// the find
elementary-data/elementary
The dbt-native data observability solution for data & analytics engineers. Monitor your data pipelines in minutes. Available as self-hosted or cloud service with premium features.
Elementary is a dbt-native data observability tool that plugs into your existing dbt project to collect test results, model run metadata, and data quality metrics, then surfaces them via a generated HTML report or Slack/Teams alerts. It's aimed at analytics engineers and data teams already using dbt who want pipeline monitoring without standing up separate infrastructure. The OSS version is a CLI + dbt package combo; the more capable features (ML-based anomaly detection, column-level lineage, BI integration) require their paid cloud product.
- Genuinely dbt-native: tests are defined as standard dbt tests in YAML, configuration lives in your dbt project, and it reads from dbt artifacts directly—no separate metadata store to maintain.
- Anomaly detection tests (volume, freshness, column-level) run as dbt tests so they participate in normal dbt test workflows and CI pipelines without any special plumbing.
- Alert routing with owner tagging and custom Slack channels is configured in dbt model properties, which means it stays in version control alongside the models themselves.
- Active development and a large contributor base (100+ contributors) with cross-warehouse support covering Snowflake, BigQuery, Redshift, Databricks, and Postgres.
- The OSS/Cloud feature split is aggressive: column-level lineage, automated ML monitors, and BI tool integration are all cloud-only, so the self-hosted version is substantially less capable than the marketing implies.
- The generated report is a static HTML file—there's no persistent backend in the OSS version, so you lose historical trend data unless you manually manage archiving or host it somewhere with versioning.
- Anomaly detection relies on statistical baselines computed over your warehouse data, which means cold-start problems on new tables and potential for noisy alerts until enough history accumulates; tuning sensitivity and training periods adds ongoing maintenance burden.
- The repo is listed as primarily HTML (the docs site), which obscures the actual Python CLI code quality; the Python package itself has a somewhat sprawling dependency surface that can conflict with dbt adapter version pinning in complex environments.