Reading this site by machine

Files

• /downloads/ledger-latest.json · .csv — the curated air ledger: one row per air permit, with generator counts, MW, NOx, water and an evidence label per figure. Schema v2, 35 columns.
• /downloads/entities-latest.json · .csv — companies, project LLCs, consultants, contractors, law firms, brokers, utilities, plus public bodies, agencies, dockets and landowner groups. Columns: entity_id, name, kind, roles, aliases, parent_id, state, municipality, website, source_url, last_verified, notes.
• /downloads/relationships-latest.json · .csv — one row per cited edge. Columns: rel_id, from_id, from_kind, to_id, to_kind, relationship, date, status, source_url, source_kind, quote, state, municipality, last_verified.
• /downloads/people-latest.json — officials and officers in a cited public role. Columns: person_id, display_name, public_role, public_role_kind, org_id, role_title, role_source_url, role_start, role_end, last_verified, notes. There are no contact fields. See the naming policy before using this file.
• /graph/<st>/<municipality-slug>/graph.json — {municipality, state, generated, nodes[], edges[]}. Node ids are <kind>:<id>; edges carry id, source, target, relationship, date, status, source_url, source_kind, quote.
• /pulse/<YYYY-MM-DD>/pulse.json — one Data Center Pulse issue: {date, url, source, threads[]}, each thread {title, subreddit, state, url, published, score, matched_site_id, matched_terms, doc_links, excerpt}. No usernames. Issues are listed in /pulse/feed.xml (RSS) and /sitemap-pulse.xml.
• Live sheet, any tab as CSV: https://docs.google.com/spreadsheets/d/1Xnc8Vp19R0Qp5PZQa8qBdXe_uiDY9bLOzQ3dqQ8ZUcg/gviz/tq?tqx=out:csv&sheet=<tab> with tab ∈ ledger, changes, sources, states, sites, permits, figures, entities, people, relationships.

ID namespaces

• site_id — <st>-<facility-slug>, e.g. pa-qts-salem-qts-data-centers. Never renamed.
• permit_id — <st>-<permit-number> when the agency issued one (pa-40-00158c), else <st>-<agency>-<agency-id>.
• entity_id — prefix says what it is: co- company, body- public body, agency-, docket- docket or order, agg- group such as “96 landowners”.
• person_id — p-<name>-<org>, scoped to one role at one organization; the same person on two bodies is two rows.
• rel_id — first 10 hex of sha1(from|to|relationship|source_url); stable across rebuilds.

Vocabularies

• Permit status: anticipated (a tracker placeholder — nothing filed) · applied · under-review · public-notice · issued · denied · withdrawn · expired · unknown.
• Permit verification: tracker < document < figures — how far a human has read behind the tracker row.
• Relationship status: recorded (a document says it happened) · pending (a vote, application or transfer not yet decided — never a prediction).
• Relationship source_kind: minutes, roster, permit, notice, tracker, docket, gis, sec, press, developer, official-site, other. A developer claim always carries a verbatim quote.
• Entity kind: operator, developer, llc-vehicle, consultant, engineer, contractor, equipment, law-firm, broker, utility, lender, investor, body, agency, docket, aggregate, other. roles is a |-joined subset — a firm can be applicant and contractor.
• Figure evidence (ledger): Permit figure, Official tracker, Developer statement, Press report, Calculated, Not published, Not yet reviewed. “Not published” is a fact about the record, not a missing value.

Rules the data follows

1. Every row cites a URL. A relationship without source_url cannot be written.
2. Company names resolve through an exact alias list; nothing is merged by similarity. “Amazon Web Services” and “Amazon Data Services, Inc.” are one entity because a human said so.
3. People rows exist only with public_role = TRUE and a document naming the role; the validator rejects any cell shaped like an email or phone number. Residents and individual land sellers are never rows.
4. Announced capacity is never mixed with permitted capacity; figures come only from documents, with an evidence label.
5. Absence means we have not read that record yet, not that nothing exists.

Page patterns

• /companies/<entity_id>/ — sites, permits, relationships, aliases; schema.org/Organization in JSON-LD. A company with a single record carries noindex until a second appears.
• /graph/<st>/<municipality-slug>/ — the relationship table is the page; the drawn graph is a view of the same rows. schema.org/Dataset with spatialCoverage; no Person markup anywhere on the site.
• /near/<st>/<place>/ — sites within 25 miles of a place; Dataset with coordinates.
• /<st>/#p=<site_id> — deep link to a site on the map.

Corrections

corrections@datacenterpermits.org, or the forms on the about and naming policy pages. Changes are logged with date and reason.