Neo4j, a graph database and analytics company, announced that it has integrated native vector search as part of its core database capabilities. The result enables customers to achieve richer insights from semantic search and generative AI applications, and serve as long-term memory for LLMs, while reducing hallucinations.
Neo4j’s graph database can be used to create knowledge graphs, which capture and connect explicit relationships between entities, enabling AI systems to reason, infer, and retrieve relevant information effectively. The result ensures more accurate, explainable, and transparent outcomes for LLMs and other generative AI applications. By contrast, vector searches capture implicit patterns and relationships based on items with similar data characteristics, rather than exact matches, which are useful when searching for similar text or documents, making recommendations, and identifying other patterns.
This latest advancement follows from Neo4j’s recent product integration with Google Cloud’s generative AI features in Vertex AI in June, enabling users to transform unstructured data into knowledge graphs, which users can then query using natural language and ground their LLMs against factual set of patterns and criteria to prevent hallucinations.