Forty vendor portals, no APIs, one warehouse

Forty suppliers, forty login portals, zero APIs. Someone in ops spends Monday mornings copying numbers from each into a spreadsheet.

Each vendor gets a container. Inside, a hoody-browser instance holds that vendor's logged-in session. A nightly cron GETs /api/v1/browser/eval with a small script that reads the stock and lead-time cells off the page, then POSTs the result as a row into the shared warehouse at /hoody/databases/supply.db. Forty containers write to one file; the host serializes them.

By 7am the supply dashboard has every vendor's numbers, timestamped. When a portal redesigns its table, one eval script changes.

The forty browsers sit idle 23 hours a day on one flat-rate server, cloned from one base, sharing disk blocks. On a per-function cloud that's forty warm Chromium functions and a bill that tracks the vendor count.

The vendor never shipped an API. You shipped it for them.