power-bi

name	power-bi
description	Power BI Development. Use when: building reports, TMDL semantic models, DAX patterns, PBIR report definitions, Fabric REST API, Direct Lake, incremental refresh, Dataverse as data source, Power BI Embedded in Code Apps, CI/CD for Power BI.

Skill: Power BI Development

When to Use

Trigger when building Power BI semantic models, reports, DAX measures, or managing Power BI via Fabric REST APIs. Covers TMDL/model.bim semantic model definitions, PBIR report definitions, DAX patterns, and deployment automation.

---

Architecture Overview

Power BI Project Structure (PBIP Format)

MyReport.pbip                    # Project file
├── MyReport.SemanticModel/
│   ├── definition/
│   │   ├── model.tmdl           # TMDL format (modern, human-readable)
│   │   ├── tables/
│   │   │   ├── Sales.tmdl       # Table definitions with columns, measures
│   │   │   ├── Date.tmdl
│   │   │   └── Product.tmdl
│   │   ├── relationships.tmdl   # All relationships
│   │   ├── expressions.tmdl     # M/Power Query expressions
│   │   ├── cultures/
│   │   │   └── en-US.tmdl       # Translations
│   │   └── roles/
│   │       └── RegionRLS.tmdl   # Row-level security
│   ├── definition.pbism         # Semantic model metadata
│   └── .platform
│
├── MyReport.Report/
│   ├── definition/
│   │   ├── report.json          # Report definition
│   │   ├── pages/
│   │   │   ├── page1/
│   │   │   │   ├── page.json    # Page config
│   │   │   │   └── visuals/
│   │   │   │       ├── visual1.json
│   │   │   │       └── visual2.json
│   │   ├── bookmarks/           # Report bookmarks
│   │   └── theme.json           # Report theme
│   ├── definition.pbir          # Report metadata
│   └── .platform

---

⚠️ REQUIRED: Load Sub-Files Before Implementation

SKILL.md is a summary only — it is NOT sufficient for implementation.

The detailed content (complete payloads, XML templates, working examples, edge-case handling) lives in sub-files in the same directory as this SKILL.md. Before writing any code, you MUST use read_file on the sub-files relevant to your task:

• TMDL — table definitions, columns, measures, partitions, Date table pattern, relationships, row-level security
• DAX Patterns — time intelligence, ranking, KPIs, field parameters, CALCULATE context patterns
• PBIR Report Definition — page.json, visual.json, theme definition
• Fabric REST API — authentication, workspace management, dataset operations, report operations, deployment pipeline
• Direct Lake & Incremental Refresh — Direct Lake mode, storage mode selection, incremental refresh setup
• Dataverse as Data Source — Power Query M connection, TDS endpoint SQL access
• Power BI Embedded — embed token generation, React integration for Code Apps
• CI/CD — GitHub Actions pipeline for automated deployment

---

Best Practices

Semantic Model

• Always create a dedicated Date table (don't rely on auto date/time)
• Use star schema: fact tables at center, dimension tables around
• Hide foreign key columns from report view
• Set summarizeBy: none on non-additive columns (IDs, codes)
• Use folders to organize measures by business domain
• Format strings on all measures (currency, %, number)

DAX

• Use variables (VAR ... RETURN) for readability and performance
• Use DIVIDE() instead of / (handles division by zero)
• Use BLANK() not 0 for missing values
• Avoid FILTER(table, ...) when column filter works — use KEEPFILTERS or VALUES
• Never use CALCULATE with row context unless you understand context transition

Reports

• Maximum 8-10 visuals per page
• Use bookmarks for showing/hiding visual states instead of multiple pages
• Apply consistent theme across all reports
• Use drill-through for detail pages, not separate reports
• Test with Row-Level Security using "View As" in Power BI Service

---

Anti-Patterns

• Auto date/time enabled (creates hidden tables, bloats model)
• Bidirectional cross-filtering (causes ambiguity, performance issues)
• Too many measures without folders (unnavigable model)
• CALCULATE(COUNTROWS(...)) when COUNTROWS alone works
• Import mode on large Dataverse tables without incremental refresh
• Publishing .pbix files directly instead of using deployment pipelines
• Hardcoding server/database names instead of using parameters
• Missing RLS when multiple business units access the same report
• Calculated columns for values that should be measures
• Not testing refresh after deployment (connection strings, credentials)

Note: Power BI Dashboards and Datasets are NOT supported in Power Platform Pipelines. Use Power BI's own deployment pipelines (Fabric REST API) for Power BI content promotion.

---

Related Skills

• dataverse — Dataverse as a Power BI data source
• code-apps — Embedding Power BI in Code Apps