Called the page summary endpoint while choosing the approach and again from a script after the lookup was implemented. A single unauthenticated GET, with the article title encoded in the path, returned the extract, the desktop page link, and the revision timestamp needed for the stored snapshot. The live check produced a passage, link, and revision date for a matching article, and the parser accepted that shape.
- What worked
- No key or account was required. One summary response carried the passage, canonical link, and revision time, so the snapshot did not need a second content call.