# PrithivirajDamodaran/FlashRank

Lite & Super-fast re-ranking for your search & retrieval pipelines.  Supports SoTA Listwise and Pairwise reranking based on LLMs and  cross-encoders and more.  Created by Prithivi Da, open for PRs & Collaborations.

Repository: https://github.com/PrithivirajDamodaran/FlashRank
Canonical: https://ross.abutalabs.com/products/flashrank
Language: Python
License: Apache-2.0
License Family: permissive
Topics: cross-encoder, full-text-search, hybrid-search, lexical-search, rag, ranking, reranking, retrieval-augmented-generation, semantic-search, vector-database, vector-search
Last push: 2026-07-11T08:06:27+00:00

## Health v2 (maintenance only)
Score: 58/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 92, release rhythm 8, longevity 71
- inputs: {"age_days": 1003, "days_push": 53, "days_rel": 642, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1003, forks 72 (observed 2026-09-03T02:15:07.082465+00:00)

## What it is
FlashRank is an ultra-light, fast Python library for re-ranking search and retrieval results using SoTA cross-encoders and listwise LLM-based rerankers. It requires no Torch or Transformers, runs on CPU, and includes models as small as ~4MB, making it well-suited for serverless and cost-sensitive deployments.

## Use cases
- rerank search results in a RAG pipeline before feeding an LLM
- add re-ranking to a vector database or hybrid search pipeline
- rerank documents on CPU without heavy dependencies like torch
- improve relevance of semantic search results with a cross-encoder
- deploy a tiny reranker in serverless functions like AWS Lambda
- rerank multilingual search results
- boost retrieval quality in a full-text or lexical search system

## When to choose
- you need fast, low-cost reranking on CPU with minimal dependencies
- you want to improve RAG retrieval quality with SoTA rerankers
- you're deploying to serverless environments where package size and cold start matter
- you need multilingual or listwise LLM-based reranking options

## When to avoid
- you need GPU-accelerated reranking of very large document sets
- you need a full search engine or vector database rather than a reranking layer
- you require models with context beyond 8192 tokens
- you need tight integration with a specific framework's native reranker APIs

## Facets
- artifact type: library
- maturity: active
- function: search-engine, rag, machine-learning, nlp
- domain: machine-learning
- platform: python, cli, cross-platform
- tags: reranking, cross-encoder, listwise-reranker, vector-search, semantic-search, lightweight, cpu-only, search, retrieval-augmented-generation, natural-language-processing

## Member repositories
- PrithivirajDamodaran/FlashRank (main) score 58

## Provenance
- Observed fields: from GitHub, fetched 2026-09-03T02:15:07.082465+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:16:04.101844+00:00, confidence not recorded.
  - readme: https://github.com/PrithivirajDamodaran/FlashRank (fetched 2026-09-03T02:15:07.082465+00:00, sha bed31c81128a)
  - registry_pypi: https://pypi.org/pypi/flashrank/json (fetched 2026-08-29T13:14:05.712628+00:00, sha 24364e2c4fec)
- Data as of 2026-08-30T08:39:29.467469+00:00.