AxiomCodex/axiom-permit-signals-sample-25k

Axiom Permit Signals (25K Sample)

Readiness Status

Public research dataset of upstream open data; see license and attribution below; no commercial or training authorization is implied.

Summary

A 25,000-row research sample of municipal building permits from four U.S. city open-data portals: Chicago, Los Angeles, New York City (Department of Buildings) and Seattle. Each row is one permit record normalized to the Axiom Codex Permit Signals schema. It has the native permit number, controlled permit type, scope and status (with the exact source status kept alongside), issue and application dates where the city publishes them, cost estimate and cost band, and an H3 resolution-8 cell for spatial joins.

The sample does not establish permit outcomes, adjudicated labels or a complete permit lifecycle.

Part of the Axiom Codex Family

Axiom Codex is the governed public-data catalog maintained by Axiomancer Labs (www.axiomcodex.io). Every Axiom Codex sample is normalized to a published schema and filtered to sources whose redistribution terms are recorded in the Codex source registry. It ships with machine-readable quality evidence (quality.json, baseline.json) and a manifest that binds each file to its upstream collection.

Full data

This sample is a preview. The full open-terms Permit Signals package covering Seattle and Los Angeles is part of Axiom Codex Commercial All-access, delivered on request after purchase. Chicago and New York City are not part of that package. See the Axiom Codex catalog. This sample grants no commercial rights.

Quick Start

from datasets import load_dataset

ds = load_dataset("AxiomCodex/axiom-permit-signals-sample-25k", split="train")
df = ds.to_pandas()
print(df.groupby(["jurisdiction_slug", "status"]).size())

Schema

Column	Type	Description
acl_tier	string	Governed source-row tier; must already be research
applied_at	string	Source-reported application date, when present
building_type	string	Nullable source building classification; populated only where a feed maps it (Los Angeles), otherwise null
closed_at	string	Source-reported close date, when present
confidence_score	double	Collector confidence score
doc_id	string	Stable upstream permit record identity
estimated_cost_tier	string	Source-derived cost band
estimated_cost_usd	double	Nullable source-reported cost estimate
ingested_at	string	Upstream collection timestamp
h3_index	string	H3 resolution 8 cell derived from source coordinates
jurisdiction_slug	string	Normalized city and state identifier
modified_at	string	Upstream normalization timestamp
normalization_version	string	Collector normalization version
occurred_at	string	Permit milestone: issue date, else application date, else the collection timestamp when the source reports neither
permit_number	string	Native jurisdictional permit number
permit_type	string	Controlled permit category from the reviewed per-feed vocabulary map
record_id	string	Stable normalized source-record identifier
sample_row_id	string	Deterministic content-derived identity for the exported permit observation
schema_version	string	Collector schema version
scope_type	string	Controlled scope classification from the reviewed per-feed vocabulary map
source_feed	string	Upstream feed identifier
source_system	string	Logical source system
source_uri	string	Source record URL
status	string	Controlled lifecycle status from the reviewed per-feed vocabulary map
status_raw	string	Exact source-native status value retained for audit
issued_at	string	Source-reported issue date, when present
unit_count	integer	Nullable source-derived unit count; populated only where a feed maps it (Los Angeles), otherwise null
valuation_usd	double	Nullable valuation amount; populated only where a feed maps it (Los Angeles), otherwise null

Sampling

The exporter keeps only research-tier rows from the registered feeds that carry an H3 cell; rows without usable source coordinates are dropped. It collapses repeated observations with the same configured source fields, then ranks sample_row_id values by SHA-256 and uses largest-remainder allocation across source_feed, so each city contributes in proportion to its clean rows. Cities are deliberately not balanced: rows that fail any quality rule are dropped, never repaired. The staged manifest binds each input parquet and collector manifest.

City	Rows	Milestone period (occurred_at)
Seattle (seattle-permits-socrata)	7,189	2024-08 to 2026-10
New York City (nyc-dob-now)	6,233	2000-01 to 2026-07
Chicago (chicago-permits-socrata)	5,799	2006-01 to 2026-10
Los Angeles (los-angeles-permits-socrata)	5,779	2013-01 to 2023-05

Demo Quality Contract

This is a source_records artifact under the Axiom Codex demo quality standard (DEMO_QUALITY_STANDARD.md) permit-signals profile. A successful transport export is not a quality-pass claim: staging must emit quality.json and baseline.json with every critical gate passing; unverified and model-derived fields are not promoted to benchmark labels by this card.

