Access — inside the guide
The Supabase schema.
Five tables, free tier, well under the limit at nine-site scale. This is the shared memory that lets every other workflow know what's already happened without you tracking it by hand.
Five tables
01
site_config
One row per site: repo name, domain, Vercel project ID, deploy hook URL. Everything the other tables need to know which site they're talking about.
02
drafts
Every generated draft, its status (pending, approved, rejected, published), which site it belongs to, and a timestamp. This is the queue.
03
pages
The live catalog — one row per published page, updated by the Inventory Agent. What's actually out there right now, not what you think is out there.
04
runs
A log of every workflow execution — which one ran, when, what it touched, whether it succeeded. This is what makes debugging possible at nine-site scale.
05
escalations
Anything the Review pipeline flags as a problem lands here, waiting on a human decision instead of getting silently fixed or silently ignored.
Buyers get the ready-to-run SQL for all five tables, or this written walkthrough to build the schema by hand. Next: what actually writes to the drafts table. The Drafter workflow.