The shipper portal pull.

The problem

Two pages of the platform ran on a spreadsheet export from the shipper's portal, and a teammate produced that export by hand every 15 minutes. The shipper's API team confirmed there would be no data access.

How it works

• Drives the portal in a real browser: set the pickup window, run the slow search, wait on the real row counter, fetch the export in-page. Legacy spreadsheet in, typed rows out.
• The portal silently caps every export at 500 rows and keeps the oldest. The bot splits the date window and unions the pieces by load number instead of ever shrinking the search.
• Every live run publishes a status row, and both dashboards show it: green when fresh, amber with the reason when the bot is failing or logged out, parked outside hours. The rules are pure functions with 26 checks.
• The bot clicks exactly two things: Search and Export. The portal's approval controls are never touched.

// the bot can only search and export; nothing on the portal is ever approved or changed by it

The result

The every-15-minutes manual export is retired. The first live write carried 529 rows, past the cap the split was built for, and the dashboards now say exactly how fresh their data is.