US Code Violations — 9 jurisdictions, one case model

Property, housing and building code violations from nine US jurisdictions, normalised into a single schema.

~4,249,022 violation records across 9 actively-maintained sources.

Source	State	Records	Last verified
Chicago	IL	~2,020,658	2026-07-29
Montgomery County	MD	~954,255	2026-07-22
Chicago (ordinance hearings)	IL	~838,859	2026-07-30
Kansas City	MO	~174,999	2026-07-01
New Orleans	LA	~111,441	2026-07-30
Mesa	AZ	~80,580	2026-07-28
Norfolk	VA	~44,835	2026-07-29
New York City (BIC)	NY	~18,026	2026-07-29
Cambridge	MA	~5,369	2026-07-22

Why this is harder than it looks

The other actors in this repo reconcile spellings. This one reconciles processes. All nine publish a "status" field, and they are not describing the same thing:

Jurisdiction	Field	Vocabulary	What it actually means
Chicago	violation_status	OPEN / COMPLIED / NO ENTRY	violation state
Montgomery County	disposition	Triennial Completed / Citation issued	inspection outcome
Kansas City	case_status	In-Compliance / In-Violation	case state
Norfolk	status	Closed - Corrected / Work in Progress	case state
NYC BIC	violation_disposition	Settled / Barred by CPLR / Judgement	court outcome
Cambridge	violation_status	Corrected / Cited	violation state
Chicago ordinance	case_disposition	Liable / Non-Suit / Default / Not Liable	hearing outcome
New Orleans	currentstatus	Violations Abated / Payment Received	abatement pipeline
Mesa	status	In Violation / Legal Action / Closed	case state

Collapsing those into one enum would look tidy and be a lie — a court settlement is not a re-inspection result. So the schema carries two orthogonal fields:

case_status  OPEN | CLOSED | UNKNOWN
resolution   CORRECTED | CITED | DISMISSED | NO_ACCESS | OTHER | NONE

plus raw_status (always the original string) and status_semantics (case / inspection / legal) so you never silently compare a court judgement against a housing re-inspection.

The mapping is substring-based and order-sensitive: "Closed - Corrected" contains both closed and corrected; "Not Corrected" contains corrected; "Settled - Pending" contains settled; and "Not Liable" contains "Liable", where a flipped order would turn ~107k dismissals into citations. test_status.py pins all 43 observed strings.

Four traps found while building this

1. Montgomery County is mostly not violations. Its 954k rows are dominated by routine triennial inspections (Triennial Completed, 547k of them). Filter resolution = CITED to find actual enforcement. Auto-inference also picked date_closed as the event date, which would have silently excluded every open case — we use date_filed.

2. Kansas City's dates run into the future. date_found contains data-entry typos as far ahead as October 2029. Set violatedBefore if you are building time series. verify.py flags this rather than reporting the feed healthy.

3. Chicago has two violation feeds. chicago is building violations; chicago_ordinance is administrative hearing outcomes — what the city actually did about them. Its violation_date contains typos reaching year 6111, so always set violatedBefore.

4. NYC BIC is not what you expect. It is the Business Integrity Commission — trade waste and public markets — not the Department of Buildings. Narrow, but the only source here with monetary penalties and court dispositions.

Who this is for

• Real-estate investors sourcing distressed-property leads
• Contractors (roofing, façade, abatement) finding properties under order
• Property managers and compliance vendors monitoring portfolios
• Journalists and researchers tracking enforcement patterns

Usage

Open violations with an address, last quarter:

{
  "sources": ["chicago", "norfolk", "cambridge"],
  "violatedAfter": "2026-05-01",
  "openOnly": true,
  "requireAddress": true,
  "maxResults": 5000
}

Actual enforcement actions only (skips routine inspections):

{ "violatedAfter": "2026-01-01", "violatedBefore": "2026-07-31", "resolution": ["CITED"] }

Local development

pip install requests
python local_run.py --coverage
python local_run.py --after 2026-05-01 --limit 10
python local_run.py --open-only --after 2026-01-01 --limit 20
python local_run.py --resolution CITED --after 2026-01-01
python verify.py         # health-check every source
python test_status.py    # status-classification regression tests

Legal

Public, logged-out municipal open-data endpoints via each jurisdiction's official Socrata API. No authentication, no account, no terms accepted, no anti-bot circumvention. Code violations are public records attached to properties; property_owner is populated only where the jurisdiction publishes it as part of that public record. Rate limits respected with exponential backoff.