SurrealDB vs. vector databases
Pinecone, Chroma, Weaviate, and Qdrant are built for one thing: vectors. SurrealDB is multi-model - native vectors alongside documents, graphs, and structured data in one database.
Vector databases store embeddings, but not the entities, relationships, or metadata that give embeddings meaning. No graph traversal, no ACID transactions, no temporal versioning. Context requires more than similarity search.SurrealDB unifies vectors with the rest of your data model.
THE LIMITATION
Vectors alone are not enough
No structured data
Vector databases store embeddings. They don't store the entities, relationships, or metadata that give embeddings meaning.
No graph traversal
You can't traverse relationships between entities. A user and their purchases live in separate systems - no join, no graph.
No transactions
Updating a vector and its source document requires two systems. No ACID - one can succeed while the other fails.
No temporal
Vector databases don't track how knowledge evolves. No bi-temporal versioning, no historical queries.
THE DIFFERENCE
Vectors as part of the whole
Co-located data
Embeddings live alongside their source documents, entities, and relationships. One record, one transaction.
Hybrid queries
Combine vector similarity, graph traversal, full-text search, and structured filters in a single SurrealQL statement.
ACID
Read-think-write loops commit atomically. Update memory and state in one transaction - both succeed or neither does.
Unified permissions
One permission model for documents, graphs, vectors, and memory. RBAC and record-level access in one place.
THE DEPENDENCY
Memory middleware and context layers still depend on vectors
Memory middleware (Mem0, Zep, Letta) and context layer platforms (Honeydew, Atlan) abstract over vector databases, but the underlying limitations persist. Similarity is not relevance. Fragments are not context. SurrealDB eliminates the abstraction by unifying vectors with graphs, documents, and temporal data in one engine.
Data fragments at every system boundary
BEYOND VECTORS
From vector search to structured memory
If you are building agent memory, vectors are just one retrieval signal. Spectron combines vector similarity with knowledge graphs, entity extraction, temporal fact tracking, and hybrid retrieval - all running on SurrealDB in a single ACID transaction.
GET STARTED
The multi-model database for AI
Documents, graphs, vectors, time-series - unified in one query language. No bolt-ons, no glue code.
