Open source -
AI Test Triage: failing tests explained in plain English
A tool that reads a failed test report and explains what broke in plain English, so no one has to dig through logs.
When a company's automated tests fail, someone has to open long, technical reports and work out what actually broke. We built a tool that does this instantly instead.
It reads the test results and posts a short, plain-English summary of what failed and why, so the team can fix it without digging through logs.
Since this post, the Action has grown up. It now labels each failure as a product bug, a test bug, an environment or flaky problem, or unclear, and it never changes the build result: a red build stays red. On a hand-written set of 40 labelled failures it classified 39 correctly, which we treat as an early test rather than proof. Adding it to your own repo takes one step in a GitHub Actions workflow, and you can try the same model in your browser on our demo page.
Written by Rushi Prajapati