Matthieu Caneill - dbt-score: a linter for your dbt model metadata | PyData Amsterdam 2024

dbt-score is an open-source linting tool for dbt model metadata, developed at Picnic and MIT licensed

The tool helps enforce consistency and quality checks across large dbt projects with hundreds or thousands of data models

Key features include:

• Linting dbt YAML metadata files
• Configurable severity levels for rules
• Custom rule creation using Python
• Score-based evaluation of models
• Integration with dbt documentation/catalogs
• CI/CD friendly with machine-readable output

Main use cases:

• Enforcing documentation standards
• Checking for required properties (owners, descriptions)
• Validating data quality tests exist
• Ensuring security and access controls
• Custom business rule validation

Benefits:

• No database connection required (metadata-only checks)
• Easy to integrate into existing Python/dbt workflows
• Extensible through custom rules
• Helps maintain consistency across large teams
• Prevents bad models from reaching production

Installation and usage:

• Installs as a companion library to dbt
• Simple CLI interface (dbt-score lint)
• Configurable via pyproject.toml
• Can be run anywhere Python/dbt runs

Customization options:

• Rule skipping for specific models
• Severity level configuration
• Custom rule namespaces
• Project-specific thresholds
• Custom scoring weights