Description
Neo4j Vector Search is a powerful tool designed to facilitate efficient and robust approximate nearest neighbor searches using vector indexes. It implements the Hierarchical Navigatable Small World (HNSW) algorithm to create layers of k-nearest neighbors, which is particularly useful for working with embeddings generated by machine learning models. This functionality allows users to find similar nodes in a graph based on their embeddings, making it an invaluable resource for applications that require similarity searches.
Read More
The tool supports creating a vector index with a specified number of dimensions and similarity functions, such as Euclidean and cosine, for both nodes and relationships. Users can query a vector index with an embedding and a top-k value, which returns nodes along with their similarity scores. Neo4j Vector Search also includes functions and procedures in the GenAI plugin to compute text embeddings with providers like OpenAI, Azure OpenAI, Google Vertex AI, and Amazon Bedrock.
For current Neo4j releases, the recommended approach is to use Cypher 25 with the SEARCH clause for vector search, and ai.text.embed() and ai.text.embedBatch() for generating embeddings. This ensures that only embeddings of the expected size are indexed, preventing mismatched-dimension queries from failing.
Hybrid search capabilities are also available, combining ranked results from multiple retrieval signals into a single result list. This is particularly useful for searches that require both vector similarity and exact term matching, such as names, acronyms, or identifiers.
Neo4j Vector Search is suitable for developers and data scientists looking to leverage graph databases for AI-powered applications. It offers comprehensive documentation and community support, making it accessible for both new and experienced users. The tool is part of Neo4j's broader ecosystem, which includes various graph database solutions and AI integrations.
- CategoryVector Databases & Retrieval
- Pricing ModelFreemium
- DeploymentSelf-hosted, Cloud
- PlatformsCloud
Neo4j Vector Search's Core Features
HNSW algorithm for k-nearest neighbors
Supports embeddings from machine learning models
Vector index creation with specified dimensions
Similarity functions: Euclidean and cosine
Integration with OpenAI, Azure, Google, Amazon
Hybrid search combining multiple retrieval signals
Cypher 25 support for vector search
GenAI plugin for text embeddings
How to use Neo4j Vector Search?
Configure: Set up vector index dimensions and similarity functions
Use: Query vector index with embeddings
Integrate: Use GenAI plugin for text embeddings
Optimize: Combine hybrid search for comprehensive results
Neo4j Vector Search's Use Cases
- Graph similarity search
- AI-powered applications
- Hybrid search
- Machine learning integration
- Data science projects
