Skip to content
CSA Loom — the Microsoft Fabric experience for Azure tenants where Fabric isn't yet available: lakehouses, warehouses, notebooks, semantic models, Activator rules, Data Agents, across Commercial, GCC, GCC-High, and DoD IL5

Tutorial 03 — Direct Lake parity (warm-cache materializer)

Author a TMDL semantic model over the Silver table from Tutorial 02, configure the Direct-Lake-Shim refresh policy, build a Power BI report. 45 minutes.

Prerequisites

  • Workspace + noaa_silver_daily Silver table from Tutorial 02
  • Power BI Desktop installed (or Loom Console Semantic Model designer if v1.1+)
  • Power BI Premium F-SKU (Gov-H / IL5) or P-SKU (GCC) workspace attached to your Loom workspace

Steps

1. Author the TMDL semantic model

Open Power BI Desktop → File → New → Composite model.

Add connection: - Type: Synapse Serverless SQL (Gov) or Databricks SQL Warehouse (Commercial) - Endpoint: from Loom Console "Warehouse" pane (Copy connection string) - Database: <workspace>_loom_silver - Tables: noaa_silver_daily

Save as noaa-semantic-model.pbip (project format, TMDL underneath).

2. Build a basic model

In Power BI Desktop Modeling view: - Mark date as Date column - Create a Date dimension table (auto-generated from Modeling → New Table) - Add measures: 

Avg Temperature = AVERAGE(noaa_silver_daily[temperature_c])
Days Recorded = COUNT(noaa_silver_daily[date])

Commit .pbip files to Git (docs/fiab/tutorials/_examples/03-tmdl/).

3. Configure Direct-Lake-Shim refresh policy

Loom Console → Semantic Models (v1.1) OR CLI in v1:

# v1 CLI
loom-dl-shim configure \
  --semantic-model noaa-semantic-model \
  --table noaa_silver_daily \
  --refresh-policy partition \
  --partition-column date

This tells the Direct-Lake-Shim service to: - Subscribe to Event Grid notifications on the noaa_silver_daily table's _delta_log/*.json - On each commit, identify the affected partition (date) - Issue TOM partition refresh against the Power BI semantic model

4. Deploy to Power BI Premium workspace

loom-semantic-model deploy \
  --pbip noaa-semantic-model.pbip \
  --workspace-id <power-bi-workspace-id>

Or use Power BI service "Get data → Power BI Project (.pbip)".

5. Build a Power BI report

In Power BI Desktop, switch to Report view: - Add a line chart: X-axis Date, Y-axis Avg Temperature - Add a card: Days Recorded - Save + publish to the Loom Power BI workspace

6. Test the freshness

Back in the Loom notebook:

# Add a new partition to Silver
new_data = spark.createDataFrame([
  ("2026-05-15", "ABC123", "WeatherStation X", 75.0),
], schema=["date_str", "station_id", "station_name", "temp_f"])

silver_new = (
  new_data
  .withColumn("date", to_date("date_str"))
  .withColumn("temperature_c", (col("temp_f") - 32) * 5/9)
)

silver_new.write.format("delta") \
  .mode("append") \
  .saveAsTable("noaa_silver_daily")

Run. Note the time.

7. Verify the semantic model refreshes

In the Power BI report, refresh the dataset. The Direct-Lake-Shim should have triggered a partition refresh within 5-30 seconds of your write.

Check refresh time: - Power BI workspace → Dataset settings → Refresh history

You should see a partition refresh for the 2026-05-15 partition within seconds of the Spark write.

8. Verify in Loom Console

Console Monitoring → Direct-Lake-Shim: - Latency p50, p95 for the recent refresh - Refresh history

Latency target: 5-30 s (per Direct Lake parity).

The honest gap

This is not Fabric's native Direct Lake sub-second freshness. You're seeing 5-30 s instead. Document this in your customer expectations. When Fabric Gov GA arrives, re-author the model with Direct Lake on OneLake storage mode → sub-second freshness.

What's next

Cleanup

  • Delete the Power BI report + dataset
  • Run loom-dl-shim unconfigure --semantic-model noaa-semantic-model
  • Drop the Silver table from notebook