All Tools
R
RAGFreeOpen Source
RAPTOR
Tree-organized retrieval with recursive summaries
MIT
ABOUT
Flat chunk retrieval misses high-level themes in long documents because embeddings only see local passages. RAPTOR clusters chunks, summarizes them recursively into a tree, and retrieves at multiple levels so RAG can answer both detail and overview questions.
INTEGRATION GUIDE
1. Build a summary tree over long documents for multi-level RAG
2. Retrieve both fine-grained chunks and high-level abstracts
3. Improve long-document QA where flat vector search loses theme
TAGS
pythonragretrievalsummarizationtreelong-contextresearch