Summary
The manual, made answerable.
QuantConnect ships its documentation as long PDFs, and the failure mode is a model that half-remembers the API and invents the rest. This keeps the text local, ranks passages instead of pasting files, and returns the excerpt with its page — so every answer can be checked against the page it came from.
What it is
A documentation search server for QuantConnect's PDF manuals. It converts the PDFs you point it at into markdown, builds a search index over them, and answers a query with ranked excerpts rather than a whole document.
What you get
- Ranked search across every converted document, with the matching passage and its page number
- Results scored by term frequency and boosted when your terms appear close together, so multi-word questions land on the right section
- The catalogue of documents it has processed, with their metadata
- The full text of a named document, or a single page of it
- Conversion skipped for files that have not changed, so restarting is not re-indexing
Requirements
Python, and the PDFs themselves — you tell it which folder holds the manuals and where the converted markdown should live. No account and no outbound call: it reads what is on your disk.