If I understand correctly, the library doesn't index the codebase but just generates a good, formatted representation of the repo. I'm curious to hear what you think about more meaningful way to represent a codebase? There is a lot of tribal knowledge in the repo that can help the LLM.