Unable to update schema in semantic model

AI response:

What you’re seeing is by design, but it’s poorly documented.

I’ll break it down cleanly and then tell you exactly how to add those columns.

The core reason (TL;DR)

You deployed a metadata-only / external semantic model into Fabric.

Once a semantic model is:

• Compatibility level 1600
• Data Source Version V3
• Deployed via Tabular Editor / XMLA

👉 Power BI Desktop no longer owns the model schema
👉 Power Query does not exist
👉 “Refresh > Schema” is permanently disabled

So:

• ❌ Desktop cannot refresh schema
• ❌ Service UI cannot refresh schema
• ❌ There is no “Get Data / Transform Data” layer anymore

This is expected behavior for XMLA-authored models.

Why your Tabular Editor change didn’t surface the columns

This is the subtle but critical part:

Editing the SQL text in a partition does NOT update column metadata

In tabular models:

• Columns are metadata objects
• Partitions only define how data is loaded
• Adding columns to SQL ≠ adding columns to the model

So when you:

updated the SQL to include two new columns

The engine happily loads them internally, but:

• The model does not know those columns exist
• They are not materialized as Table.Columns[]
• Power BI will never show them

Why “Refresh Metadata” doesn’t exist

In SSAS / Fabric semantic models:

• There is no automatic schema discovery
• Metadata refresh is not supported
• Schema changes must be explicit

This is different from:

• Import-mode PBIX
• Power Query–based datasets

The correct way to add columns (this WILL work)

You must explicitly add the columns in Tabular Editor.

Option A — Add columns manually (recommended & safest)

1. Open the model in Tabular Editor
2. Expand the table
3. Right-click Columns → Create New → Column
4. Set:
• Name
• Data Type
• Source Column (must match SQL column name exactly)
5. Save
6. Deploy back to the Fabric workspace
7. Run a data refresh

✅ Columns will appear immediately

Option B — Recreate the table metadata (faster but riskier)

Only do this if the table is simple.

1. Script out:
• Measures
• Relationships
• Sort-by columns
2. Delete the table
3. Recreate the table from scratch with updated SQL
4. Reattach measures/relationships
5. Deploy

Easy to break dependencies — use carefully.