Standard RAG (retrieval-augmented generation) works by chopping documents into text chunks and searching them when you ask a question. It is the most common way to make an AI answer questions about your own files. The problem: chunks have no awareness of how ideas connect across a book or document. Ask a nuanced question that spans multiple chapters and you will often get a shallow or contradictory answer.
BookGraph, a new open-source project on GitHub, takes a different approach. Instead of storing flat text chunks, it parses books into a knowledge graph - a structured web of concepts, entities, and the relationships between them. When you query it, the AI reasons over those connections rather than just pattern-matching against nearby sentences.
The practical difference matters most for long, complex documents. A textbook on economics, for example, introduces concepts in chapter 2 that get applied in chapter 8. Traditional RAG might retrieve both passages independently and miss the link. A graph structure preserves that link explicitly, so the AI can follow the reasoning chain the author intended.
BookGraph is early-stage and focused specifically on books, not general enterprise documents. For anyone building internal knowledge bases or research tools, though, it is a clean demonstration of where RAG needs to go. The "just chunk everything" era is hitting its ceiling, and graph-based approaches like this one are the most promising path past it.