Problem statement

OpenSRE does not yet have a clearly documented, first-class Apache Airflow integration for DAG-level and task-run investigation. Teams using Airflow need direct workflow context rather than relying on indirect or partial signals.

Proposed solution

Add a first-class Apache Airflow integration that supports:

• onboarding, authentication, and connectivity verification
• reading DAG state, task runs, recent failures, retries, and scheduler context
• using Airflow evidence directly during investigations and in final RCA output
• connecting Airflow failures to downstream data platform and infrastructure signals when relevant

Acceptance criteria

• the integration is usable from the investigation workflow end to end
• setup and usage are documented clearly
• the implementation is backed by end-to-end tests in tests/e2e or equivalent
• the PR includes a screen video showing setup plus a successful investigation flow using Apache Airflow
• vibecoding without testing will not be allowed; PRs without working end-to-end coverage and screen-video proof should be rejected