dbt-score (GitHub Repo)

dbt-score is a linter that scores data transformation models on metadata quality to enforce documentation, testing, and governance standards across growing dbt projects.

What: A linting tool for dbt projects that evaluates models against configurable rules for documentation completeness, test coverage, ownership metadata, naming conventions, and SQL complexity, assigning numerical scores (0-10) to individual models and entire projects.

Why it matters: As dbt projects scale to hundreds or thousands of models, inconsistent documentation and missing tests become major problems; this tool provides automated quality checks that can fail CI/CD builds when standards slip, preventing technical debt accumulation.

Takeaway: Install via pip in your dbt-core environment and run `dbt-score lint` from your project root to get immediate quality feedback, or integrate into CI pipelines with fail thresholds.

Deep dive

Provides a 0-10 scoring system for both individual dbt models and overall project health, with configurable badge thresholds (gold/silver/bronze icons)
Ships with built-in rules covering common needs: column descriptions, model documentation, ownership metadata, test presence, and SQL line count limits
Supports custom rules written as simple Python functions using decorators, enabling organization-specific governance requirements (e.g., requiring business owner metadata)
Configuration via pyproject.toml allows setting CI fail thresholds, disabling specific rules, adjusting rule severity levels, and customizing rule parameters
Integrates with dbt's native selection syntax for targeted linting (e.g., --select staging.* for staging models only, --select state:modified for changed models)
Can auto-generate the dbt manifest via --run-dbt-parse flag, eliminating the need for separate dbt parse step
Exit codes (0 for pass, 1 for fail) make CI/CD integration straightforward across different platforms
Warnings include severity levels (low/medium/high) and specific guidance on what's missing (e.g., which columns lack descriptions)
Addresses real scaling problems: inconsistent documentation, missing tests on critical models, naming convention violations, and compliance metadata gaps
Requires Python 3.10+ and dbt-core 1.5+, installed in the same environment as dbt

Decoder

dbt (Data Build Tool): An open-source framework for transforming data in warehouses using SQL SELECT statements, organizing data pipelines as versioned models
Linter: A static analysis tool that checks code or metadata against style rules and best practices without executing it
Manifest: A JSON file generated by dbt containing metadata about all models, tests, and dependencies in a project

Original article

dbt-score is a linter for dbt metadata quality. It scores models and projects against rules for docs, tests, ownership, naming, and SQL complexity, so teams can enforce standards in CI/CD and catch weak models early. It supports custom rules for org-specific governance.