Using Graph Databases

Graph Databases

Purpose

This skill guides selection and implementation of graph databases for applications where relationships between entities are first-class citizens. Unlike relational databases that model relationships through foreign keys and joins, graph databases natively represent connections as properties, enabling efficient traversal-heavy queries.

When to Use This Skill

Use graph databases when:

• Deep relationship traversals (4+ hops): "Friends of friends of friends"
• Variable/evolving relationships: Schema changes don't break existing queries
• Path finding: Shortest route, network analysis, dependency chains
• Pattern matching: Fraud detection, recommendation engines, access control

Do NOT use graph databases when:

• Fixed schema with shallow joins (2-3 tables) → Use PostgreSQL
• Primarily aggregations/analytics → Use columnar databases
• Key-value lookups only → Use Redis/DynamoDB

Quick Decision Framework

DATA CHARACTERISTICS?
├── Fixed schema, shallow joins (≤3 hops)
│   └─ PostgreSQL (relational)
│
├── Already on PostgreSQL + simple graphs
│   └─ Apache AGE (PostgreSQL extension)
│
├── Deep traversals (4+ hops) + general purpose
│   └─ Neo4j (battle-tested, largest ecosystem)
│
├── Multi-model (documents + graph)
│   └─ ArangoDB
│
├── AWS-native, serverless
│   └─ Amazon Neptune
│
└── Real-time streaming, in-memory
    └─ Memgraph

Core Concepts

Property Graph Model

Graph databases store data as:

• Nodes (vertices): Entities with labels and properties
• Relationships (edges): Typed connections with properties
• Properties: Key-value pairs on nodes and relationships

(Person {name: "Alice", age: 28})-[:FRIEND {since: "2020-01-15"}]->(Person {name: "Bob"})

Query Languages

Common Cypher Patterns

Reference references/cypher-patterns.md for comprehensive examples.

Pattern 1: Basic Matching

// Find all users at a company
MATCH (u:User)-[:WORKS_AT]->(c:Company {name: 'Acme Corp'})
RETURN u.name, u.title

Pattern 2: Variable-Length Paths

// Find friends up to 3 degrees away
MATCH (u:User {name: 'Alice'})-[:FRIEND*1..3]->(friend)
WHERE u <> friend
RETURN DISTINCT friend.name
LIMIT 100

Pattern 3: Shortest Path

// Find shortest connection between two users
MATCH path = shortestPath(
  (a:User {name: 'Alice'})-[*]-(b:User {name: 'Bob'})
)
RETURN path, length(path) AS distance

Pattern 4: Recommendations

// Collaborative filtering: Products liked by similar users
MATCH (u:User {id: $userId})-[:PURCHASED]->(p:Product)<-[:PURCHASED]-(similar)
MATCH (similar)-[:PURCHASED]->(rec:Product)
WHERE NOT exists((u)-[:PURCHASED]->(rec))
RETURN rec.name, count(*) AS score
ORDER BY score DESC
LIMIT 10

Pattern 5: Fraud Detection

// Detect circular money flows
MATCH path = (a:Account)-[:SENT*3..6]->(a)
WHERE all(r IN relationships(path) WHERE r.amount > 1000)
RETURN path, [r IN relationships(path) | r.amount] AS amounts

Database Selection Guide

Neo4j (Primary Recommendation)

Use for: General-purpose graph applications

Strengths:

• Most mature (2007), largest community (2M+ developers)
• 65+ graph algorithms (GDS library): PageRank, Louvain, Dijkstra
• Best tooling: Neo4j Browser, Bloom visualization
• Comprehensive Cypher support

Installation:

# Python driver
pip install neo4j

# TypeScript driver
npm install neo4j-driver

# Rust driver
cargo add neo4rs

Reference: references/neo4j.md

ArangoDB

Use for: Multi-model applications (documents + graph)

Strengths:

• Store documents AND graph in one database
• AQL combines document and graph queries
• Schema flexibility with relationships

Reference: references/arangodb.md

Apache AGE

Use for: Adding graph capabilities to existing PostgreSQL

Strengths:

• Extend PostgreSQL with graph queries
• No new infrastructure needed
• Query both relational and graph data

Reference: Implementation details in examples/

Amazon Neptune

Use for: AWS-native, serverless deployments

Strengths:

• Fully managed, auto-scaling
• Supports Gremlin AND SPARQL
• AWS ecosystem integration

Graph Data Modeling Patterns

Reference references/graph-modeling.md for comprehensive patterns.

Best Practice 1: Relationships as First-Class Citizens

Anti-pattern (storing relationships in node properties):

// BAD
(:Person {name: 'Alice', friend_ids: ['b123', 'c456']})

Pattern (explicit relationships):

// GOOD
(:Person {name: 'Alice'})-[:FRIEND]->(:Person {id: 'b123'})
(:Person {name: 'Alice'})-[:FRIEND]->(:Person {id: 'c456'})

Best Practice 2: Relationship Properties for Metadata

// Track interaction details on relationships
(:Person)-[:FRIEND {
  since: '2020-01-15',
  strength: 0.85,
  last_interaction: datetime()
}]->(:Person)

Best Practice 3: Bounded Traversals for Performance

// SLOW: Unbounded traversal
MATCH (a)-[:FRIEND*]->(distant)
RETURN distant

// FAST: Bounded depth with index
MATCH (a)-[:FRIEND*1..4]->(distant)
WHERE distant.active = true
RETURN distant
LIMIT 100

Best Practice 4: Avoid Supernodes

Problem: Nodes with thousands of relationships slow traversals.

Solution: Intermediate aggregation nodes

// Instead of: (:User)-[:POSTED]->(:Post) [1M relationships]

// Use time partitioning:
(:User)-[:POSTED_IN]->(:Year {year: 2025})
       -[:HAS_MONTH]->(:Month {month: 12})
       -[:HAS_POST]->(:Post)

Use Case Examples

Social Network

Schema and implementation in examples/social-graph/

Key features:

• Friend recommendations (friends-of-friends)
• Mutual connections
• News feed generation
• Influence metrics

Knowledge Graph for AI/RAG

Integration example in examples/knowledge-graph/

Key features:

• Hybrid vector + graph search
• Entity relationship mapping
• Context expansion for LLM prompts
• Semantic relationship traversal

Integration with Vector Databases:

# Step 1: Vector search in Qdrant/pgvector
vector_results = qdrant.search(collection="concepts", query_vector=embedding)

# Step 2: Expand with graph relationships
concept_ids = [r.id for r in vector_results]
graph_context = neo4j.run("""
  MATCH (c:Concept) WHERE c.id IN $ids
  MATCH (c)-[:RELATED_TO|IS_A*1..2]-(related)
  RETURN c, related, relationships(path)
""", ids=concept_ids)