How the system works

Source, evidence, and alert methodology

The deterministic and evidence-constrained workflow behind every record and alert.

1. Ingest and preserve

Official identifiers, metadata, URLs, raw snapshots, normalized snapshots, checksums, and first and last seen timestamps are stored. Connectors use cursors, retries, schema validation, and idempotent upserts.

2. Resolve status and dates

Document lifecycle stages are classified without collapsing proposals, final rules, guidance, notices, and corrections. Dates are typed so publication, comment, effective, compliance, enforcement-discretion, and expiration dates are not conflated.

3. Draft against evidence

Candidate obligations, parties, dates, thresholds, exceptions, and definitions retain source locators. Draft claims can publish only when every citation validates against the current source version.

4. Match and deliver once

Watchlist matches record transparent reasons. Immutable match events and idempotency keys prevent duplicate alerts; material source changes trigger a new, traceable event.

5. Fail visibly

Missing evidence holds a summary. Source drift, date anomalies, job failures, and delivery errors surface in operational health.