Inspiration
Shipping UI changes without a clear visual gate burns reviewer time and CI cycles. We wanted a guardian that fails loudly on real visual regressions—and makes the environmental cost of flaky or failing runs visible, so teams care about both quality and footprint.
What it does
Lumi is a Visual Regression Guardian with CO₂ measurement for GitLab pipelines. It compares UI screenshots across builds, flags meaningful diffs, and estimates CO₂ impact tied to failed or wasteful CI work—so you catch UI breaks early and see the cost of letting them slip.
How we built it
- GitLab CI job(s) that capture and compare visual baselines
- Diff / regression signals surfaced in the pipeline (not only in a local folder)
- A lightweight CO₂ estimate path tied to CI outcomes so the story is quality + sustainability
- Project packaging in our Transcend showcase repo for Path A submission
Challenges
Balancing sensitivity (catch real UI breaks) vs noise (ignore harmless layout jitter), keeping the guardian usable in a hackathon CI setup, and telling a clear demo story in under 3 minutes on video.
What we learned
Visual gates need clear baselines and reviewable artifacts. Pairing regression signals with a simple CO₂ narrative helps the submission stand out beyond “another screenshot diff tool.”
What's next
Tighter thresholds, richer review UI for diffs, and better CO₂ factors once we can measure real runner energy profiles.
Log in or sign up for Devpost to join the conversation.