Skip to content

Latest commit

 

History

History
37 lines (23 loc) · 1.84 KB

README.md

File metadata and controls

37 lines (23 loc) · 1.84 KB

PostgreSQL + pgvector Setup

About pgvector

pgvector is an open-source vector similarity search for Postgres, allowing you to store and query vectors alongside your relational data. It supports exact and approximate nearest neighbor search, L2 distance, inner product, and cosine distance, and is compatible with any language that has a Postgres client.

Container Image

The provided Containerfile builds a PostgreSQL 15 + pgvector image, with pgvector built from source. You can deploy this container like any other PostgreSQL image.

Alternatively, you can use the prebuilt image available at https://quay.io/repository/rh-aiservices-bu/postgresql-15-pgvector-c9s.

Deployment

To deploy a PostgreSQL+pgvector instance, follow these steps:

  1. Create a secret using 01_db_secret.yaml, which defines the database name, username, and password.
  2. Create a Persistent Volume Claim (PVC) using 02_pvc.yaml, which persists the database data. Make sure to add a default storage class if necessary.
  3. Deploy the PostgreSQL server using 03_deployment.yaml.
  4. Expose the PostgreSQL server using 04_services.yaml.

After applying these files, you should have a running PostgreSQL+pgvector server accessible at postgresql.your-project.svc.cluster.local:5432.

Enabling PgVector Extension

To enable the PgVector extension, you need to connect to the running server Pod as a Superuser and execute the following command:

psql -d vectordb -c "CREATE EXTENSION vector;"

Replace vectordb with the name of your database if you changed it in the Secret.

If the command succeeds, it will print CREATE EXTENSION.

Troubleshooting

If you encounter any issues, make sure to check the pod logs and verify that the PgVector extension is enabled.