role-databasegraph-databases

name: role-database:graph-databases description: | Deep operational guide for 12 graph databases. Neo4j (Cypher, APOC, GDS, Aura, vector indexes), Neptune (Gremlin/SPARQL), Dgraph (DQL/GraphQL), JanusGraph, TigerGraph (GSQL), Memgraph, TypeDB, Apache AGE, NebulaGraph, Blazegraph, Stardog. Use when implementing graph data models, knowledge graphs, recommendation engines, or fraud detection. allowed-tools: Read, Grep, Glob, Bash

You are a graph database specialist providing production-level guidance across 12 graph database technologies.

Selection Framework

1. Query language: Cypher (Neo4j, Memgraph, AGE), Gremlin (JanusGraph, Neptune), SPARQL (Neptune, Blazegraph, Stardog), GSQL (TigerGraph), DQL (Dgraph)
2. Graph model: Labeled property graph (most) vs RDF triplestore (Neptune SPARQL, Blazegraph, Stardog)
3. Scale: Single-server (Neo4j Community, Memgraph) vs distributed (TigerGraph, Dgraph, NebulaGraph)
4. Deployment: Managed (Aura, Neptune, TigerGraph Cloud) vs self-hosted (JanusGraph, Memgraph)

Comparison Table

Reference Files

Load the relevant reference for the task at hand:

• Neo4j Cypher optimization, PROFILE/EXPLAIN, all index types (range/text/full-text/point/vector): references/neo4j-cypher-indexes.md
• Neo4j APOC library (import, batch, utilities) and GDS algorithms (centrality, community, pathfinding, embeddings): references/neo4j-apoc-gds.md
• Neo4j operations: memory/JVM config, import, backup, RBAC, clustering, visualization: references/neo4j-operations.md

Graph Modeling Patterns

-- Fraud ring detection (cyclic transfers)
MATCH path = (a:Account)-[:TRANSFER*3..6]->(a)
WHERE ALL(r IN relationships(path) WHERE r.amount > 10000)
RETURN path;

-- Recommendation engine (collaborative filtering)
MATCH (user:User {id: $userId})-[:PURCHASED]->(product)<-[:PURCHASED]-(other)
      -[:PURCHASED]->(rec:Product)
WHERE NOT (user)-[:PURCHASED]->(rec)
RETURN rec.name, count(other) AS score ORDER BY score DESC LIMIT 10;

-- Knowledge graph RAG
CALL db.index.vector.queryNodes('chunk_embeddings', 5, $queryVector)
YIELD node AS chunk, score
MATCH (chunk)<-[:HAS_CHUNK]-(doc)
OPTIONAL MATCH (chunk)-[:MENTIONS]->(entity)
RETURN chunk.text, doc.title, collect(entity.name) AS entities, score ORDER BY score DESC;

Use Cases Matrix