Neo4j
A graph database officially released in 2007 and considered a benchmark product in the graph database domain. It is widely used in areas such as financial risk control, social recommendation, and fraud detection.
Core Facts
The GraphRAG approach extracts entities and relationships from documents, stores them as nodes and edges in a graph database, and during retrieval first uses vector search to recall candidate nodes before performing multi-hop reasoning along the graph's edges.
Microsoft's 2024 research experiments demonstrated that GraphRAG significantly outperforms standard RAG on global questions requiring synthesis and reasoning across multiple documents, but with increased index construction costs and query latency.
Graph databases (such as Neo4j and Amazon Neptune) use nodes and edges as their native data structures, allowing multi-hop queries without the need for expensive table join operations.
Timeline (last 90 days)
OntoGr runs on top of FalkorDB or Neo4j, and can optionally use pgvector or Postgres; a background consolidation task handles deduplication, summarization, scoring, and archiving; the LLM and embedder are bring-your-own.
Common implementations of property graph databases include Neo4j and JanusGraph.
Agentic GraphRAG finds paragraphs through Neo4j's native full-text index and vector index, and uses reciprocal rank fusion (RRF) to fuse keyword rankings and semantic rankings.
The Agentic GraphRAG project stores the knowledge layer, working memory layer and decision memory layer in the same Neo4j graph, with the three layers connected to each other by edges.
Graphiti uses the Apache-2.0 license, supports Neo4j, FalkorDB, or Amazon Neptune, provides a temporal knowledge graph with validity intervals on both entities and relationships, and can query what was true on a given day; it is delivered as an engine rather than a service, requiring you to build identity, authentication, queues, backups, and application APIs yourself.
Graphiti is Zep's open-source temporal graph engine, supporting time-validity intervals for entities and relationships, able to answer "what was true at a given point in time", requires pairing with a graph database (such as Neo4j, FalkorDB, Neptune), and provides a Server and an MCP adapter for self-hosting.
Cypher is a graph query language proposed by Neo4j, and FalkorDB is compatible with its syntax.
Relational SQL (Postgres/MySQL) collapses in performance on recursive JOIN cascades and CTE traversal to arbitrary depth; vector databases are non-deterministic and cannot tell you whether function A strictly calls function B; traditional graph databases (such as Neo4j) have a heavy JVM runtime, slow cold starts, and high memory usage, making them unsuitable for developer CLI tools.
In persistent graph databases such as Neo4j, creating a separate graph for a hypothetical what-if analysis usually involves disk writes and heavier initialization costs, and frequent creation and deletion is noticeably expensive.
FalkorDB is compatible with the OpenCypher standard, so developers familiar with Neo4j can migrate at low cost.
24 more timeline events
All Facts (20)
The GraphRAG approach extracts entities and relationships from documents, stores them as nodes and edges in a graph database, and during retrieval first uses vector search to recall candidate nodes before performing multi-hop reasoning along the graph's edges.
65%VerifiedMicrosoft's 2024 research experiments demonstrated that GraphRAG significantly outperforms standard RAG on global questions requiring synthesis and reasoning across multiple documents, but with increased index construction costs and query latency.
65%VerifiedGraph databases (such as Neo4j and Amazon Neptune) use nodes and edges as their native data structures, allowing multi-hop queries without the need for expensive table join operations.
65%VerifiedNeo4j uses the Cypher query language.
65%UnverifiedNeo4j采用属性图模型存储数据，使用免索引邻接（Index-free Adjacency）技术，使图遍历操作的时间复杂度仅与遍历的局部子图大小相关
90%UnverifiedNeo4j是图数据库领域最主流的代表，被LinkedIn、eBay等公司广泛采用
85%Unverified转型罗盘工具的技能图谱本质上是知识图谱的垂直应用，将技能节点、依赖关系、岗位要求结构化为图数据库（如Neo4j）中的实体与关系
85%UnverifiedNeo4j和TigerGraph等图数据库用于存储和查询代码知识图谱中的关系
85%UnverifiedGraph RAG通常以Neo4j等图数据库作为底层支撑
80%Partially VerifiedCypher is a declarative query language designed by Neo4j for property graph databases, and FalkorDB is compatible with a subset of its syntax.
65%UnverifiedKnowledge graphs store knowledge in the form of 'entity–relation–entity' triples, with Neo4j being the mainstream implementation tool.
60%UnverifiedThe graph structure of a knowledge graph natively supports graph traversal algorithms, allowing queries such as 'find all places that call function A' to complete in milliseconds.
60%UnverifiedOntoGr runs on top of FalkorDB or Neo4j, and can optionally use pgvector or Postgres; a background consolidation task handles deduplication, summarization, scoring, and archiving; the LLM and embedder are bring-your-own.
50%UnverifiedCommon implementations of property graph databases include Neo4j and JanusGraph.
50%UnverifiedThe Agentic GraphRAG project stores the knowledge layer, working memory layer and decision memory layer in the same Neo4j graph, with the three layers connected to each other by edges.
50%UnverifiedAgentic GraphRAG finds paragraphs through Neo4j's native full-text index and vector index, and uses reciprocal rank fusion (RRF) to fuse keyword rankings and semantic rankings.
50%UnverifiedGraphiti uses the Apache-2.0 license, supports Neo4j, FalkorDB, or Amazon Neptune, provides a temporal knowledge graph with validity intervals on both entities and relationships, and can query what was true on a given day; it is delivered as an engine rather than a service, requiring you to build identity, authentication, queues, backups, and application APIs yourself.
50%UnverifiedGraphiti is Zep's open-source temporal graph engine, supporting time-validity intervals for entities and relationships, able to answer "what was true at a given point in time", requires pairing with a graph database (such as Neo4j, FalkorDB, Neptune), and provides a Server and an MCP adapter for self-hosting.
50%UnverifiedCypher is a graph query language proposed by Neo4j, and FalkorDB is compatible with its syntax.
50%UnverifiedRelational SQL (Postgres/MySQL) collapses in performance on recursive JOIN cascades and CTE traversal to arbitrary depth; vector databases are non-deterministic and cannot tell you whether function A strictly calls function B; traditional graph databases (such as Neo4j) have a heavy JVM runtime, slow cold starts, and high memory usage, making them unsuitable for developer CLI tools.
50%