Once you have set up your JFrog ML account and your AI catalog entitlement, all you need to do to start allowing models is follow these simple steps:
Workflow for Using the AI Catalog
Access the AI Catalog
From the JFrog Platform menu, navigate to AI/ML > Models to open the AI Catalog.
The AI Catalog has three main tabs:
Registry: This is the default view. It lists the models approved for use in your organization. If you are a new user, this tab is empty until models are approved.
Discovery: Explore all available models provided by approved (supported) providers that you can review and approve for use.
Detection: View and manage all package* type models found in your organization’s artifacts as scanned by JFrog Xray.
To explore all available models, switch to the Discovery tab. This tab displays the full list of models that can be reviewed and allowed for use within your organization.
Note
While all users can consume allowed models, only Admin level users have the permission to decide which models are allowed for use within your organization.
By default, on first use, no models have been allowed.
Finding Your First Model
Since the Registry tab is empty on first use, your first action is to switch to the Discovery tab. Browse the comprehensive catalog of external APIs, model packages, and your own custom models to discover the best fits for your project requirements.
Use the Filter & Search functionality to quickly pinpoint the most suitable models.
The JFrog platform automatically scans each model for vulnerabilities, and offers transparent license information to help you avoid compliance issues.
Note
Even if your AI Catalog is empty, that does not mean your organization is not using any models. It maybe be using unmanaged models. Using the Shadow AI detection feature, which uses Xray to scan your JFrog artifacts, you can detect which models are already being used, and manage them in the AI Catalog.
Allow Models
Next, governance becomes straightforward and effective. Admin users can easily allow models (from the Discovery tab) for secure use within your organization. Models can be approved on a per-project basis, with a comprehensive list of allowed models available on the Registry (allowed models) tab.
You can create an allow-list of models, ensuring that if a model is not on the list, it cannot be used, thereby preventing unvetted models from entering your supply chain.
Using an Allowed Model
Once a model has been successfully allowed in a project, it moves into the Registry tab and is ready for use. The steps for model consumption vary depending on the model type (Package or External API).
See: Allow Your First Model | Discover and Allow Models | Get Started with JFrog ML