Hi! I really love Cursor, but it still can’t quickly find the right context in index to answer an arbitrary question.
I was extremely impressed with Augment Code’s indexing speed and fast search and analysis capabilities. I checked on several rather large repositories (up to 10 MB of code), and after indexing Augment Code gives instant answers to arbitrary questions about the codebase (for example, where some specific algorithm is in the project, how it works and where is it used).
After indexing, it himself offers examples of questions on the codebase and generally makes it seem as if it understands the whole codebase.
Their indexing algorithm is most likely their know-how, but maybe you can implement something similar?
This algorithm might also be useful for working with one’s documentation or knowledge base (“second brain”), not just for codebase.