AUTOMATED ETL OBSERVABILITY

Trace your ETL lineage
with AI precision.

Tracepipe maps how every column is built and catches logic mistakes before they reach your production dashboards.

Install Star on GitHub
Native support for DuckDB Databricks + Unity Catalog Airflow DAGs
cursor — tracepipe
// what it does

Three things, done well.

Ask in plain English from your IDE. Tracepipe handles the rest.

01 / LINEAGE

Trace any column to its source.

Walks the dependency graph through CTEs, JOINs, CASE WHENs, and Spark UDFs. Tells you exactly where a number comes from.

› how is risk_level calculated?
↳ silver.fact_jobs · L112
02 / PEER REVIEW

Catches what CI doesn’t.

Reads your git diff. Flags syntax bugs, grain mismatches, misleading aliases. Assigns a risk level so you know whether to ship.

✗ SUM(salary) AS avg_salary
✗ SELECT col not in GROUP BY
03 / IMPACT

See what breaks before it breaks.

Pick a table you’re about to change. See every downstream dependency, direct and transitive, with a risk score.

● YELLOW · 3 affected
conformed.churn_risk + 2 more
// peer review

The senior engineer
you wish was reviewing your PRs.

Real output from a real diff. Risk levels, technical issues, downstream impact — all in your terminal.

It catches the bugs your test suite never sees.

SELCT, FRON, JOUN. Aggregations missing GROUP BY. Misleading aliases. Grain mismatches that silently corrupt metrics three weeks later.

Optional git pre-commit hook blocks any RED diff from leaving your machine.

PEER REVIEW · 02_corporate_logic.sql
● RED

Senior notes

silver.fact_jobs feeds 5 downstream tables.

conformed.company_stats — changed

AVG(salary) GROUP BY company_name SUM(salary) GROUP BY company_id

Technical issues

SELECT company_name not in GROUP BY
Misleading alias: SUM(...) AS avg_salary
// 60 seconds

One command. Wizard does the rest.

Picks your DB, scans your SQL, configures Cursor or Claude Desktop’s MCP for you.

$ curl -fsSL https://raw.githubusercontent.com/oarbel11/tracepipe_ai/master/install/install.sh | bash
Works with DuckDB, Databricks & Airflow · Python 3.8+ required
// contact

Questions, feedback, ideas?

Drop a line. Goes straight to the creator’s inbox.

Free, open source, MIT.

Built by a data engineer who got tired of breaking dashboards. PRs, bug reports, and odd use cases all welcome.

★ Star on GitHub