I wrote 'baler' over a couple of weekends to fix this. Now you can run your repository through baler and get the fewest files before interacting with an LLM.

You can also tune output file's size, filter input files based on their size or line-count.

Please give it a try, and I hope it makes some of your work easier.