Introducing Neo4j’s Native Vector Data Type

Neo4j has introduced a new native Vector data type to enhance the handling of embedding vectors within its graph database, providing benefits such as simplified code, improved data integrity, and enabling future optimizations specific to vectors. This first-class Vector type, which replaces the previous method of storing vectors as lists of numbers, is fully integrated into Neo4j's system from drivers and Bolt protocol to the Cypher query language and storage engine. It addresses common issues like type mismatches and facilitates the implementation of vector-specific functions and indexes, making it ideal for applications in semantic search and other GenAI patterns. Users can now enforce property-type constraints to ensure data consistency and utilize Neo4j's vector indexes, powered by Apache Lucene's HNSW implementation, for efficient semantic retrieval. Existing projects can continue using list properties, but the new Vector type is recommended for new initiatives, especially with the 2025.10 rollout in Neo4j Aura, provided users employ Cypher 25 and version 6 drivers.