[naveedjanmo]

The document is first broken up into chunks. When a user prompts, Unriddle finds the most relevant chunks and then feeds them into GPT-3.