vector

vector (pgvector) is an extension that provides native vector data types and indexing capabilities for Postgres. All Crunchy Bridge clusters come with the extension available, but you need to enable it with the following command:

CREATE EXTENSION vector;
Copy

Once enabled, you can create a vector column with your tables:

CREATE TABLE my_table (
    id SERIAL PRIMARY KEY,
    embedding vector(1536)
)
Copy

Then, to create an index on the vector data, you can run:

CREATE INDEX ON my_table USING ivfflat (embedding vector_l2_ops) WITH (lists = 500);
Copy

Note that the vector size (that is, the number in the parenthesis for VECTOR(#)) on the column definition is important. Without it, you will see the following error when creating an index:

ERROR:  column does not have dimensionsCopy

We've created a few tutorials for using vector:

• Beginning vector: What's Postgres Got To Do With AI?
• Using vector: Performance Tips for Developers Using Postgres and vector
• See also our AI blog series

vector documentation:

• vector