PrithivirajDamodaran/FlashRank
Ultra-lightweight Python library for re-ranking search results in RAG pipelines using cross-encoders and LLM-based listwise rerankers, running on CPU without Torch or Transformers.
Not currently ranked — collecting fresh signals.
star history
FlashRank provides state-of-the-art pairwise and listwise re-ranking capabilities for search and retrieval pipelines. It integrates cross-encoder-based rerankers for fast pointwise/pairwise scoring and LLM-based listwise rerankers for more sophisticated ranking. The library is designed as a lightweight (~4MB models), CPU-friendly addition to RAG pipelines that operates without heavy ML framework dependencies. Users feed initial retrieval results through rerankers before passing them to downstream LLMs to improve answer quality.
Frequently asked
- What is PrithivirajDamodaran/FlashRank?
- Ultra-lightweight Python library for re-ranking search results in RAG pipelines using cross-encoders and LLM-based listwise rerankers, running on CPU without Torch or Transformers.
- Is FlashRank open source?
- Yes — PrithivirajDamodaran/FlashRank is open source, released under the Apache-2.0 license.
- What language is FlashRank written in?
- PrithivirajDamodaran/FlashRank is primarily written in Python.
- How popular is FlashRank?
- PrithivirajDamodaran/FlashRank has 995 stars on GitHub.
- Where can I find FlashRank?
- PrithivirajDamodaran/FlashRank is on GitHub at https://github.com/PrithivirajDamodaran/FlashRank.