@crowi/plugin-search-mongo

The default, infra-free search driver for Crowi 2.0. It searches live data — the Page collection and each page's current Revision — with a case-insensitive MongoDB $regex over path / title / body. No external service, no separate search index: MongoDB alone powers search.

This is what makes the slim deployment (local file storage + mongo search) searchable out of the box, with MongoDB as the only required infrastructure.

When to use

• ✅ Small / mid-size wikis that don't want to run Elasticsearch / OpenSearch.
• ✅ Japanese (and other CJK) content: a $regex substring match is more practical than MongoDB $text, whose CJK tokenisation is weak.
• ❌ Large installs. A non-anchored $regex cannot use an index, so search scans the collection. Run @crowi/plugin-search-elasticsearch instead at scale.

How it works

• No index maintenance. index() and remove() (called from the page-saved hook) are no-ops; there is nothing to keep in sync because the page body already lives in Page / Revision. rebuild() is omitted.
• Two-pass query. Path/title hits (the page title is its path) are found first, then body hits are resolved via a single bulk Revision.find({ revision: { $in }, body: regex }) over the viewer-visible candidate pages (capped at CANDIDATE_CAP = 5000 to bound the scan). Path hits rank ahead of body-only hits.
• Grant-aware. Results respect page visibility: anonymous viewers see public pages only, admins see everything, and other viewers see public pages plus pages they created or that are shared with them. Drafts, deleted pages and redirects are always excluded.
• Snippets. Best-effort substring around the first match, with the match wrapped in <mark>. Not HTML-escaped — the web client sanitises before render.

Configuration

None. The driver uses the same MongoDB connection as the rest of the app, so its configSchema is empty. Select it with search.driver: 'mongo' in the runner's crowi.config.json (this is the schema default).