Rankify

Overview of Rankify

What it is

Rankify is a Python toolkit for retrieval, re-ranking, and retrieval-augmented generation (RAG). It integrates 40 pre-retrieved benchmark datasets and supports 7+ retrieval techniques, 24+ state-of-the-art reranking models, and multiple RAG methods. The framework is designed to be modular and efficient, working with state-of-the-art models for retrieval, ranking, and RAG tasks.

Key features

• Integrates 40+ pre-retrieved benchmark datasets with 1,000 pre-retrieved documents each - Supports 7+ retrieval techniques including BM25, DPR, ANCE, Contriever, ColBERT, BGE, and online retrieval via CrawAI and ReACT - Includes 24+ state-of-the-art reranking models such as FlashRank, Mono T5, LiT5, and others - Provides multiple RAG methods and a pipeline API for one-line usage (search, rerank, rag)
• Offers a RankifyAgent for AI-powered model selection and a Rankify Server for REST API deployment - Integrates with LangChain and LlamaIndex

Use cases

Rankify can be used for document retrieval, document re-ranking, and full RAG pipelines. It is suitable for question answering, search systems, and any application requiring retrieval and generation of information from large document collections. The pre-retrieved datasets allow for quick benchmarking and evaluation.

Getting started

bash conda create -n rankify python=3.10 conda activate rankify pip install torch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1 --index-url https://download.pytorch.org/whl/cu124 pip install rankify # or for full functionality: pip install 'rankify[all]' For the latest development version from GitHub: bash git clone https://github.com/DataScienceUIBK/rankify.git cd rankify pip install -e '.[all]'

License and pricing

The license is not stated in the provided metadata or README. The project is listed as open source.

Based on the project's public README and repository metadata.