Vector Databases and Graph Databases Explained, and Who Actually Sells Them
This piece explains vector databases (meaning as coordinates, similarity search, retrieval-augmented generation) and graph databases (entities and relationships as first-class records) in plain English, then shows why AI agents moved both from niche to mainstream through RAG, agent memory and GraphRAG, including Microsoft's open-source GraphRAG work. It maps the vendor landscape as checked against vendor sites on 6 August 2026 — Pinecone, Weaviate, Qdrant, Milvus/Zilliz and Chroma; vector search inside Postgres, MongoDB, Elastic, Redis and the hyperscalers including Amazon S3 Vectors; Neo4j, Neptune, Memgraph, Arango, TigerGraph and FalkorDB in graph; Zep, Mem0 and Letta in agent memory — with UK hosting availability, published pricing, the documented cost-unpredictability of usage-based tiers, and the requirement-led questions that decide whether a dedicated purchase is needed at all.
Daniel Thomas 15 September 2026