Matching storage to your workload

Not every workload needs a dedicated vector database. Postgres with pgvector covers most simple use cases; purpose-built engines (Milvus, Qdrant, Pinecone) earn their complexity at scale, and text-search engines like OpenSearch solve a different problem (keywords, not meaning). The decision is really about scale, filters, latency budget, and what your team already operates.