Integrating your AI models in Collibra

When assessing the value and risks associated with an AI use case, it's important to be able to quickly identify which AI model is used by the use case. From the AI Use Case asset page, you can easily link to any AI Model assets (or assets of child asset types) that exist in your Collibra environment.

Tip For complete details on the out-of-the-box attribute types and relation types specific to the AI Model asset type, go to AI Governance operating model.

When you're linking to an AI model, all relevant assets appear in an Assets drop-down list. But what if you can't find your AI model?

Can't find your AI model?

You're trying to link an AI model to your use case, but no assets appear in the Assets drop-down list, or you can't find the right one. What then?

As shown in the following image, you have three options.

Supported integrations

We offer the following Collibra-supported AI integrations:

• AWS Bedrock AI
• AWS SageMaker AI
• Azure AI Foundry (beta)
• Azure ML
• Databricks Unity Catalog
• Google Vertex AI
• MLflow AI (beta)
• SAP AI Core

Important These integrations are only available via Edge, not via Jobserver.

During an integration, Machine Learning (ML) model metadata is ingested as assets on the Collibra Platform. The following table shows the asset types associated with each integration.

Integration	Assets of this type are ingested in Data Catalog	Parent asset type
AWS Bedrock AI	AWS Bedrock AI Model	AI Model
AWS SageMaker AI	AWS SageMaker AI Model	AI Model
Azure AI Foundry (beta)	Azure AI Foundry Model	AI Model
	Azure AI Foundry Agent	AI Agent
Azure ML	Azure AI Model	AI Model
Databricks Unity Catalog	Databricks AI Model	AI Model
Google Vertex AI	Vertex AI Model	AI Model
MLflow AI (beta)	MLflow AI Model	AI Model
SAP AI Core	SAP AI Model	AI Model

These integration-specific asset types are child asset types of the AI Model asset type.

All assets of these types in your Collibra environment appear in the Assets drop-down list when linking AI models to your registered AI use cases.

Custom integration

If our Collibra-supported integrations don't suit your needs – maybe you use a different AI model provider or you want to integrate a proprietary system – you can perform a custom integration.

On the Collibra Developer Portal, you can find a tutorial explaining how to use Python to create and synchronize AI Models in Collibra.

Metadata from your ML models is ingested as assets in Collibra. All AI Model assets (and assets of child asset types) appear in the Assets drop-down list when linking AI models to your registered AI use cases.

Manually create an AI Model asset

You can create AI Model assets via the global Create button.

Requirements and permissions

• You have a resource role with the Asset > Add resource permissions. for the community or domain in which the AI Model asset will be created.
  Tip This permission is granted through a resource role, for example the out-of-the-box Business Steward role, that is assigned to you either on the domain or on the community to which the domain belongs (via inheritance). For more information, go to Responsibilities.

Steps

1. On the main toolbar, click .
   The Create dialog box appears.
2. Click the Assets tab, then click AI Model.
   The Create Asset dialog box appears.
3. Enter the required information.

   Field	Description
   Asset type	The asset type of the asset that you are creating. In this case, AI Model.
   Domain	The domain to which the asset will belong. Tip Ensure that the domain type of the domain you select is assigned to the AI Model asset type.
   Asset name	A name to identify the asset. Asset names have a character limit of 2000 characters. Tip  You can simultaneously create multiple assets: type the first name, then click the drop-down or press Enter and type the next name. Depending on the settings, asset names may need to be unique in their domain.

4. Click Create.
   A message stating that one or more assets are created appears in the upper-right corner of the page.