FlexRAG
An open-source framework for quickly building, evaluating and reproducing retrieval-augmented generation systems, covering text-based, multimodal and web-accessible scenarios with an end-to-end pipeline from data preparation to evaluation.
FlexRAG provides a modular architecture that researchers can customize and extend for their own retrieval-augmented generation experiments, aiming to make it fast to reproduce published RAG systems and to build demonstrative prototypes around a new algorithm. Its retriever component can load indexes directly from the Hugging Face Hub, and the project publishes benchmark results on single-hop question answering tasks along with a dedicated run_retriever entry point and several information retrieval metrics.
It installs with pip, including prebuilt wheels for Windows, and ships English and Chinese documentation covering installation, a quickstart and a longer tutorial.