Skip to main content
Tracecat requires pgvector 0.8.0 or newer in the public schema of your application database, even when semantic search is disabled. The Temporal database does not need pgvector. There are two setup steps: install the extension files on the PostgreSQL server, then enable the extension in the application database. Tracecat’s migration performs the second step automatically if its database role has permission.

Install the server files

For your own server, follow the official pgvector instructions for your installation method: For RDS, follow AWS’s PostgreSQL extension setup guidance. Check that your engine supports the required version and that rds.allowed_extensions permits vector, if you use an allowlist. When upgrading an existing database, keep a backup and preserve a compatible PostgreSQL and OS base. Adding pgvector should not silently replace the libraries that your existing indexes depend on.

Enable the extension

Connect to the Tracecat application database and check what the server provides:
If the query returns no rows, install the server files or select a managed engine that supports pgvector. A null installed_version means the files are available but the extension is not enabled in this database. Tracecat’s migration runs:
If your migration role cannot create extensions, have a database administrator run that command in the application database before retrying the migration. You can keep the application role’s existing permissions.

Resolve a migration error

Pgvector setup failures include a link to this page in the migration logs. After correcting the prerequisite, rerun your deployment’s migration step. The migration does not upgrade or relocate an existing extension, and downgrading Tracecat retains it.