I maintain a popular pytest plugin[1] and throughout its life have supported and observed many developers struggling with GitHub actions.
- It's hard to identify what caused a failure given the length of some ci logs - Multiple ci jobs can fail with the same cause meaning it's noisy - It's unclear how to prioritise fixes to these failures
This Github app gives you a prioritised, de-duplicated list of issues relating to your GitHub actions failure.
It uses LLMs (GPT3.5 at the moment) to identify the most likely root cause, highlight relevant source files, and order the issues by priority.
Feedback welcome!