Vector vs Graph Memory Fresh

Choose the right memory architecture for your use case.

Vector Memory

Best for: semantic search, similarity matching, content-based retrieval.

• Fast approximate nearest neighbor search
• Good for "find similar" queries
• Simple setup and maintenance

Graph Memory

Best for: relationship tracking, multi-hop queries, entity networks.

• Preserves relationships between entities
• Enables path-based queries ("Who does Alice know who works at X?")
• Requires graph database backend (Neo4j, Memgraph, etc.)

Hybrid (Recommended)

Combine both for comprehensive memory:

config = {
    "vector_store": {"provider": "qdrant", "config": {...}},
    "graph_store": {"provider": "neo4j", "config": {...}},
}
memory = Memory.from_config(config)

Vector handles semantic search; graph handles relationship queries.