Pro BPA Scanner

BPA scanning for Fabric semantic models in Pro workspaces — no Premium capacity or XMLA endpoints required.

Combines Fabric CLI (REST API extraction) with Tabular Editor 2 (BPA rule engine) to analyze semantic models headlessly, enabling CI/CD automation for any Fabric workspace.

Why This Tool?

Approach	XMLA Required	Premium Required	CI/CD Ready
Tabular Editor direct	Yes	Yes	Partial
Fabric Notebooks (semantic-link-labs)	No	Yes	Partial
This tool	No	No	Yes

Output

Quick Start

# Scan all models in a workspace (.Workspace suffix optional)
./scan-workspaces.ps1 -Workspaces "YourWorkspace"

# Scan multiple workspaces
./scan-workspaces.ps1 -Workspaces "Production", "Development"

# Scan specific model
./scan-workspaces.ps1 -Workspaces "YourWorkspace" -Model "YourModel.SemanticModel"

# Filter by pattern
./scan-workspaces.ps1 -Workspaces "YourWorkspace" -Filter "Sales.*"

# With custom rules
./scan-workspaces.ps1 -Workspaces "YourWorkspace" -CustomRulesFile "./my-rules.json"

How It Works

See docs/architecture.md for detailed component documentation.

Parameters

Parameter	Required	Description
-Workspaces	Yes	One or more workspace names (.Workspace suffix optional)
-Model	No	Specific model to scan (.SemanticModel suffix required)
-Filter	No	Regex pattern to filter models
-RulesFile	No	Path to BPA rules JSON (default: ./BPARules-PowerBI.json)
-CustomRulesFile	No	Path to custom rules JSON (merged with default rules)
-OutputDir	No	Output directory for reports (default: ./output)

View Results

# Python dashboard (recommended)
python bpa-dashboard.py

# Or process JSON directly
cat output/bpa-analysis-*.json | jq '.violations'

Prerequisites

See PREREQUISITES.md for installation requirements:

• Fabric CLI (fab) - https://aka.ms/fabric-cli
• Tabular Editor 2 - https://github.com/TabularEditor/TabularEditor/releases
• PowerShell 5.1+
• Python 3.8+ with rich library (for dashboard only)

Output Format

{
  "metadata": {
    "timestamp": "2025-12-15T12:00:00Z",
    "model": "Sales Model",
    "workspace": "Production"
  },
  "modelStructure": {
    "tables": 10,
    "measures": 25,
    "relationships": 8
  },
  "violations": [
    {
      "rule": "DAX_DIVISION_COLUMNS",
      "object": "Sales.Revenue Ratio",
      "severity": 3,
      "message": "Avoid division (use DIVIDE function instead)"
    }
  ]
}

Severity Levels

Level	Name	Action
3	Error	Fix before deployment
2	Warning	Consider fixing
1	Info	Optional improvement

CI/CD Integration

See .github/workflows/bpa-scan.yml for a GitHub Actions template.

Required secrets:

• FABRIC_CLIENT_ID - Service principal app ID
• FABRIC_CLIENT_SECRET - Service principal secret
• FABRIC_TENANT_ID - Azure AD tenant ID

Project Structure

pro-bpa-scanner/
├── scan-workspaces.ps1              # Main scanner entry point
├── lib/                      # PowerShell modules
│   ├── Discover-Models.ps1   # Workspace model discovery
│   ├── Extract-Model.ps1     # TMDL extraction via Fabric CLI
│   ├── Analyze-Model.ps1     # BPA analysis via Tabular Editor 2
│   └── Format-Report.ps1     # JSON report generation
├── bpa_dashboard/            # Python dashboard package
├── bpa-dashboard.py          # Dashboard entry point
├── BPARules-PowerBI.json     # BPA rule definitions
├── output/                   # JSON analysis reports (generated)
├── temp/                     # TMDL extraction (temporary, auto-cleaned)
└── PREREQUISITES.md          # Installation guide

Known Limitations

Calculated Table Column Detection

When comparing scanner results with Tabular Editor 3 GUI connected directly to a model, the scanner may report fewer violations. This is due to a limitation in Tabular Editor 2 CLI mode when analyzing TMDL exports.

Affected rules (violations on Calculated Table Columns may be missed):

• APPLY_FORMAT_STRING_COLUMNS - Provide format string for visible numeric columns
• META_AVOID_FLOAT - Do not use floating point data types
• META_SUMMARIZE_NONE - Don't summarize numeric columns

Root cause: Tabular Editor 2 CLI analyzing TMDL exports does not evaluate certain metadata rules against Calculated Table Columns the same way TE3 GUI does when connected to a live model.

Impact: Scanner results are directionally accurate but may undercount by ~5-15% for models with many Calculated Table Columns. Critical violations (DAX issues, unused columns, relationship naming) are fully detected.

Acknowledgments

• Kurt Buhler - fabric-cli reference documentation used in developing this solution
• Kurt Buhler - claude-goblin TUI design that inspired the dashboard
• Daniel Otykier - Tabular Editor and Best Practice Analyzer
• TabularEditor/BestPracticeRules - Official BPA rule definitions

License

MIT