It leverages an ML energy model to estimate the power of the current machine executing the pipeline and correlates that with the carbon grid intensity of the public IP.

It then can show directly in the Pull-Request how much energy and carbon is used.

That functionality is paired with an external dashboard that can be hosted which is also free and open source and can show the carbon emissions over time.

Here is an example link where we have for instance been tracking the carbon emissions of Django on GitHub: https://metrics.green-coding.io/ci.html?repo=green-coding-so...