parthsarthi03/raptor
A tree-based retrieval system that recursively processes documents into hierarchical summaries for improved RAG performance.
Not currently ranked — collecting fresh signals.
star history
RAPTOR implements a novel retrieval approach that organizes documents into a recursive tree structure, enabling more efficient and context-aware retrieval for language models. The system processes documents through abstractive summarization, clustering them into hierarchical levels to support multi-granularity retrieval. This enables language models to access relevant information at different abstraction levels during inference.
Frequently asked
- What is parthsarthi03/raptor?
- A tree-based retrieval system that recursively processes documents into hierarchical summaries for improved RAG performance.
- Is raptor open source?
- Yes — parthsarthi03/raptor is open source, released under the MIT license.
- What language is raptor written in?
- parthsarthi03/raptor is primarily written in Python.
- How popular is raptor?
- parthsarthi03/raptor has 1.7k stars on GitHub.
- Where can I find raptor?
- parthsarthi03/raptor is on GitHub at https://github.com/parthsarthi03/raptor.