Learn the product from its public site and docs (answers at once; the inventory follows).

Sleeper reads the product's public site and docs (its pages, llms.txt, sitemap and OpenAPI description, up to 24 pages, obeying robots.txt) and lists its features, who it is for, what it claims and what it costs, each with the page it came from. When the product's app is connected to Sleeper — a sign-in Sleeper has tested — Sleeper also signs in with that connection and walks the app's own screens, honouring any areas you have blacked out. With no connection, nothing but public pages is read: no login, cookie or key is sent for them, and private or reserved addresses are refused. It runs in the background: the answer is the run (learning), and the inventory reads ready in a minute or two. For a product project only. Free, up to 10 runs a day. siteUrl is the product's https address; leave it out to read the last run's site again. connectionId names the project's own app connection to walk; leave it out to use its verified one (public pages only when it has none). 202 with the new run; 200 with alreadyRunning: true and the run already learning (no second run starts, so no Idempotency-Key). Poll GET until the run is ready or failed. 400 validation_failed for an address that is not https or not public, or for a connectionId that is not this project's own connection (field connectionId); 409 project_not_product for a story project; 403 usage_limit_reached past the daily limit.