Run the app with:
python main.py

The HuggingFace dataset configs are stored in the same directory the app is run from.  It will be in a .cache directory in the root directory where main.py is.  This directory is included in the .gitignore file so it isn't a part of the GitHub repo itself.  If you delete the downloaded GitHub repo it will delete the cached downloads also.

I have included the results of a full run of the app:

- analysis_summary.txt
- analysis_summary_paragraphs.txt (this is not generated by the app; this is my own analysis)
- log files
- csv files
- charts in the figures directory

If you run the app, it just overwrites all of these.  Currently with the same data, it will produce the same thing, but if the data did change, it will create new versions overwriting the older ones.

A note on q2_review_overview.png.  Category Distribution is 'What percentage of ALL COMMENTS are about correctness?" Compared to Primary Category which is "For each PR, what category got the MOST comments?  The percentage in Primary Category is of counting how many PRs belong to each category."

Category Distribution is working at the overall comment level, not distinguishing PRs.  Primary Category is working on the PR level.