Import Models Using Lakehouse SQL Endpoint Cannot Refresh Without Creating an Undocumented SQL Serve

[barztown]

When publishing an Import-mode semantic model built on the Lakehouse SQL Analytics endpoint, the dataset fails to refresh with the error:

“This semantic model uses a default data connection without explicit credentials.”

The Data source credentials pane in the Service is greyed out.
The Service provides no way to enter credentials.
The UI gives no instructions for fixing this.

After hours of debugging, the only working solution is:

1. Go to Manage connections and gateways

2. Create a Cloud connection

3. Set Connection type = SQL Server

4. Manually enter the Lakehouse SQL endpoint server + database

5. Authenticate with OAuth

6. Rebind the semantic model to this connection

This is:

• Completely undocumented

• Counterintuitive (why SQL Server instead of Lakehouse or Fabric SQL?)

• Not discoverable through the UI

• Impossible for a normal user to figure out

• A huge blocker for Import mode adoption

What’s worse:
Users naturally try Fabric SQL database, Lakehouse, or Azure SQL —
but these do not work and they never show the required server/database fields.

Only SQL Server works, even though the endpoint is not SQL Server.

Why this matters:
The entire reason I used Import mode was because Direct Lake cannot be shared in a Power BI App unless every user is added to the workspace.
Import mode should be the “safe/compatible” option — but in this case it simply breaks unless you know a hidden workaround.

Ask / Proposal:

• Fix the refresh pipeline so Import models using Lakehouse SQL endpoints do not depend on a hidden SQL Server cloud connection
  OR

• Update the UX so that creating a Lakehouse SQL endpoint connection is intuitive:

  • Expose server/database fields for Lakehouse connections

  • Provide a visible “Fix this connection” button

  • Remove or explain the “default data connection” state

• Add proper documentation explaining how to configure cloud connections for Lakehouse SQL endpoints

This is a major UX gap that makes Fabric far more confusing and error-prone than it needs to be.