Unable to Add a New Table to an Existing Microsoft Fabric Semantic Model

The required recovery depends on whether the model is Direct Lake on OneLake or Direct Lake on SQL; identify that first, then use the matching add-table path rather than removing parameterization again.

1. Identify the Direct Lake type

Use Power BI Desktop to inspect the model’s source expression:

1. Open a new instance of Power BI Desktop.
2. Select OneLake catalog and choose the existing semantic model.
3. From Connect, select Edit.
4. Open TMDL view and drag the Semantic model node into the editor.
5. Find the shared Expression near the bottom:
   • AzureStorage.DataLake(...) means Direct Lake on OneLake.
   • Sql.Database(...) means Direct Lake on SQL.

These are the documented source-expression types for the two Direct Lake architectures.^1^

Do not edit the TMDL yet; this inspection only determines which procedure applies.

2. If it is Direct Lake on OneLake

Add the Warehouse table through Desktop live editing, bypassing the failing browser operation:

1. While live-editing the model, select OneLake catalog.
2. Select the Fabric Warehouse containing the new table.
3. Select the table and choose OK.
4. Select Refresh to perform a schema refresh and reframe the Direct Lake tables.
5. Verify that the table appears in Model view.

Changes are applied directly to the semantic model in the Fabric workspace; there is no separate publish operation.^2^

Alternatively, in Fabric:

1. Semantic model → Open data model.
2. Switch from Viewing to Editing.
3. Select OneLake catalog.
4. Choose the Warehouse and table.
5. Select Refresh after adding it.^3^

3. If it is Direct Lake on SQL

Direct Lake on SQL supports one Fabric source and does not support composite models. Therefore:

• If the new table belongs to the model’s existing Warehouse, use Fabric Open data model → Editing → Edit tables, select the table, and refresh.
• If it belongs to another Warehouse or Fabric item, it cannot be added to that Direct Lake on SQL model as another Direct Lake source.
• Do not use Get data to add an Import table to this model; combining Direct Lake and Import tables requires Direct Lake on OneLake.^4^

For a table from another source, either create a separate compatible semantic model or migrate to Direct Lake on OneLake. Microsoft documents a TMDL migration, but it is not recommended when the current model uses SQL analytics endpoint views.^2^

4. Correct the identity or source binding

The exact message “Direct Lake query is missing workspace or artifact identity” is not documented as a named parameterization limitation. It points more plausibly—but not conclusively—to missing source context or an invalid identity/connection binding.

Have the relevant administrator or owner check:

1. Semantic-model owner: Confirm the source Warehouse still exists in the expected workspace and was not moved or replaced.
2. Workspace administrator: Confirm both the model owner and editing user have access to the Warehouse. Direct Lake requires the semantic-model owner to retain read access for framing.
3. Semantic-model or connection owner: Open the semantic model’s Settings, review its cloud connection, and determine whether it uses:
   • Microsoft Entra ID SSO, where the current user needs source access; or
   • A fixed cloud-connection identity, where that identity needs source access.
4. For Direct Lake on OneLake without OneLake security, grant the effective identity Read and ReadAll on the source item. For Direct Lake on SQL, grant source-item Read plus SELECT permission on the relevant tables.^1^
5. Return to the model and select Refresh.

Also verify the table is not already represented under another model-table name: adding the same source table more than once is unsupported in Desktop and web modeling.^4^

Since removing parameterization previously worked but no longer does, repeating that workaround is unlikely to distinguish the current failure. If both Desktop and web modeling still return the same identity error after the source and permissions are corrected, open a Microsoft Fabric support request and include the screenshot’s Activity ID, UTC timestamp, workspace region, semantic-model ID, Warehouse ID, and whether the expression is AzureStorage.DataLake or Sql.Database.

References

1. Integrate Direct Lake Security - Microsoft Fabric
2. Direct Lake in Power BI Desktop - learn.microsoft.com
3. Direct Lake in web modeling
4. Direct Lake overview