Manage inventory and curate assets in Amazon SageMaker Unified Studio
To use Amazon SageMaker Unified Studio to catalog your data, you must first bring your data (assets) into your project's inventory. Assets in a project's inventory are discoverable only to that project's members. After you curate an asset's metadata, you can publish it to the Amazon SageMaker Catalog to make it discoverable by all domain users, and share it with other projects.
Note
To curate, publish, or unpublish an asset, you must be the owner or a contributor of the project the asset belongs to.
Navigate to assets
To navigate to the Assets page
-
Navigate to Amazon SageMaker Unified Studio using the URL from your admin and log in using your SSO or AWS credentials.
-
From the project selector dropdown at the top of the page, choose the project the asset belongs to.
-
At the bottom of the left navigation pane, choose Manage.
-
Under Catalog management, choose Assets.
The Assets page lists the data assets and data products for the project. Use the search box to find assets, and the Filters panel to narrow the list:
-
Catalog scope – choose which assets to show:
-
Entire catalog – everything published to the catalog in this domain.
-
Published by this project – items this project has published to the catalog.
-
Owned by this project – all cataloged items in this project's inventory (default).
-
Subscribed – assets this project is subscribed to.
-
-
Asset type, Glossary terms, and Date range (Metadata last updated or Created at).
-
Custom – choose Add to filter by additional attributes.
You can switch between list and grid views using the icons in the top right.
Asset details page
Choose an asset's name to open its details page. The header shows the asset's name and publishing status (for example, Published), business name, asset type, and source location. The page has the following tabs.
| Tab | Description |
|---|---|
| Overview | The asset's summary, glossary terms, README, metadata forms, asset details, and associated data products. |
| Schema | The asset's columns, with their business names, data types, descriptions, glossary terms, and metadata forms. |
| Asset filter | Asset filters let you share a subset of the asset's rows or columns with subscribers instead of the whole asset. Choose Create asset filter to create one. Filters are used when you share with Filtered access (see Share an asset). |
| Preview data | A sample of the asset's rows. Use the search box to find rows. |
| Lineage | A graph of the asset's lineage, from upstream job runs and source tables to the inventory and published versions of the asset. Use the toolbar to zoom, fit, and refresh the graph. Toggle Show dataset nodes only or Display in event timestamp order, and set the Graph depth to control how many levels are shown. |
| Data profile | Profiling results for the asset's data. This tab appears only for assets that your project owns. Choose Run data profile to start a profiling run. The results appear after the run completes. |
| Data quality | Data quality monitoring for the asset. Rulesets lists the data quality rulesets for the table, Rule results shows the outcome of each rule, and Run history lists past evaluation runs. Choose Create to create a ruleset. |
| Subscriptions | Requests lists subscription requests for the asset. Access lists the projects that have been granted access, with each grant's status and date. Use the search box and Status filter to narrow either list. |
| History | The asset's revisions. See View asset revisions. |
Curate asset metadata
Each edit to an asset's business or technical metadata creates a new revision of the asset. See View asset revisions.
Name, business name, and description
In the asset header, choose the edit icon next to the Business name or description to update them.
Overview
On the Overview tab, choose the edit icon on any of the following sections:
-
Summary – a description of the asset's contents and use case.
-
Glossary terms – glossary terms that apply to the asset as a whole.
-
Readme – rich descriptions for the asset. The README editor supports Markdown (up to 4096 characters), so you can format descriptions and highlight key information for consumers. Select Side By Side Preview to see the rendered output while you edit, then choose Save.
-
Metadata forms – see Update metadata forms.
The Overview tab also includes the following sections:
-
Asset details – the asset's publish status, domain unit, owning project, creator, creation and update times, latest inventory revision, published revision, subscriber count, last modifier, and asset source identifier. You can edit the following settings by choosing the edit icon next to them:
-
Catalog visibility – whether the asset is Searchable in the catalog.
-
Subscription approval – whether subscription requests require approval (Required by default).
-
-
Associated data product – data products that include this asset. Choose Add to existing data product to add the asset to one.
Generate suggestions
Choose Generate suggestions to use AI to generate business names and descriptions for the asset and its columns. If automated business name generation was enabled when the data source was created, suggestions are generated automatically during data source runs. You can review suggestions and accept or reject them individually or all at once.
Update metadata forms
Metadata forms let you extend an asset's metadata model with custom-defined attributes (for example, sales region, sales year, and sales quarter). Metadata forms attached to an asset type are applied to all assets created from that asset type, and metadata forms attached to a domain are attached to all assets published to that domain. You can also add metadata forms to individual assets as part of a data source run or after the asset is created. To create new forms, see Create a metadata form.
