Direct Lake Operations

This skill covers working with Direct Lake semantic models in Semantic Link Labs.

When to Use This Skill

Use this skill when you need to:

• Migrate models to Direct Lake
• Check and fix Direct Lake fallback issues
• Sync schema between lakehouse and model
• Work with Direct Lake guardrails
• Warm Direct Lake cache

Overview

Direct Lake is a storage mode for Power BI semantic models that reads data directly from Delta tables in OneLake, providing fast query performance without data import.

Key Files

Direct Lake Functions

Core Functions

Schema Management

Guardrails

Cache Management

Connection Updates

Checking Fallback Reasons

Direct Lake models can fall back to DirectQuery mode. Use checkfallbackreason to diagnose:

from sempy_labs.directlake import check_fallback_reason

# Check specific model
result = check_fallback_reason(
    dataset="My Direct Lake Model",
    workspace="My Workspace"
)

# Returns DataFrame with fallback reasons per table
print(result)

Common Fallback Reasons

Schema Synchronization

Compare Schemas

from sempy_labs.directlake import direct_lake_schema_compare

# Compare model schema with lakehouse tables
comparison = direct_lake_schema_compare(
    dataset="My Direct Lake Model",
    workspace="My Workspace"
)

# Shows differences between model and lakehouse
print(comparison)

Sync Schema

from sempy_labs.directlake import direct_lake_schema_sync

# Sync model schema from lakehouse
direct_lake_schema_sync(
    dataset="My Direct Lake Model",
    workspace="My Workspace",
    add_columns=True,      # Add new columns
    remove_columns=False,  # Keep columns not in lakehouse
)

Guardrails

Direct Lake has limits based on capacity SKU:

from sempy_labs.directlake import get_direct_lake_guardrails

# Get guardrails for a model
guardrails = get_direct_lake_guardrails(
    dataset="My Direct Lake Model",
    workspace="My Workspace"
)

# Shows limits vs current values
print(guardrails)

Guardrail Limits

Generating Direct Lake Models

Create a new Direct Lake model from lakehouse tables:

from sempy_labs.directlake import generate_direct_lake_semantic_model

# Generate model from lakehouse
generate_direct_lake_semantic_model(
    dataset="New Direct Lake Model",
    lakehouse="My Lakehouse",
    workspace="My Workspace",
    lakehouse_workspace="Lakehouse Workspace",  # Optional
)

Updating Connections

Change the lakehouse a Direct Lake model points to:

from sempy_labs.directlake import update_direct_lake_model_connection

# Update to different lakehouse
update_direct_lake_model_connection(
    dataset="My Direct Lake Model",
    workspace="My Workspace",
    target_lakehouse="New Lakehouse",
    target_workspace="Target Workspace",
)

Cache Warming

Warm the Direct Lake cache after model refresh:

Using IsResident

from sempy_labs.directlake import warm_direct_lake_cache_isresident

# Warm columns that were previously in memory
warm_direct_lake_cache_isresident(
    dataset="My Direct Lake Model",
    workspace="My Workspace"
)

Using Perspective

from sempy_labs.directlake import warm_direct_lake_cache_perspective

# Warm columns defined in a perspective
warm_direct_lake_cache_perspective(
    dataset="My Direct Lake Model",
    perspective="Cache Warming",
    workspace="My Workspace"
)

Migration to Direct Lake

The library provides migration tools in src/sempy_labs/migration/:

Migration Workflow

1. Migrate calculated tables to lakehouse as Delta tables
2. Create Direct Lake model with same structure
3. Migrate measures and other objects
4. Validate and test

Example Migration

from sempy_labs.migration import (
    migrate_calctables_to_lakehouse,
    migrate_model_objects_to_semantic_model,
)

# Step 1: Migrate calculated tables to lakehouse
migrate_calctables_to_lakehouse(
    dataset="Source Import Model",
    workspace="My Workspace",
    lakehouse="Target Lakehouse",
    lakehouse_workspace="Lakehouse Workspace",
)

# Step 2: Migrate other objects (measures, etc.)
migrate_model_objects_to_semantic_model(
    source_dataset="Source Import Model",
    target_dataset="Target Direct Lake Model",
    source_workspace="My Workspace",
    target_workspace="My Workspace",
)

Best Practices

Do's

• ✅ Always check guardrails before deploying
• ✅ Use schema sync after lakehouse changes
• ✅ Warm cache after scheduled refreshes
• ✅ Monitor fallback reasons regularly

Don'ts

• ❌ Don't ignore fallback warnings
• ❌ Don't exceed guardrail limits
• ❌ Don't modify lakehouse schema without syncing model
• ❌ Don't use unsupported DAX patterns

Troubleshooting

Model Falls Back to DirectQuery

1. Run checkfallbackreason() to identify cause
2. Check column mappings with directlakeschema_compare()
3. Verify data types match between model and lakehouse
4. Check guardrails with getdirectlake_guardrails()

Schema Mismatch

1. Run directlakeschema_compare() to see differences
2. Run directlakeschema_sync() to fix mismatches
3. Verify partition entity mappings

Performance Issues

1. Check cache state with warmdirectlakecacheisresident()
2. Verify partitions are properly configured
3. Review query patterns for unsupported filters

Related Resources