authorized with the Harvy workspace token; source credentials are read only from server environment variables.

Configure server-side credentials

Variable	Used for
SPIDER_API_KEY	Spider public-page collection
WEBSHARE_PROXY_URL	Proxy for Spider target requests and direct API calls
JUDYRECORDS_API_KEY	JudyRecords official API
COURTLISTENER_API_TOKEN	Optional CourtListener REST API token; the API also permits limited unauthenticated searches
CASELAW_API_TOKEN	Optional Case.law API token
DOCKETALARM_USERNAME, DOCKETALARM_PASSWORD	Docket Alarm API login
MAPILLARY_ACCESS_TOKEN	Mapillary Graph API
BASIN_INGEST_URL, BASIN_INGEST_TOKEN	Optional Basin archive; see BASIN.md

Do not put API keys in templates, job arguments, or browser requests. Restart the server after changing its environment.

Source behavior

• CourtListener: REST API search, capped at one page of ten results per job. It is rate-limited; configure a token for ongoing use.
• Case.law: attempts the documented REST endpoint. The endpoint returned HTML rather than JSON during implementation checks, so the report marks this source unavailable unless it returns JSON.
• JudyRecords: uses its documented /search API via Spider, with the API key sent in a server-side request header. The site’s terms state its API is the permitted programmatic access path.
• Docket Alarm: API search only when credentials exist. The connector uses indexed search; it does not fetch live PACER data or download documents.
• Arrests.org: public search page via Spider + the configured Webshare proxy, using the supplied state and first/last name. It saves the raw response; any result must be independently checked against the court source.
• CourtRecord.net: link-out only. Its posted terms restrict copying and commercial reuse of its materials.
• Google Scholar: link-out only; no official public search API was verified.
• Mapillary: image-radius API query (50 m, up to 10 results). Creator identifiers are not requested.
• Google Maps: python -m v1.scrape.google_maps <https://www.google.com/maps/place/...> --out <folder> uses Spider’s /unlimited/scrape endpoint and the shared proxy secret helper. It preserves the raw response and normalizes only exposed JSON-LD business/review data. It does not call Spider’s metered /fetch Maps extractor; review coverage remains partial and explicitly marked incomplete.
• Zoom Earth and OpenSfM: links only. Zoom Earth has no public API documented; OpenSfM is a local photogrammetry toolkit, not a hosted API.

Search names can match different people; an arrest record is not proof of guilt. Court and legal data can be incomplete or stale. Restrict access to downloaded archives and Basin tables, and set appropriate retention before enabling persistent storage.