Refactor a File Dump Into a Retrieval-Ready Knowledge Pack
Reshape messy source files into a smaller, clearer knowledge pack optimized for retrieval.
Source dump: eight overlapping files, vague names, mixed topics, and a slide deck that duplicates policy text from a markdown file. The assistant often cites the wrong thing. Messy corpus -> clearer signals -> better retrieval The trap is to keep every historical file “just in case.” That feels safe to humans while making ranking and source selection much worse for the assistant. Identify the canonical source Mark one file as the current authority for each topic and decide which duplicates can be removed or archived. If you do not know the canon, the assistant definitely does not know it.…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in