Given the limited context length of most LLMs, is there value in turning in an entire codebase into a doc to feed it into an LLM? I think cherry-picking releva

Given the limited context length of most LLMs, is there value in turning in an entire codebase into a doc to feed it into an LLM? I think cherry-picking relevant sections would be necessary to make it function effectively. Has anyone tried using tree-sitter to recursively feed it the source for functions used in the section we want to analyze to optimize for this?