Devo started from a very real pain. CI and CD fails. Everyone gets a red build. Then the team wastes time doing the same boring detective work. Scroll logs, search Slack, guess the root cause, and repeat.
I built Devo to make that whole loop faster and way less annoying. It pulls failure logs, groups similar failures, and explains what most likely broke in simple language. It also keeps history, so if the same issue happened before, it points you to the old fix instead of making you rediscover it.
The system does three big things. First, it cleans and structures messy logs so they become searchable and comparable. Second, it categorizes failures, like flaky tests, dependency issues, infra timeouts, or real code bugs. Third, it gives an action plan, like which step failed, what changed recently, and what to try next. It also tracks trends, so you can see which failure type is getting worse over time.
This project was super fun because it mixes engineering and ML vibes. You have to deal with noisy real world logs, build good grouping logic, and still keep the UX simple. The goal is not to be fancy. The goal is to save developer time and keep teams moving.