[ANN] FlashRank.jl: Fast and Efficient Document Ranking

Introducing FlashRank.jl: Fast and Efficient Document Ranking

We are excited to announce the release of FlashRank.jl, a Julia package that provides a fast and efficient way to rank documents relevant to a given query without requiring GPUs or large dependencies. It’s heavily inspired by Python’s FlashRank.

Key Features

  • Four ranking models with varying sizes and accuracy
  • Lightweight dependencies for ease of integration
  • Fast ranking speeds, with the smallest model capable of ranking 100 documents in ~0.1 seconds on a laptop

Getting Started

To get started with FlashRank.jl, simply add it to your Julia environment using Pkg.add("FlashRank"). Import the package and you can use the RankerModel to rank documents for a given query.

Example Usage

using FlashRank

ranker = RankerModel()
query = "How to speedup LLMs?"
passages = ["...", "...", "..."]
result = rank(ranker, query, passages)

Integrated with PromptingTools.jl

FlashRank.jl is designed to work seamlessly with PromptingTools.jl, allowing you to easily integrate it into your Retrieval Augmented Generation (RAG) pipelines. See our documentation for more details.

Try it out!

Give FlashRank.jl a try today and experience fast and efficient document ranking for your RAG pipelines!