Associate workspaces with an AI use case to organize assets so they align with the phases of an AI solution. For example, associate a project or space for assets in the Development phase, associate a project or space for assets in the Validation phase, and associate a space for assets in the Operation phase.
As you plan an AI use case to define and address a business problem, consider the organization of assets you plan to track. The following sequence is a typical scenario.
- Create a use case and provide the basic details.
- Associate a project for developing and testing an asset for your business problem. When you start tracking the asset in the associated use case, facts for the asset creation and testing are recorded in the Development phase of the use case and associated factsheet.
- Next, associate a workspace for validation. For a prompt template, you might use a second project. Copy the asset to the validation project, upload validation data, and evaluate the asset. For a machine learning model, you can associate a pre-production deployment space where you can evaluate or monitor a deployed model. Facts that are captured in the project or space are recorded in the Validation section of the use case and associated factsheet.
- Finally, associate a production deployment space with the AI use case where you can monitor assets in production. Facts that are captured from the deployed assets in production spaces are recorded in the Operation phase of the use case and associated factsheet.
Rules for associating workspaces
Note these requirements when you associate workspaces with an AI use case:
- A project can be associated with just one AI use case. Multiple projects might contribute assets or validations to a single AI use case, but each project is restricted to associating with a single-use case.
- A project can be associated with a single lifecycle phase for an AI use case. That is, you cannot use the same project to develop and validate an asset.
- A space can be associated with multiple AI use cases. That is, if you use a single space for deployed models in production, assets in that space might be tracked in different AI use cases. This distinction from how projects relate to an AI use case reflects how spaces are often used in ModelOps to manage a collection of deployments in a particular state.
The following graphic illustrates the relationship of associated workspaces with an AI use case.
Additional requirements for the Sydney data center
If you are using a watsonx.governance service provisioned from the Sydney region, an additional level of access control is required to track assets in an AI use case. Users with admin access to projects and spaces must explicitly authorize access to the metadata for assets in the workspace. If this access is not granted, users will get errors when trying to track assets or view metadata in factsheets.
To grant access to asset metadata:
- Click the Manage tab for the project or space.
- On the General tab, Click Grant access for the control labeled Reporting on asset metadata is allowed.
Associating a project or space with an AI use case
Follow these steps to associate a workspace with an AI use case.
Organize your workspaces before you start tracking assets in an AI use case. After asset tracking is enabled, you cannot update the associated workspace without losing tracking information for the assets. If you do change an associated workspace, you must enable tracking for assets in the workspace.
- Create a new AI use case.
- Click Associate workspaces.
- Choose the lifecycle phase for the workspace: Development, Validation, or Operation.
- Choose an existing project or space, or create a new one.
- Repeat to associate workspaces for the remaining lifecycle phases.
The associated workspaces tables provide a dashboard where you can view workspaces you can access, your role, and an indicator showing association with one or more AI use cases.
Migrating existing workspaces to an AI use case
If you tracked assets in an AI use case before associated workspaces were required, a migration process helps you associate workspaces with existing AI use cases.
When your software is updated to include the Associated workspaces feature, a migration process automatically associates workspaces when the required criteria are met. Migration is attempted for each AI use case that you open or otherwise update by tracking or untracking an asset.
Rules for migrating projects
An existing project is automatically associated with an AI use case when all of the following criteria are met at the time of the update:
- There is at least one tracked asset in this project.
- All AI assets in this project are tracked in the same AI use case.
- All tracked assets are in the same lifecycle phase (Development or Validation).
When these criteria are not met, you must manually resolve the organizational issues. For example, create one or more new projects and use the export/import features to move assets from the original project to the new projects to achieve a structure that aligns with the organizational requirements for tracking assets in an AI use case.
Rules for migrating spaces
An existing space is automatically associated with an AI use case when all of the following criteria are met at the time of the update:
- There is at least one tracked asset in the space.
- All tracked assets in the space are associated with the same lifecycle phase. For example, if the tracked assets in a space are all in the Operation phase, the space is automatically associated with the AI use case or cases for the tracked assets.
When these criteria are not met, you must manually resolve the organizational issues. For example, create new spaces and use the export/import features to move assets to new workspaces to achieve a structure that aligns with the organizational requirements for tracking assets in an AI use case.
Troubleshooting migration issues
When you begin the migration, a progress bar shows the process. If the migration fails or is incomplete, follow these guidelines to resolve problems with migration of workspaces.
Resolving issues with invalid project structures
Untrack, delete, or move assets from projects until you reach a one-to-one correspondence with project to AI use case. A project can have multiple assets that are tracked in the same use case if they are all in the same phase (development, validate, or operate).
Resolving issues with invalid space structures
Untrack, delete, or move assets from spaces until you reach a one-to-one correspondence with space to AI use case. A space can track multiple assets in the same use case or multiple use cases as long as they are all in the same phase (development, validate, or operate) for each use case.
Resolving issues with the Platform access catalog
Follow these steps to resolve issues with the Platform access catalog
- If you get an error that a Platform access catalog is not available, work with an administrator to set up or access a Platform access catalog, then restart the migration process.
- If the Platform access catalog is present but you have insufficient access, work with an administrator to resolve the access issues, then restart the migration process.
Parent topic: Setting up an AI use case