enquire-mcp API reference - v3.9.0-rc.4
    Preparing search index...

    Function findSimilar

    • Lexical-hybrid similarity over vault-native signals — finds notes related to the target without any embeddings.

      Combines four structural signals: tag Jaccard (×3.0), title character 3-gram Jaccard (×1.5), shared-outbound link overlap (×2.0), and co-backlink Jaccard (×2.0). Tag overlap dominates by design — that's the strongest "this is the same topic" signal a human would use. Skips body cosine on purpose: structural signals converge fast at vault scale (5k × 5KB) without a full TF-IDF pass per call.

      Use this when the agent has one specific note and wants neighbors. For "find notes about " use searchHybrid or semanticSearch.

      Parameters

      • vault: Vault

        The vault to search.

      • args: { limit?: number; min_score?: number; path?: string; title?: string }

        One of path or title is required to identify the target. limit defaults to 10. min_score (default 0.05) prunes weak matches.

      Returns Promise<SimilarNote[]>

      Sorted SimilarNote[] (desc by score), capped at limit. Empty array if the target was excluded by --exclude-glob.

      If neither path nor title is provided, or the target cannot be resolved.

      const related = await findSimilar(vault, {
      path: "Reference/Hybrid Retrieval.md",
      limit: 5
      });
      for (const n of related) {
      console.log(n.path, n.score, n.signals);
      }