Elementary

Elementary is a dbt package that layers automated test logging, anomaly detection, and schema-change tracking on top of the dbt project. It runs entirely inside dbt — no separate service to deploy — and writes its results into a dedicated ${ENV}__elementary schema in ClickHouse.

What it gives you

A persistent log of every dbt test run (dbt_test_results, elementary_test_results)
Anomaly tests that flag deviations the static dbt tests would miss
Schema-change detection that flags column adds, drops, and type changes
A standalone HTML report (edr report) for review outside the dbt CLI

Project wiring

Field	Value
Package	`elementary-data/elementary` (in `dbt/dependencies.yml`)
Schema	`${ENV}__elementary` (e.g. `dev__elementary`, `main__elementary`)
Project-specific views	`dbt/models/elementary_extensions/` (`daily_test_summary`, `run_results_by_date`, `test_results_by_date`)

Initialize the schema

After bumping the package or recreating an environment, build the elementary infrastructure tables:

uv run dbt run --select elementary

Elementary intentionally ignores --full-refresh; to recreate its tables from scratch, set the Elementary-owned variable instead:

uv run dbt run --select elementary --vars "elementary_full_refresh: true"

Enable tests on a model

Add the Elementary tests in the model’s properties.yml alongside the standard dbt tests:

models:
  - name: matches
    config:
      elementary:
        timestamp_column: event_date
    tests:
      - elementary.volume_anomalies:
          timestamp_column: event_date
      - elementary.schema_changes

Available test types:

Test	Detects
`volume_anomalies`	Row-count drops or spikes vs. historical volume
`freshness_anomalies`	Tables that haven’t been written to recently
`column_anomalies`	Per-column null %, distinct count, min/max drift
`dimension_anomalies`	Distribution shifts within categorical columns
`schema_changes`	Column additions, removals, type changes

Where results land

Test runs write to the ${ENV}__elementary schema:

dbt_test_results — every dbt test execution
elementary_test_results — Elementary anomaly tests specifically
alerts_* views — failed-test summaries grouped for downstream alerting

The project-specific views in dbt/models/elementary_extensions/ (daily_test_summary, run_results_by_date, test_results_by_date) sit on top of these and are the easiest entry points for ad-hoc queries.

Standalone report

Outside the dbt CLI, the edr tool generates an HTML report:

pip install 'elementary-data[clickhouse]'
edr report --profile elementary

The elementary dbt profile lives alongside the main clickhouse profile in the same profiles.yml.

Integrations

dbt: Elementary runs as a dbt package — every dbt build populates its tables
ClickHouse: the only target — Elementary writes to ${ENV}__elementary
Airflow: since Elementary tests run as part of dbt build, the daily Airflow runs continuously refresh the test history