Major Industrial Preconstruction Permit Intelligence

Worker #3 detects major industrial projects entering environmental preconstruction permitting in Louisiana, West Virginia, and Texas. It uses only official state sources and retains project lifecycle changes rather than treating each run as a disposable scrape.

Version 0.9.3 is the validated production candidate. It does not modify or depend on Worker #1 or Worker #2.

The public Actor is available at https://apify.com/beaming_tellurium/major-industrial-preconstruction-permit-intelligence. Its production schedule runs daily at 06:30 America/Los_Angeles against the validated v2 observation, delivery, and health stores.

Public example tasks:

• Track Texas Data Center Air Permit Projects
• Monitor Advanced Manufacturing Permit Projects
• Find Major Industrial Permits in LA, WV and TX

What it includes

• PSD and GHG PSD projects
• new major stationary sources and major modifications
• explicit industrial construction, capacity, equipment, or manufacturing expansions
• data centers, power generation, LNG, chemical, steel/metals, and advanced manufacturing projects
• application, completeness, revision, draft/preliminary decision, public notice, final issuance, and withdrawal stages when the official source exposes them

What it rejects

• routine permit renewals or no-change Title V/Part 70 actions
• administrative changes, minor actions, and general permits
• temporary or small concrete batch plants
• portable crushers, well pads, and ordinary compressor stations
• Texas rows without usable official English evidence
• Texas name-only matches; project facts must come from the official summary PDF

Official sources

• Louisiana DEQ Public Notices: https://www.deq.louisiana.gov/public-notices
• West Virginia DAQ NSR Permit Applications: https://dep.wv.gov/daq/permitting/Pages/NSR-Permit-Applications.aspx
• Texas TCEQ Pending Air Permit Applications: https://www.tceq.texas.gov/permitting/air/newsourcereview/airpermits-pendingpermit-apps

The Texas concrete-table section is intentionally excluded. The Actor uses the official English Plain Language Summary first and falls back to official English NORI/NAPD evidence only when the PLS is unavailable, unreadable, or mislabeled. The Actor fails instead of silently truncating when a configured document ceiling is too low.

Data and history model

Each delivered event includes stable project, company, facility, and source-event IDs; permit and lifecycle fields; project facts; PSD flags; qualification reasons; official provenance; and observation history:

• event_date, observed_at, first_seen_at, last_seen_at
• change_class, changed_fields
• previous_values, current_values
• source_url, source_hash, supporting_documents_hash

Observation history and customer delivery state use separate named stores. All qualified source observations are committed append-only after source-health gates pass. maxResults and spending limits commit only the records actually delivered to the delivery store, so deferred results remain deliverable later.

Texas emits one current lifecycle observation per permit. When NORI/NAPD links appear or change, the same stable source event becomes CHANGED, while the earlier observation remains append-only. West Virginia retains distinct official lifecycle documents, including revisions and withdrawals.

Main inputs

• states: any of LA, WV, TX; the public prefilled default is Louisiana only so marketplace QA and first-time evaluation finish comfortably within five minutes
• onlyNew: deliver only records not previously delivered
• includeChanged: include changed records with onlyNew
• maxResults: output limit independent of observation capture; the public prefilled default is 25
• company, county, projectType, minimumStage: customer filters
• permitNumbers: focused validation/monitoring; requires exactly one selected state
• maximumTexasDocuments, maximumWestVirginiaDocuments: fail-closed download ceilings
• productionMode: enforce source-health gates

Advanced inputs expose separate observation, delivery, and health store names plus bounded document/storage concurrency.

Pricing

The Actor uses transparent pay-per-event pricing with platform usage included:

• Validated official-source run: $0.15 once, after all selected source-health and evidence gates pass.
• Intelligence record: $0.003 for each record written to the default dataset.
• Actor start: Apify's standard $0.00005 synthetic event.

A current three-state bootstrap producing 61 records costs approximately $0.33. A stable onlyNew: true monitoring run with no new or changed records costs approximately $0.15. The minimum maximum-charge setting is $0.16; spending limits defer unaffordable records without incorrectly marking them delivered.

Run locally

Requires Node.js 22 or newer.

npm ci
npm test
npm start

For a full current-source verification:

node scripts/validate-full-network.mjs

The validation script performs hundreds of official Texas PDF requests and should be used deliberately. See VALIDATION.md for the validated results and CLOUD_LAUNCH.md for the public-release handoff.

Important limitation

This product surfaces project signals from public permitting records. It is not legal advice, a compliance determination, or proof that construction or financing will occur. event_date_inferred: true identifies records where the source exposes no reliable document date and the observation date is used.