Khata is a pre-commit Git hook that runs on the change that is upon submission to evaluate how risky the change is.
To setup Khata, developers need to place the files in this repository under directory .git/hooks/ in their projects. As long as pre-commit file is under the mentioned directory, Khata runs every time git commit command is run and displays the probabilty of the change being defective.
Currently, the model Khata is using to predict defectiveness of commits is JITGNN, a graph neural network model trained on thousands of commits to predict the probability of a given commit being defective. To extend Khata with other models, one need to build a subclass from HookInterface in file hook_interface.py. HookInterface contains the files that are modified in attribute modifieds and any subclass from this class has access to this attribute.
This class has an abstract method called run_model that should be implemented based on the new models that are extending Khata. This method essentially displays the probability of the commit being defective and exits with value 0 (sys.exit(0)). Exiting with any value other than zero aborts the commit which is not the desired behavior.