This is about creating a script that takes the output of the synthetic generation as an input and using an LLM as a Judge separates the question-answer list into different groups based on a predefined list of groups (eg by component).

The script should be added to the same repository that generates the Q&A.

Outcome: The script and a set of files with the different questions-answer groups.