PageIndex
Tree-based document index for RAG without vectors or chunking
What it is
PageIndex builds a hierarchical tree index of each long document, such as a PDF report, and has an LLM reason through that tree to find the relevant sections instead of running similarity search over embeddings. Answers point back to explicit sections of the source. Developers building question-answering over financial filings, legal texts, manuals, or textbooks can run it locally as a Python SDK with their own model API key, or plug its tools into an existing agent framework.