Search

SwarmVault builds a SQLite FTS (Full-Text Search) index over all wiki pages during compilation.

How It Works

  1. During compile, all wiki pages are indexed into the local SQLite shard under state/retrieval/
  2. When you run query, the search index finds relevant pages
  3. Matching pages are passed as context to the LLM provider
  4. The provider generates an answer grounded in your wiki content

Retrieval Index

The index is rebuilt on every compilation, and saved output pages are also indexed immediately when you use query or explore. Use swarmvault retrieval status|rebuild|doctor to inspect or repair it. It supports:

  • Full-text search across all page content
  • Ranking by relevance
  • Fast lookups for the query pipeline
  • Stored kind, status, and project_ids metadata for the local graph workspace filters

Query Persistence

By default, query answers are persisted as wiki output pages in wiki/outputs/, indexed immediately, and linked back into related source, concept, and entity pages through the lightweight artifact sync path. Use --no-save to skip persistence.

swarmvault graph serve uses the same search index through /api/search.

The local workspace can filter results by:

  • page kind
  • lifecycle status
  • project scope