Why should you care?

PostgresVerse recommendation engine stores embeddings next to orders — one database for OLTP and similarity search.

See it live — copy this example

Run in pgAdmin or psql.

CREATE EXTENSION IF NOT EXISTS vector;

CREATE TABLE product_embeddings (
  product_id bigint PRIMARY KEY REFERENCES products(product_id),
  embedding vector(3)
);

INSERT INTO product_embeddings VALUES
  (1, '[0.1,0.2,0.9]'),
  (2, '[0.11,0.19,0.88]');

What happened?

• vector extension adds vector type.
• Store fixed-dimension embeddings beside relational product_id.
• Next lesson queries nearest neighbors.

Practice next

1. Install pgvector on your PostgreSQL build.
2. CREATE EXTENSION vector in PostgresVerse.
3. Create table and insert sample vectors.
4. List extensions: SELECT * FROM pg_available_extensions WHERE name LIKE 'vector%';
5. Plan IVFFlat index after loading sample data.

Remember

Extensions add types and functions without fork. pgvector is standard for embeddings in Postgres. Keep inference outside DB; store vectors inside.

PostgresVerse recsys bootstrap

Data team adds pgvector; nightly job writes embeddings from Python service.

Outcome: Similar products API queries same DB as catalog — no sync lag.