Turning scattered procurement pages into reviewable bid intelligence

Public opportunities are spread across inconsistent agency pages and procurement platforms. Titles, deadlines, documents, locations, and access rules vary, making it difficult for estimators to assemble a reliable view without losing the official source or overstating weak evidence.

TradeOps converts source records into a canonical bid model, classifies relevant trades, collapses near-duplicates, tracks previously seen records, applies source-rights decisions, and produces estimator-facing HTML, JSON, and CSV artifacts.

The product boundary is deliberate: it supports research and review but does not decide whether a contractor is qualified or should bid.

Source pages and adapters → Pydantic normalization → trade classification → duplicate handling and SQLite state → rights-aware export gate → local reports

The operational system supports asynchronous collection, static extraction, platform-aware adapters, and Playwright fallback. The public-safe demo uses only local, project-owned HTML fixtures and requires no network connection.

• Missing or unreviewed source rights default to quarantine.
• Provenance and reason codes remain visible.
• The demo has no network, messaging, scheduler, CRM, feed, or deployment path.
• Bid/no-bid, rights, recipient, and publication decisions remain human decisions.

The release produced a deterministic, reviewable research path and replaced a broad contractor-matching theory with a more defensible category-and-location intelligence workflow. The evidence describes the synthetic demo and engineering verification—not customer results or market-wide accuracy.