Restaurant Inspection Scores & Violations (Official City Data)

Official government restaurant / food-service health inspection results — pulled straight from each city's open-data portal (no scraping, no third-party aggregators) and normalized into one unified schema across cities.

Currently covered (every source verified live against the portal's API):

City	Source (official)	Dataset	Freshness at build (2026-07-19)	Row grain
New York City, NY	NYC DOHMH via NYC Open Data	43nn-pn8j	newest rows 2026-07-16	one row per cited violation
Chicago, IL	Chicago Dept. of Public Health via Chicago Data Portal	4ijn-s7e5	newest rows 2026-07-16	one row per inspection
Austin, TX	Austin Public Health via City of Austin Open Data	ecmv-9xxi	newest rows 2026-05-22	one row per inspection
Cincinnati, OH	Cincinnati Health Dept. via Cincinnati Open Data	rg6p-b3h3	newest rows 2026-07	one row per cited violation

Unified record schema

{
  "city": "Chicago",
  "state": "IL",
  "business_name": "OLD TOWN POUR HOUSE",
  "address": "120 N LA SALLE ST CHICAGO IL 60602",
  "inspection_date": "2026-07-16T00:00:00.000",
  "inspection_type": "License",
  "result_or_grade": "Fail",
  "score": null,
  "violations_summary": "3. MANAGEMENT, FOOD EMPLOYEE ... (truncated at 2000 chars)",
  "source_url": "https://data.cityofchicago.org/resource/4ijn-s7e5.json",
  "raw": { "...every original field from the city, verbatim..." }
}

raw always carries the complete original record (minus phone numbers), so nothing the city published is lost — the unified fields are a convenience layer on top, never a replacement.

Honest coverage notes (read before comparing cities)

Cities grade differently; this actor does not translate between grading systems — it reports each city's own fields verbatim:

• NYC — numeric score where lower is better (0 = clean), plus letter grade (A/B/C; N/Z/P are pending states; often empty until a grade is assigned). One row per cited violation: an inspection with 3 violations = 3 rows sharing date/score/grade.
• Chicago — result_or_grade = Pass / Pass w/ Conditions / Fail / Out of Business / No Entry… No numeric score. Full violation narrative in violations_summary (truncated at 2000 chars; full text always in raw.violations).
• Austin — numeric score where higher is better (100 = perfect). No letter grade or pass/fail, and no violation detail in the city's feed.
• Cincinnati — result_or_grade = the inspection's action status (Approved / Approved - Minor Violations / Not In Compliance…). No numeric score. One row per cited violation, with Ohio food-code section, description, and inspector comments.

So: do not rank a NYC 10 against an Austin 90 — they are different scales in opposite directions. Compare within a city, or use result_or_grade.

Input

Field	Type	Default	Notes
cities	array	all 4	nyc, chicago, austin, cincinnati
businessNameContains	string	—	case-insensitive substring on the establishment name (server-side)
resultContains	string	—	case-insensitive substring on result/grade, e.g. fail; skips Austin (score-only) with a warning
since	string	—	ISO date YYYY-MM-DD; only inspections on/after (server-side)
maxRecordsPerCity	integer	1000	newest first

Records are always returned newest first, so capped runs surface the most recent inspections. If a city portal fails mid-run, the actor fails open: it keeps everything already collected, logs a loud warning naming the city, and continues with the other cities.

Who uses this

• Acquisition / franchise diligence — pull the full inspection history of a target restaurant (or every location of a brand via businessNameContains) before a deal.
• Food-safety analytics & insurers — track failure rates, repeat violations, and closure actions across markets.
• Journalists & researchers — closures (action in NYC raw), failed inspections, pest violations, neighborhood patterns.
• Restaurant-tech & delivery platforms — surface official grades next to listings.

Data custody

Values are verbatim from the official source — no inference, no cleanup beyond whitespace/assembly of address parts, no cross-city score translation. Phone numbers are stripped. Each record carries source_url so every row is independently checkable against the city portal.