Search

DocPlatform includes an embedded full-text search engine (Bleve) that indexes all content automatically. No external service to configure — search works out of the box.

Using search

Cmd+K dialog

Press Cmd+K (macOS) or Ctrl+K (Windows/Linux) anywhere in the web editor to open the search dialog.

┌──────────────────────────────────────────┐
│  🔍  Search documentation...             │
├──────────────────────────────────────────┤
│                                          │
│  📄 Getting Started                      │
│     Install and configure DocPlatform... │
│                                          │
│  📄 API Authentication                   │
│     JWT tokens, OAuth2, and session...   │
│                                          │
│  📄 Docker Deployment                    │
│     Run DocPlatform as a container...    │
│                                          │
│  ↑↓ Navigate   ↵ Open   Esc Close       │
└──────────────────────────────────────────┘

What’s indexed

The search engine indexes:

• Page title (boosted weight for ranking)
• Page description (boosted weight)
• Full page content (body text, code blocks, lists, etc.)
• Tags (keyword field, boosted)
• Page path

Matching behavior

Queries are plain keywords — there is no special operator syntax:

Permission filtering

Search is workspace-scoped: every query is constrained to the workspace it’s issued against at the search-engine level, and the search endpoint requires read access to that workspace. There is no cross-workspace search.

Indexing

Automatic indexing

Content is indexed incrementally via an async job queue:

1. A page is created or updated (via editor or git sync)
2. Content Ledger emits an event
3. A search indexing job is queued
4. The Bleve indexer processes the job and updates the index

There’s a brief delay (typically under 1 second) between saving a page and the updated content appearing in search results.

Rebuild the search index

If the search index gets out of sync (rare — usually after a crash or manual file manipulation), rebuild it:

docplatform rebuild --workspace {workspace-id} --search

This re-indexes the workspace’s pages from the filesystem (the --search flag includes the search index in the rebuild).

Index health

Check overall instance health (FS/DB consistency, broken wikilinks, backups, and more) with the doctor command:

docplatform doctor

If search results look stale or incomplete, rebuild the index with docplatform rebuild --workspace {id} --search.

Search in published docs

Published documentation sites do not currently include a visitor-facing search interface — search is available to authenticated workspace members in the editor. For AI agents and crawlers, published sites expose llms.txt / llms-full.txt and sitemap.xml for content discovery instead.

Search engine internals

For users who want to understand how search works under the hood:

Analyzer

Bleve uses the English analyzer by default, which includes:

• Tokenization — splits text on whitespace and punctuation
• Lowercasing — case-insensitive matching
• Stop word removal — filters common words (the, is, at, etc.)
• Stemming — matches word variants (running → run, documented → document)

Field boosting

Not all fields are weighted equally in relevance scoring:

This means a query matching a page’s title ranks higher than the same query matching deep in the body text.

Storage

The Bleve index is stored at {DATA_DIR}/search-index/ using the Bleve scorch storage engine (backed by bbolt). The index is separate from the SQLite database and can be safely deleted and rebuilt with docplatform rebuild.

Performance

Search performance scales linearly with content volume. For workspaces exceeding 10,000 pages, a future release will offer optional Meilisearch integration.