RunLens for Apache Airflow

[rahrajlat]

RunLens — Compare DAG Runs and Spot Performance Regressions in the Airflow UI

When an Airflow DAG suddenly takes longer than usual, one of the first questions is:

What actually changed between this run and the last good run?

Answering that often means opening multiple task instances and manually comparing durations, retries, and states.

I built RunLens, an open-source plugin for Apache Airflow 3 that makes these comparisons directly inside the Airflow UI.

RunLens lets you:

• Compare any two DAG runs task-by-task
• See task duration changes and identify performance regressions
• Compare task states and retry counts
• Detect tasks that were added or removed between runs
• View DAG runtime trends over the last 10, 30, 60, or 90 days
• Compare the latest task duration against its historical median

The goal is to make two questions easier to answer:

What changed between these two runs?

Is this DAG behaving differently from normal?

RunLens runs as a native Airflow plugin, with a FastAPI backend and React/TypeScript frontend, so the diagnostics stay alongside the DAG rather than requiring a separate monitoring tool.

Tech stack

Apache Airflow 3 · Python · FastAPI · React · TypeScript

The project is still early, and I’m exploring additional run-level diagnostics and regression detection features.

It’s open source under the Apache License 2.0.

GitHub: https://github.com/rahrajlat/airflow-run-lens

[raphaelauv]

typo in the link , it's https://github.com/rahrajlat/airflow-run-lens

thanks for sharing this , would be great to package and publish it in pypi like https://github.com/AlvaroCavalcante/airflow-calendar-plugin

[rahrajlat]

Thanks , sure let me work on that