Neo4j has introduced a new vector index in version 5.11 to efficiently perform semantic search over unstructured text or other embedded data modalities, making it a great fit for most RAG applications that rely on both structured and unstructured data. The Neo4j Vector Index implementation in LangChain allows customization of the node label, text, and embedding property names. Users can customize the node label to store text chunks under the specified node label, where the info property is used to store text, and the vector property holds the text embedding representation. The index implementation also creates a unique node property constraint on the id property for faster imports and allows users to load additional documents into an instantiated vector index. Additionally, users can use custom retrieval queries to collect, transform, or calculate any additional graph information they want to return from the similarity search. The newly added vector index implementation in Neo4j makes it a perfect fit for highly complex and connected datasets.