| The query_documents tool only matches the title, doccontent and doccontentraw Solr fields. The standard search UI (Main.SolrSearchConfig) also matches name, filename, attcontent, objcontent, author_display, creator_display, comment, attauthor_display and spaces, with the same title and content weights. As a result an agent cannot find a page by the text of its PDF or Office attachments, by its object properties or by a file name unless it hand-writes an attcontent: or filename: clause, which models do inconsistently. Align the tool's query fields and phrase fields with the platform's DOCUMENT configuration so that an agent searching with plain keywords gets the same hits a user gets in the search box. Keep type:DOCUMENT as the result unit, so an attachment match still returns the page holding the file. Acceptance:
- A keyword that only appears in an attached PDF returns the holding page with no field prefix.
- A keyword that only appears in an object property or a file name returns the page.
- Ranking for title and body matches is unchanged.
- The man query_documents page drops the need for attcontent: and documents the field prefixes as an optional narrowing.
|
