Neo4j Vector Search Skill

Status: Draft / WIP — Content is a placeholder. Reference files to be added.

When to Use

• Creating a vector index on a node property (embeddings)
• Running vector similarity search (semantic/nearest-neighbor lookup)
• Storing embeddings on graph nodes as part of an ingestion pipeline
• Using the new SEARCH clause (Neo4j 2026.02.1+) or the legacy db.index.vector.queryNodes() procedure
• Choosing similarity function (cosine vs euclidean) and embedding dimensions
• Post-filtering vector results with graph traversal (but retrieval_query patterns → graphrag-skill)

When NOT to Use

• GraphRAG pipelines (retrieval_query, HybridCypherRetriever) → use neo4j-graphrag-skill
• Fulltext / keyword search (FULLTEXT INDEX, db.index.fulltext.queryNodes) → use neo4j-cypher-skill
• GDS node embeddings (FastRP, Node2Vec) → use neo4j-gds-skill