On the Overview tab, under Metadata forms, do one of the following:
-
Choose the edit icon next to an existing form to edit its values.
-
Choose the edit icon next to the section header to re-arrange forms or remove them.
-
Choose Add Form and enter values for each field to attach an additional form.
Use Collapse all to collapse the forms' field lists.
Adding new fields or other updates to existing metadata forms will cause previously attached forms to become outdated. These forms display an Outdated chip next to their name in the Overview tab. Choosing the edit icon displays the difference between the form currently used by the asset and the latest form revision, and lets you update existing and newly added fields.
Some forms might be required as part of a metadata rule or custom asset type, or are inherited system forms from AWS Glue or other services. Such forms have a Required chip next to their name in the Overview tab, and cannot be deleted.
Curate the schema
To curate the schema
-
On the asset details page, choose the Schema tab. Use the search box to find columns by name, data type, description, or glossary term.
-
For the column you want to curate, choose the edit icon next to the value you want to change:
-
Name – the column's business name.
-
Description – the column's description.
-
Glossary terms – the column's glossary terms.
-
Metadata Forms – the column's metadata forms.
-
To edit a column's glossary terms, choose the edit icon in the Glossary terms column. In the Edit Glossary terms panel, choose Add terms to add terms, or choose the delete icon to remove a term, then choose Save changes.
Publish an asset
You can publish assets and their metadata from the project inventory to the Amazon SageMaker Catalog, making them discoverable by all domain users. Only the latest revision of an asset can be published.
Note
For Amazon Redshift assets, make sure that the Amazon Redshift clusters associated with both publisher and subscriber clusters meet all the requirements for Amazon Redshift data sharing so that Amazon SageMaker Unified Studio can manage access for Amazon Redshift tables and views. See Data sharing concepts for Amazon Redshift.
If you didn't choose to publish assets automatically when you created the data source, open the asset's details page and choose Publish asset.
By default, all assets require subscription approval, which means a data owner must approve all subscription requests to the asset. To change this setting, on the Overview tab under Asset details, choose the edit icon next to Subscription approval, then re-publish the asset.
Re-publish an asset
After you edit a published asset, a new revision is created and a New revision is available to publish message appears under Re-publish asset. Choose Re-publish asset to publish the latest revision to the catalog.
After re-publishing, the catalog shows the updated asset and the Re-publish asset button no longer appears.
In the project, Asset details shows both the Latest inventory revision and the Published revision. In the catalog, only the latest published revision appears.
Asset actions
On the asset details page, open the Actions menu to choose from the following:
-
Query – query the asset.
-
Create notebook – open the asset in a new notebook.
-
Share – share the asset with other projects. See Share an asset.
-
Add to existing data product – add the asset to a data product.
-
Create new data product with this asset – create a data product that contains the asset.
-
Unpublish – remove the asset from the catalog. See Unpublish an asset.
Share an asset
You can grant other projects access to a published asset.
To share an asset
-
On the asset details page, open the Actions menu and choose Share.
-
Under Select a project, choose the projects to share the asset with. Selected projects appear under Selected principals (up to 10).
-
Under Data permissions, the asset is shared with Read-only permissions, which let principals view data without the ability to modify, delete, or share it. Choose one of the following:
-
Full access – share the whole asset.
-
Filtered access – share only the rows and columns that an asset filter allows.
-
-
Choose Share.
Unpublish an asset
When you unpublish an asset, it no longer appears in global search results. New users can't find or subscribe to the asset in the catalog, but all existing subscriptions remain the same.
To unpublish an asset
-
On the asset details page, open the Actions menu and choose Unpublish.
-
In the confirmation dialog, choose Unpublish.
The asset is removed from the catalog. You can re-publish the asset at any time by choosing Publish asset.
View asset revisions
Amazon SageMaker Unified Studio creates a new revision of an asset whenever you edit its business or technical metadata. These edits include changes to the asset name, description, glossary terms, column names, metadata forms, and metadata form field values, whether made manually, by data source runs, or through API operations. After a new revision is created, you must re-publish the asset (see Re-publish an asset) for subscribers to see the update.
To view an asset's revisions, open the asset's details page and choose the History tab. The Revisions list shows each revision's Revision ID, Created by, and Created on. Revisions created by data source runs show SYSTEM as the creator. Use the search box to find a revision, and choose a revision ID to see the asset as it was at that point.
