Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
Use Microsoft OneLake as a knowledge source for Microsoft Foundry. You can connect directly and securely to OneLake from Foundry, index unstructured and semi structured files stored in OneLake (including files that arrive through shortcuts), and then use that indexed content as a knowledge source inside agents in Foundry.
With this integration, you can ground your agents on the same enterprise data that already lives in OneLake, instead of creating new copies of files in separate AI specific stores. Permissions and governance are enforced through the same OneLake and Fabric controls that you use for analytics workloads.
Prerequisites
A lakehouse in Fabric. If you don't have a lakehouse, follow the steps in Create a lakehouse with OneLake.
- Files in the Files folder of the lakehouse.
A Foundry project. If you don't have one, follow the steps in Create a project.
An Azure AI Search service at the Basic tier or higher. If you don't have one, follow the steps in Create an Azure AI Search service.
The search service must be in the same tenant as your Fabric workspace.
In this article, you create an assign a managed identity for the search service. To create a managed identity, you must be an Owner or User Access Administrator roles. To assign roles, you must be an Owner, User Access Administrator, Role-based Access Control Administrator, or a member of a custom role with Microsoft.Authorization/roleAssignments/write permissions.
Index data from OneLake files
Use Azure AI Search to configure a OneLake files indexer to make your lakehouse data searchable as a knowledge source.
Review the prerequisites in Index data from OneLake files and shortcuts > Prerequisites.
Then, follow the steps for system managed identity in Index data from OneLake files and shortcuts > Grant permissions.
Create a OneLake connection in Foundry
Sign in to Microsoft Foundry.
Make sure the New Foundry toggle is On. The steps in this article refer to Microsoft Foundry (new).
Open the project that you want to work in.
Select Build from the navigation menu, then select Knowledge from the left pane.
Select your AI Search resource.
Select Create a knowledge base.
Select Microsoft OneLake as the knowledge type. Select Connect.
Provide your Fabric workspace ID and lakehouse ID.
You can retrieve both of these IDs from your lakehouse URL:
https://app.powerbi.com/groups/<WORKSPACE_ID>/lakehouses/<LAKEHOUSE_ID>.
Select Create.
Select Save knowledge base.