We are pleased to announce the availability of a new Persistent Data Service (PDS) option: pgvector, a PostgreSQL extension for storing and querying vector embeddings.

pgvector is now available to all LC users in both the CZ and RZ via LaunchIT:

• CZ: https://launchit.llnl.gov or https://lit.llnl.gov
• RZ: https://rzlaunchit.llnl.gov

What is pgvector and why would I use it?

With growing interest in AI/ML workflows at LC, many users need a convenient way to store and query dense vector embeddings (for example, from language models or other feature extractors). pgvector provides:

• A familiar PostgreSQL database environment
• A vector data type for storing embeddings

Documentation discussing pgvector, and it's use within the LC ecosystem can be found here:

https://hpc.llnl.gov/services/cloud-services/persistent-data-services-pds/embedded-storage-pgvector

Documentation for provisioning a new PDS instance via Launchit is available here: 

https://hpc.llnl.gov/services/cloud-services/launchit

And documentation for working with postgresql in LC can be found here:

https://hpc.llnl.gov/services/cloud/pds/postgresql