pgvector vs Dedicated Vector DBs

The one-line verdict

• Data already in Postgres? pgvector — vectors as a column, similarity in SQL, one backup to manage.
• Outgrowing it? Qdrant / Pinecone / Chroma (see our vector-DB showdown) — purpose-built scale and filtering.

Official homes: pgvector · Supabase AI guide · Qdrant.

pgvector in 5 lines

CREATE EXTENSION IF NOT EXISTS vector;
CREATE TABLE docs (
  id bigserial PRIMARY KEY,
  content text,
  embedding vector(1536)
);
CREATE INDEX ON docs USING hnsw (embedding vector_cosine_ops);
SELECT content FROM docs ORDER BY embedding <=> '[0.1, 0.2, ...]' LIMIT 4;

When dedicated wins

• Billions of vectors or strict p99 latency budgets.
• Metadata filtering at a scale where a single Postgres box sweats.
• You want managed uptime someone else wakes up for.

Until then, pgvector plus the chunking and eval discipline from our RAG chapters beats a fancier database with sloppy retrieval.