modaryn 0.1.1

modaryn analyzes dbt projects to score model complexity and structural importance, helping teams identify high-risk and high-impact data models.

modaryn

modaryn analyzes dbt projects to score model complexity and structural importance, helping teams identify high-risk and high-impact data models.

Overview

modaryn is a Python-based CLI tool that analyzes dbt projects and scores each model based on three pillars:

• Complexity — SQL metrics (JOINs, CTEs, conditionals, WHERE clauses, character count)
• Importance — Structural metrics (downstream model/column counts)
• Quality — Test coverage metrics (test count, column coverage %)

Final score: raw_score = complexity_score + importance_score - quality_score (higher = riskier)

The SQL dialect is auto-detected from manifest.json. Column-level lineage is traced via sqlglot to compute downstream column impact.

Installation

pip install modaryn

Usage

score command

Analyzes and scores all dbt models, displaying a combined scan and score report.

modaryn score --project-path . --apply-zscore --format html --output report.html

Option	Short	Description	Default
--project-path	-p	Path to the dbt project directory	.
--dialect	-d	SQL dialect (bigquery, snowflake, duckdb, etc.). Auto-detected from manifest.json if omitted.	auto
--config	-c	Path to a custom weights YAML file	None
--apply-zscore	-z	Apply Z-score normalization to scores	False
--format	-f	Output format: terminal, markdown, html	terminal
--output	-o	Path to write the output file	None
--select	-s	Filter models by selector (repeatable, OR logic)	None
--verbose	-v	Show detailed warnings (missing SQL, skipped columns)	False

--select selector syntax:

# Model name glob
modaryn score --project-path . --select "fct_*"

# Path prefix
modaryn score --project-path . --select path:marts/finance

# dbt tag
modaryn score --project-path . --select tag:daily

# Multiple selectors (OR logic)
modaryn score --project-path . --select path:marts/customer --select path:marts/finance

---

ci-check command

Checks model scores against a threshold for use in CI/CD pipelines. Exits with code 1 if any model exceeds the threshold, 0 otherwise.

modaryn ci-check --project-path . --threshold 20.0 --apply-zscore

Option	Short	Description	Default
--project-path	-p	Path to the dbt project directory	.
--threshold	-t	Maximum allowed score (required)	—
--dialect	-d	SQL dialect. Auto-detected if omitted.	auto
--config	-c	Path to a custom weights YAML file	None
--apply-zscore	-z	Check against Z-scores instead of raw scores	False
--format	-f	Output format: terminal, markdown, html	terminal
--output	-o	Path to write the output file	None
--select	-s	Filter models by selector (repeatable, OR logic)	None
--verbose	-v	Show detailed warnings	False

---

impact command

Traces all downstream columns affected by a change to a specific column (BFS column-level impact analysis).

modaryn impact --project-path . --model fct_orders --column order_id

Option	Short	Description	Default
--project-path	-p	Path to the dbt project directory	.
--model	-m	Model name to trace impact from (required)	—
--column	-c	Column name to trace impact from (required)	—
--dialect	-d	SQL dialect. Auto-detected if omitted.	auto
--select	-s	Filter models by selector (restricts lineage scope)	None
--verbose	-v	Show detailed warnings	False

---

Missing compiled SQL (N/A columns)

Complexity metrics require compiled SQL from target/compiled/. If dbt compile has not been run or a model failed to compile, those columns will show N/A in the report. A warning summary is printed at the end of the output. Use --verbose to see the full list of affected models.

⚠ 3 model(s) show N/A for complexity columns because compiled SQL was not found.
Run `dbt compile` to enable full analysis: model_a, model_b, model_c

---

Report Columns and Calculation Logic

1. SQL Complexity Metrics

Metric	Calculation	Example
JOINs	Count of all JOIN clauses	JOIN, LEFT JOIN, CROSS JOIN each count as 1
CTEs	Count of all CTEs defined	WITH a AS (...), b AS (...) = 2
Conditionals	Count of IF expressions (each WHEN branch in a CASE)	A CASE WHEN ... WHEN ... END with 2 branches = 2
WHEREs	Count of WHERE clauses including subqueries	Main WHERE + subquery WHERE = 2
SQL Chars	Total character count of the compiled SQL	—

2. Structural Importance Metrics

Metric	Calculation	Example
Downstream	Number of dbt models that directly reference this model	Models B and C use A → A has 2
Col. Down	Total count of downstream column references	B's col1 and col2 both reference A's id → 2

3. Quality Metrics

Metric	Calculation	Example
Tests	Total dbt tests attached to the model	4 column tests → 4
Coverage (%)	% of columns with at least one test	8 of 10 columns tested → 80%

---

Scoring Formula

1. Complexity Score = (JOINs × w1) + (CTEs × w2) + (Conditionals × w3) + (WHEREs × w4) + (Chars × w5)
2. Importance Score = (Downstream Models × w6) + (Col. Down × w7)
3. Quality Score = (Tests × w8) + (Coverage % × w9)

Raw Score = Complexity Score + Importance Score − Quality Score (minimum 0)

Z-Score Normalization

When --apply-zscore is used: Z-Score = (Raw Score − Mean) / Standard Deviation

---

Custom Weights Configuration

Override default weights by passing a YAML file via --config:

sql_complexity:
  join_count: 2.0
  cte_count: 1.5
  conditional_count: 1.0
  where_count: 0.5
  sql_char_count: 0.01

importance:
  downstream_model_count: 1.0

quality:
  test_count: 0.5
  column_coverage: 1.0

Unknown sections or keys are reported as warnings at runtime.

---