Run the batch. Keep the provenance.

The problem

Repeated searches can lose progress when a process stops. Flattening results to one domain list also loses the query and location that explain each match.

What I did

I defined the batch, recovery and export requirements, directed the AI-assisted implementation, and reviewed synthetic fixtures and exact release evidence. My practical owner walkthrough is deferred.

How it works

• Edit location and search-template CSVs, then inspect the expanded plan. Configure query, page and request bounds before running; retries and resumed attempts count toward the request budget.
• Commit page results and progress together in SQLite. Resume with the same input fingerprint, preserve completed work, and report partial coverage when limits or failures prevent completion.
• Export one consistent snapshot as domains.csv for distinct full hosts and matches.csv for query, location, source URL and rank provenance. Keep www hosts and subdomains distinct and neutralize spreadsheet formulas in CSV cells.

What you can verify

• The bundled offline demo makes four queries and eight successful page attempts, producing three reserved .test hosts and twelve match rows. These are fictional results, not leads.
• 43 maintained tests passed at the public release commit on Windows and Ubuntu CI, covering process exits, retries, quota/depth limits, safe CSV exports and competing writers.
• Requires Node.js 24.15.0 or a later Node 24 release. The finite public source has 33 files, no third-party packages and no install step; Windows launchers and operator instructions are included.