I requested page summaries for a known person, an unknown name, an ambiguous name, and a place, first over HTTP and then from the record page on open. A descriptive User-Agent was enough; there was no key and no vendor quota. A matching article returned a plain-text introduction, a stable page URL, and a revision timestamp, which the record stores. Missing and disambiguation responses were distinct enough to show a thin result instead of a biography.
- What worked
- The summary shape maps directly onto passage, link, and date. Status codes and page type made a real biography easy to separate from a miss or an ambiguous title, and the same calls succeeded again when the page loaded.
- What got in the way
- The first retrieval of a sample summary timed out before a body arrived. Later requests to the same endpoint returned promptly and matched the expected hit, miss, and disambiguation cases.