Answers for the compliance officer in the room

A closed taxonomy of regulatory tags is enforced inside the dataset — free-form strings are never accepted. The six seed jurisdictions are seeded in src/lib/business/compliance/taxonomy.ts: EU Green Deal, US EPA Clean Air Act, US EPA Clean Water Act, California Air Resources Board, UK Environment Agency, and Australia NGER.

Live alerts drawn from data/regulatory-feed.json extend that surface as rules land — EPA NESHAP refinery fenceline monitoring, EPA CWA steam-electric ELG updates, CARB Advanced Clean Fleets phase-in, NYSDEC Part 212 NOx RACT updates, NYC Local Law 97 penalty schedule, EU CBAM Phase 2 embedded emissions, UK EPR public register changes, and Australian Safeguard Mechanism baseline tightening are all shipped live. Each entry already carries its regulatoryTags + effectiveDate so the platform cross-references new rules against the closed taxonomy as soon as the feed is updated.

Reports are rendered from the HTML template at src/lib/business/compliance/template.ts and turned into a regulator-ready PDF by the installed pdf module. Every report ships with the same structure: a site / reporting-period meta block, regulatory-tag chips from the closed taxonomy, totals (readings counted, sensors reporting, metrics tracked), and the readings table itself with each reading's measured value, threshold, and pass/warn / fail status.

Want to see the shape before signing in? Each tier card on /pricing#tiers prints a live sampleOutput line taken straight from the production data model, and you can generate a full sample PDF in under a minute from the same page.

Three tiers, no add-on fees. The exact deliverables per tier are defined on src/app/(custom)/pricing/page.tsx and rendered live on /pricing:

• Pilot — up to 3 compliance reports per month, the regulatory alerts feed (read-only), a basic risk-flag summary per report, and regulator-ready PDF export.
• Professional — unlimited reports, the full alerts feed with site-level risk scoring, the report history + audit trail, and email notifications on threshold breaches.
• Operations — everything in Professional, plus field crew dispatch and scheduling, the multi-site monitoring dashboard, and cron-automated report generation.

Every per-tenant resource is exposed through an authenticated REST surface — you read and mutate app data exclusively via @/lib/api-client (apiFetch). The shipped endpoints today:

• /api/reports and /api/reports/[id] — report CRUD, scoped by where: { userId: user.id }; see src/app/api/reports/route.ts and src/app/api/reports/[id]/route.ts.
• /api/regulatory/alerts — the live regulatory feed.
• /api/regulatory-tags — the closed taxonomy as JSON.
• /api/crews/schedule — field crew dispatch.
• /api/billing/summary — billing state for the authenticated tenant.
• /api/pdf/document — programmatic PDF render backed by the installed pdf module (renderDocumentPdf).

Webhook honesty: outbound webhooks for report-published and threshold-breach events are on the Operations tier roadmap — they don't ship as a generic /api/webhooks/ surface yet. In the meantime, authenticated clients can poll /api/reports for new completions. Tell us your consumer format and we'll flag the priority of the rollout.

Per-tenant isolation lives in the route handlers, not the database: requireAuth() resolves the session, and every read and write is scoped by where: { userId: user.id } (or the matching aggregated group key for shared workspaces). There is no implicit global query path — if a handler forgets the scope, the data simply isn't returned.

Shared workspaces: every report carries optional orgId and orgLabel fields, so a compliance group can attach readings and reports to an organisation namespace without anyone outside that namespace seeing them.

Retention policy is explicit, not automated — there is no cron-driven purge job silently dropping rows. Reports persist until the account owner deletes them; deletion is irreversible and audited. If your procurement requires a fixed retention window (e.g. 7-year archive with tamper-evident hashing), say so during kickoff and we'll scope the Operations-tier tenant configuration to match.

Model training: your customer report data is never used to train shared or foundation models. The only place it lands is your tenant's own query results and exported PDFs.