Coverage & Caveats

• Different periods per city. The Los Angeles open-data permit feed has not been updated since May 2023. Seattle rows come from its current permit feed (August 2024 onward); Seattle rows that report neither an issue nor an application date take the collection timestamp as occurred_at. Do not compare cities over time without aligning periods.
• Status is a snapshot. status is the permit's status when collected, not a verified outcome. Chicago, Los Angeles and Seattle do not publish a close date, so closed_at is null there even for closed permits.
• Field availability by city (share of rows missing the value):

Field	Chicago	Los Angeles	NYC	Seattle
estimated_cost_usd	8.1%	0%	0%	0%
valuation_usd	100%	0%	100%	100%
building_type	100%	65.8%	100%	100%
unit_count	100%	79.4%	100%	100%
applied_at	0%	100%	0%	28.3%
issued_at	0%	0%	21.1%	44.4%

The release audit enforces a maximum missing rate per city: 15% for estimated_cost_usd. valuation_usd, building_type and unit_count are disclosed rather than capped, because most feeds do not publish them. permit_type, scope_type, status and estimated_cost_tier are required on every row.

• Dropped rows. Rows whose permit type or status has no reviewed mapping, and duplicate permits, are dropped at collection. About a third of fetched Chicago records have no permit status and are excluded, so Chicago over-represents permits that report one. Dallas is a registered feed but publishes no status field, so it is not in this sample.

Provenance & Source Licensing

This release contains rows from four municipal open-data feeds (Chicago, Los Angeles, NYC and Seattle), each under its own documented terms. The table lists every registered permit feed; only the four marked "yes" contribute rows. The Hugging Face other metadata names this mixed source posture; it does not combine or replace the per-feed terms below. Per-feed collector manifests and input parquet digests are recorded in the staged manifest.

Los Angeles rows come from LADBS under the Creative Commons Attribution 4.0 International license. Axiom Codex modified them: records were normalized to this schema, permit type, scope and status were mapped to controlled vocabularies, H3 cells and cost bands were derived, and the fields listed under Known Redactions / Exclusions were omitted.

Registry source	In this sample	Registry terms	Source	Attribution condition
chicago-permits-socrata	yes	LicenseRef-Chicago-Open-Data	https://data.cityofchicago.org	Source: City of Chicago open data portal. Retain source_uri per record.
dallas-permits-socrata	no rows	LicenseRef-Dallas-Open-Data	https://www.dallasopendata.com	Source: City of Dallas open data portal. Retain source_uri per record.
houston-permits-adc	no rows	LicenseRef-Houston-Open-Data	https://data.houstontx.gov	Source: City of Houston open data portal. Retain source_uri per record.
los-angeles-permits-socrata	yes	CC-BY-4.0	https://data.lacity.org	LADBS. Los Angeles building permits dataset licensed CC-BY-4.0; retain source_uri per record and identify modifications.
nyc-dob-now	yes	LicenseRef-NYC-Open-Data	https://a810-bisweb.nyc.gov	Source: City of New York — Department of Buildings open data portal. Retain source_uri per record.
philadelphia-permits-ais	no rows	LicenseRef-Phila-Open-Data	https://ais.phila.gov	Source: City of Philadelphia open data portal. Retain source_uri per record.
phoenix-permits-socrata	no rows	LicenseRef-Phoenix-Open-Data	https://www.phoenix.gov/opendata	Source: City of Phoenix open data portal. Retain source_uri per record.
seattle-permits-socrata	yes	LicenseRef-Seattle-Open-Data	https://data.seattle.gov	Source: City of Seattle open data portal. Dataset Public Domain; Seattle Open Data Policy permits commercial use. Retain source_uri and version and identify modifications.

Known Redactions / Exclusions

The public sample omits the person-bearing applicant, owner, and contractor structs because they can contain names and contact details. The description_raw free-text field is excluded because it may contain personal details that cannot be reliably separated from permit text. The address struct and raw lat/lng coordinates are excluded because they can identify a private residence. The retained h3_index is derived at resolution 8 from source coordinates for coarse spatial joins. The derived description_normalized field is not emitted by these collectors. Legacy table fields not emitted by the upstream collector (completion_date, approved_at, expired_at, inspected_at, permit_valuation_tier, permit_volatility_12mo, declared_valuation, and ftm) are not part of this release projection. The supplied extracts carry parcel_id and naics_code only as null-only schema columns, so neither key is included in the candidate. Every row carries a per-record DLP receipt in the staged permit release audit.

