pgvector is the de-facto standard extension for vector similarity search in PostgreSQL. Its strongest argument is operational: if you already run Postgres, vector search is one CREATE EXTENSION away - no additional database to deploy and operate.
Hybrid search works with plain SQL: combine vector similarity with Postgres full-text search and ordinary joins/metadata filters in one query (see Smart Vector Database Usage).
For larger scale there are drop-in performance extensions (pgvectorscale, VectorChord) with faster index builds and better throughput at high recall - benchmarks in 2026 show Postgres-based setups competitive with dedicated vector databases well into the tens of millions of vectors.
Team Insights
We run pgvector in production in customer projects. The operational simplicity is the deciding factor: Postgres is already operated and understood, so pgvector adds RAG-scale vector search without a new infrastructure component.
Recommendation
If Postgres is already part of your stack, evaluate pgvector before introducing a dedicated vector database - for typical RAG workloads it is usually sufficient and cheaper to operate. Revisit dedicated options when scale or feature requirements demand it.