Settings

Values saved here override the server's environment defaults, persist across restarts, and apply to the next lookup - no restart needed. Fields marked override currently differ from the environment.

AI module (via OpenRouter)

All models are served through OpenRouter; the fallback model is used automatically on rate limits or errors. Keys are stored server-side only and never shown again.

Summary modele.g. openai/gpt-oss-120b OpenRouter API key

clear

Fallback summary model Fallback API key

clear

AI document reading read each downloaded PDF for amounts, legal description, signatories

Document-reading model vision-capable; reads the scanned PDFs (e.g. openai/gpt-5-nano) Fallback document-reading model Documents read per run newest first · 1–200

Without a key, summaries fall back to the deterministic template narrative (facts still shown; clearly labelled) and document reading is skipped.

Pipeline

Limits and caching for each lookup run.

Run timeout (seconds) 30–3600 · expiry returns clearly-marked partial results ZIP bundle cap (documents) long histories keep the N most recent in the ZIP

Data sources

NJ parcel layer used for Block/Lot resolution. The state retires/renames these now and then; if resolution fails with a service error, open the current "Parcels and MOD-IV Composite of NJ" item page, copy its service URL (add /0 for the layer), and paste it here. Built-in fallback endpoints are tried automatically.

NJGIN parcel service URL

PRESS portal (Essex County Register)

Scraper behavior against press.essexregister.com. Delays keep the scraping polite - sequential requests, never parallel.

Portal base URL Headless browser off = show the Chromium window (local debugging)

Min delay between actions (s) Max delay between actions (s)

Demo / development

Mock services run the whole flow on bundled fixtures - no network, no portal