shelfmark-find-context

Find Shelfmark Context

Find the smallest useful set of documents while respecting catalogue freshness and sharing boundaries.

Search Workflow

1. Call corpus_stats() first. Note freshness, available roots, document types, context types, and the rights split before forming a query.
2. If the user names a folder or only describes a broad area, call browse_folder() to learn the catalogue's actual structure and labels.
3. Call search_docs() with filename-style or title-style terms. Shelfmark indexes paths, filenames, authors, company metadata, OOXML titles, and slide titles, not document body text.
4. Add only filters supported by the corpus summary or prior results. Use root, client, doc_type, context_type, authored-year bounds, or path_contains to narrow a large result set.
5. Call get_file() on the strongest candidates to confirm rights, confidentiality, authorship, dates, slide titles, duplicate copies, and on-disk status.
6. Return a short ranked list with exact catalogue paths, why each result fits, and any governance or freshness caveat.

Apply Governance

• Set shareable_only=true whenever paths or derived material may leave the user's local environment. This excludes confidential and unreviewed UNKNOWN material.
• Set own_only=true only when the user specifically needs material classified as authored by them.
• Never infer or probe RESTRICTED material. Shelfmark deliberately makes restricted paths indistinguishable from absent paths.
• Do not describe metadata matches as proof of document-body content. Ask to open a selected file with a separately authorized file-reading tool when its contents are needed.

Recover From Weak Results

• Remove unsupported filters before broadening the query.
• Try filename, project, organization, author, or slide-title vocabulary rather than sentence fragments from a document.
• Use browse_folder() when terminology is uncertain.
• State that a true no-result search is recorded locally by Shelfmark as product feedback; it is not sent anywhere.