Self-healing for your dbt warehouse.
Cairn auto-detects metric anomalies in your dbt warehouse, attributes them to the offending commit, and opens a verified revert PR. Detect, attribute, recover. Zero human in the loop.
Self-healing console · incident attribution · sandbox recovery proof
One loop. Three motions.
Cairn runs after dbt changes ship. It watches the metric contracts, finds the commit that broke them, and verifies the safest revert before opening a PR.
Detect metric regressions.
Continuous anomaly detection runs on the dbt metric contracts your team defines: volume drops, freshness lapses, row-count regressions, ratio shifts, and null-rate spikes.
Trace the offending commit.
A cognition graph walks lineage, collects sample evidence, and uses LLM-aided root cause analysis to return a top cause plus ranked alternates with confidence scores.
Open a verified revert PR.
Every candidate revert runs against a zero-copy clone of the warehouse and the customer dbt project. If recovery is confirmed, Cairn opens the PR with evidence attached.
Three things had to be true.
Verified, not heuristic.
Every revert runs against a real zero-copy clone of the customer's warehouse and the customer's dbt project before any PR is opened. No fix ships because a model guessed.
Attributed, not blamed.
Every incident carries an evidence trail: lineage from broken metric to offending commit, sample-row deltas, and recovery proof. The audit trail is part of the product.
Autonomous, opt-in.
Cairn starts in Observe mode, drafts PRs in Assisted mode, and can auto-merge verified reverts in Autonomous mode. Teams graduate when the signal earns trust.
Cairn closes the loop after dbt changes ship.
Let Cairn watch the next broken metric.
Cairn's alpha is open to teams running dbt on BigQuery. Start in Observe mode with real production incidents, then move to Assisted or Autonomous recovery when your team trusts the signal.