# Rankify

Rankify is a Python toolkit for retrieval, re-ranking and retrieval-augmented generation, integrating 40 pre-retrieved benchmark datasets, more than 7 retrieval techniques, more than 24 reranking models, and multiple RAG methods.

Rankify is a modular framework for retrieval, reranking and RAG tasks, designed to work with a range of state-of-the-art models across those three stages. It bundles a Pipeline API for running a full retrieve-rerank-generate flow, a RankifyAgent, a Rankify Server, and integrations with external tools, alongside a Streamlit-based web playground and demo.

It ships CLI indexing commands for retrieval methods such as BM25, DPR, ANCE, Contriever, ColBERT and BGE, and optional installation extras for retriever, reranking, RAG or all components together. Recent updates added more than 15 dense retrievers, including LLM-based bi-encoders such as SFR, E5 and GritLM, and reasoning-augmented retrieval models.

It is written in Python, is installed into a conda environment, and recommends installing a CUDA build of PyTorch when a GPU is available since many of its evaluation metrics are optimized for GPU use.

Website: https://rankify.readthedocs.io/