article
Work wiki's API can't search page text, so the server reads every page
The task tool's documents API can't search full page content, so my server pulls one document, matches the phrase itself, and answers with excerpts and honest counts instead of page bodies.
Read more 4m read