What is Property Graph?
A property graph is a graph data model where both nodes and edges can have properties (key-value pairs) and labels (types or categories), providing a flexible and expressive way to represent complex, heterogeneous data. Unlike simple graphs that just connect entities, property graphs attach rich information to both entities and relationships. For example, a Person node might have properties like name, age, and email, while a KNOWS relationship might have properties like since_date and relationship_type.
The property graph model supports multiple labels per node (allowing entities to belong to multiple categories) and typed, directed edges with properties. This flexibility makes property graphs well-suited for real-world knowledge representation where entities and relationships are rarely simple or uniform. Queries can filter based on node labels, edge types, and property values, and can traverse relationships while accessing the rich property information attached to each element.
Property graphs are the data model used by databases like Neo4j, Amazon Neptune, and Azure Cosmos DB. They differ from RDF triple stores, which use a more rigid subject-predicate-object model without direct support for properties on relationships. In AI agent systems, property graphs provide an excellent foundation for knowledge graphs that need to represent nuanced, real-world information with entities that have multiple types and relationships that carry important contextual information. The query language Cypher, designed for property graphs, makes it easy to express complex graph patterns and traversals.