​
Search Bookmarks

Perform a hybrid search across the user’s bookmarks. Combines three strategies for maximum recall:
  1. Full-text search (PostgreSQL FTS) — exact word matching with stemming
  2. Fuzzy matching (trigram similarity via pg_trgm) — tolerates typos
  3. Semantic search (pgvector embeddings) — understands meaning
curl -H "Authorization: Bearer $ACCESS_TOKEN" \
  "https://api.unisave.io/v1/search/bookmarks?q=react+hooks+tutorial&limit=20"
Query Parameters
ParamTypeRequiredDescription
qstringYesSearch query
limitintegerNoMax results. Default: 20.
Response 200 
{
  "results": [
    {
      "clientBookmarkId": "abc-123",
      "url": "https://react.dev/reference/react/hooks",
      "title": "React Hooks Reference",
      "summary": "Complete guide to built-in React hooks...",
      "tags": ["react", "hooks", "frontend"],
      "thumbnailUrl": "https://react.dev/og.png",
      "score": 0.92
    }
  ],
  "total": 1
}
Results are ranked by relevance score (higher = better match). The score blends full-text rank, trigram similarity, and cosine similarity from embeddings.

​
Search Fields

The query matches against these bookmark fields:
FieldFTSFuzzySemantic
TitleYesYesYes
URL / domainYesYesNo
SummaryYesYesYes
TagsYesNoYes
DescriptionYesYesYes
Semantic search uses AI-generated vector embeddings (created during enrichment). This enables meaning-based search:
  • Query “cooking tips” finds bookmarks about recipes
  • Query “frontend framework” finds React, Vue, and Svelte bookmarks
  • Query “investment advice” finds finance articles
Bookmarks that haven’t completed enrichment don’t have embeddings yet. They’re still findable via full-text and fuzzy search, but won’t appear in semantic results.

​
Pro Gating

The saveWhy field in search results follows the same Pro gating as the enrichment status endpoint — free users receive null.