Project-level Machine Learning Model Persistence in Aura Graph Analytics

We are excited to announce a major upgrade to how machine learning models are saved in Aura Graph Analytics. You can now persistently store and manage trained machine learning models in the Model Catalog at the project level!

Previously, models created within Aura Graph Analytics (AGA) were ephemeral and limited to the session in which they were created, while existing persistence in AuraDS was limited to the instance or database level. With this update, your models are stored safely in a persistent catalog and can be reused in later sessions across the project, even after your original Graph Data Science (GDS) session has ended.

Scope and Availability 🌐

Models are now securely scoped to your environment:

Project-wide access
: Models can be accessed by GDS Sessions created within the same Aura project.
Provider & Region
: Models are available to sessions operating in the same cloud provider and region where they were stored.
User Ownership
: Access to a model is currently limited to the specific user who originally trained and saved it (i.e., the owner of the session that stored the model initially). (
Note: Sharing models with other users via

gds.model.publish

is coming soon!
)

How to use it 🛠️

You can start taking advantage of the project-level Model Catalog right away through:

Cypher
: Manage your models directly using standard Cypher model catalog operations (
note: access via the Cypher API requires an additional

sessionName

parameter
).
GDS Python Client
: Available natively for users on version 1.21 and newer.

For complete details on supported model catalog operations, check out the official documentation.