dbt Pipeline Log-Gap Watcher via Axiom

Many dbt failures are silent — the orchestrator dies before it emits a failure, so nothing alerts. This workflow inverts the logic: it asks Axiom whether each expected model logged a successful completion inside its SLA window. Any model that should have run but produced no success log is treated as a missed run and surfaced immediately.

Use it when absence of a signal is the real risk. Threshold alerts fire on bad data that arrives; this catches data that never arrives at all because the job never ran. Best for teams shipping dbt run telemetry to Axiom.

1. 1A schedule fires at each model's expected-completion checkpoint.
2. 2Axiom is queried for `dbt.run.success` events in the trailing window.
3. 3A logic step diffs the expected model list against models that logged success.
4. 4A branch ignores models inside a grace period to avoid false alarms.
5. 5For each genuinely missing run, the last observed completion time is pulled from Axiom.
6. 6A Slack alert names each silent model and how long since its last success.