dbt-cost 0.2.0

Estimate BigQuery costs for dbt models before execution

dbt-cost

Estimate BigQuery costs for dbt models before you run them.

dbt-cost uses BigQuery's free dry-run API to predict bytes scanned for each model, converts that to dollars, and gives you a clear cost breakdown — in your terminal, as JSON for CI, or as a markdown comment on your PR.

Install

pip install dbt-cost

Requires Python 3.9+ and a compiled dbt manifest (dbt compile).

Quick Start

# Cost of a single model
dbt-cost estimate dim_customers

# Cost of a model + all downstream
dbt-cost estimate stg_orders+

# Full project cost report
dbt-cost report

# Compare costs between two branches (for PRs)
dbt-cost diff --base-manifest base_manifest.json --pr-manifest pr_manifest.json

Commands

estimate

Estimate the cost of specific models using dbt selector syntax.

# Single model
dbt-cost estimate dim_customers

# Model + downstream
dbt-cost estimate stg_orders+

# Model + 2 levels downstream
dbt-cost estimate stg_orders+2

# Upstream + model + downstream
dbt-cost estimate +dim_customers+

# By tag
dbt-cost estimate --selector tag:marketing

# Top 5 most expensive, JSON output
dbt-cost --format json estimate stg_orders+ --top 5

Options:

Flag	Default	Description
--price	6.25	Price per TB in USD (BigQuery on-demand)
--concurrency	10	Number of concurrent dry-runs
--top	all	Show only the N most expensive models

report

Full project cost breakdown by layer.

dbt-cost report
dbt-cost report --top 20
dbt-cost report --layer staging
dbt-cost --format json report

Options:

Flag	Default	Description
--top	10	Show top N most expensive models
--layer	all	Filter to a specific layer
--price	6.25	Price per TB in USD
--concurrency	10	Concurrent dry-runs (max 50)

diff

Compare costs between two manifests — designed for CI pipelines and PR reviews.

# Basic diff
dbt-cost diff --base-manifest target/base.json --pr-manifest target/pr.json

# Markdown output for PR comments
dbt-cost --format markdown diff --base-manifest base.json --pr-manifest pr.json

# Fail CI if cost increase exceeds $50
dbt-cost diff --base-manifest base.json --pr-manifest pr.json --threshold 50

Example markdown output:

dbt-cost: Cost Impact of this PR

| Model | Before | After | Change |
|-------|--------|-------|--------|
| dim_customers | $0.80 | $45.00 | +$44.20 (+5,525%) |
| stg_orders | $2.10 | $2.10 | no change |

Total: $2.90 -> $47.10 (+$44.20)

2 models changed, 0 added, 0 removed

Options:

Flag	Default	Description
--base-manifest	required	Path to base branch manifest.json
--pr-manifest	required	Path to PR branch manifest.json
--threshold	none	Max allowed cost increase in USD; exit 1 if exceeded
--include-unchanged	false	Include unchanged models in output
--price	6.25	Price per TB in USD
--concurrency	10	Concurrent dry-runs

Global Options

These apply to all commands:

Flag	Default	Description
--manifest	target/manifest.json	Path to dbt manifest
--credentials	ADC	Path to GCP service account key JSON
--format	table	Output format: table, json, or markdown

GitHub Actions

Add automated cost checks to your PRs. Copy examples/dbt-cost-pr.yml to your project's .github/workflows/ directory.

The workflow:

1. Compiles dbt on both the base and PR branches
2. Runs dbt-cost diff to compare costs
3. Posts a sticky PR comment with the cost impact
4. Fails the check if cost increase exceeds your threshold

Authentication

dbt-cost needs read-only BigQuery access to run dry-runs. It supports:

1. Application Default Credentials (recommended) — run gcloud auth application-default login
2. Service account key — pass --credentials /path/to/key.json

Required IAM permissions: bigquery.jobs.create (BigQuery Job User role) and read access to referenced tables.

Safety: dry-run mode is hardcoded — dbt-cost can never execute queries.

How It Works

manifest.json (from dbt compile)
       |
   dbt-lens parses the manifest + dependency graph
       |
   Selector resolves which models to estimate
       |
   BigQuery dry-run API returns bytes_processed (free, no execution)
       |
   bytes / 1 TB * $6.25 = estimated cost
       |
   Table, JSON, or Markdown output

Development

git clone https://github.com/ishanyash/dbt-cost.git
cd dbt-cost
pip install -e ".[dev]"

# Run checks
ruff check src/ tests/
mypy src/dbt_cost/
pytest tests/ -v

License

MIT