ML Split

Not an ML train/test split: This dataset is provided as a single train split purely for HuggingFace dataset tooling compatibility.

Citation

Suggested source attribution and license notices for this research sample:

Hugging Face license metadata: other.

• chicago-permits-socrata — LicenseRef-Chicago-Open-Data; https://data.cityofchicago.org. Source: City of Chicago open data portal. Retain source_uri per record.

• dallas-permits-socrata — LicenseRef-Dallas-Open-Data; https://www.dallasopendata.com. Source: City of Dallas open data portal. Retain source_uri per record.

• houston-permits-adc — LicenseRef-Houston-Open-Data; https://data.houstontx.gov. Source: City of Houston open data portal. Retain source_uri per record.

• los-angeles-permits-socrata — CC-BY-4.0; https://data.lacity.org. LADBS. Los Angeles building permits dataset licensed CC-BY-4.0; retain source_uri per record and identify modifications.

• nyc-dob-now — LicenseRef-NYC-Open-Data; https://a810-bisweb.nyc.gov. Source: City of New York — Department of Buildings open data portal. Retain source_uri per record.

• philadelphia-permits-ais — LicenseRef-Phila-Open-Data; https://ais.phila.gov. Source: City of Philadelphia open data portal. Retain source_uri per record.

• phoenix-permits-socrata — LicenseRef-Phoenix-Open-Data; https://www.phoenix.gov/opendata. Source: City of Phoenix open data portal. Retain source_uri per record.

• seattle-permits-socrata — LicenseRef-Seattle-Open-Data; https://data.seattle.gov. Source: City of Seattle open data portal. Dataset Public Domain; Seattle Open Data Policy permits commercial use. Retain source_uri and version and identify modifications.

No commercial or training authorization is implied.

Contact

Maintained by Axiomancer Labs as part of Axiom Codex. For questions, corrections, or removal requests, open a discussion in this repository's Community tab or reach us through www.axiomcodex.io.

Governed Dataset Card Metadata

• Source / license posture: Research-tier rows from four registered municipal open-data feeds (Chicago, Los Angeles, NYC DOB, Seattle), each under its own city open-data terms with attribution and a per-record source_uri.
• Intended use: Research, exploration, and evaluation of municipal permit activity, permit-type and status mix, and H3-level construction signals.
• Forbidden use: No identification of property owners, applicants or contractors, no decisions about individual properties or people, and no treatment of status snapshots as verified permit outcomes.
• Known missingness / bias: Cities cover different periods (Los Angeles ends May 2023); status is a collection-time snapshot; building_type, unit_count and valuation_usd exist only for Los Angeles; parcel_id and naics_code are excluded as null-only.
• Update cadence: Static sample of one collection per feed; later municipal updates are not reflected.
• Geographic / temporal coverage: Four U.S. cities identified by jurisdiction_slug, with h3_index at H3 resolution 8 on every row (no addresses or raw coordinates), plus source permit dates applied_at, issued_at, closed_at, and modified_at; occurred_at is the permit milestone: issue date, else application date, else the collection timestamp when the source reports neither.
• Label semantics: No outcome or task labels are projected.
• Confidence / provenance method: source_feed/source_uri/provenance, content-addressed collector receipts, and a sealed per-record DLP receipt in the permit release audit; person-bearing party structs are omitted.
• Training policy: Not authorized for training under the Axiom Codex governed training-use policy; the single train split is a Hub convention, not an ML split.

Release Scope & Quality Evidence

• Scope: source_records; 25000 staged rows from upstream collector Parquet inputs.
• Observed time window: 2000-01-11T00:00:00Z to 2026-10-07T04:04:20Z.
• Machine-readable evidence: manifest.json, quality.json, and baseline.json.
• Packaging validation command: python3 tools/hf_packaging/validate.py --staged <staged-dir> --dataset permit-signals --release-candidate --tier research. This records the build's validation entry point and requires the Axiom Codex packaging environment, which is not included in this Hugging Face data repository. Use the quickstart above to load the data. This card does not claim a model baseline or an ML split unless a future task/evaluation configuration supplies one.
• Promotion: a release candidate requires the snapshot-bound quality report, baseline, current source-policy replay, and all applicable thresholds; blocked or unknown gates remain blocked.