#3737th of 67 in RAG, knowledge and search

flashrag

Python toolkit for reproducing and building RAG pipelines on benchmarks

Stars
3.6k
License
MIT
Last commit
Oct 2026
Last release
Oct 2026
Language
Python
Stars, last 29 weeks+192
25 Apr 2026: 3,422 stars2 May 2026: 3,432 stars9 May 2026: 3,441 stars16 May 2026: 3,452 stars23 May 2026: 3,455 stars30 May 2026: 3,460 stars6 Jun 2026: 3,465 stars13 Jun 2026: 3,472 stars20 Jun 2026: 3,478 stars27 Jun 2026: 3,484 stars4 Jul 2026: 3,489 stars11 Jul 2026: 3,494 stars18 Jul 2026: 3,501 stars25 Jul 2026: 3,509 stars1 Aug 2026: 3,517 stars8 Aug 2026: 3,522 stars15 Aug 2026: 3,525 stars22 Aug 2026: 3,538 stars29 Aug 2026: 3,549 stars5 Sept 2026: 3,553 stars12 Sept 2026: 3,557 stars19 Sept 2026: 3,567 stars26 Sept 2026: 3,575 stars3 Oct 2026: 3,581 stars10 Oct 2026: 3,588 stars17 Oct 2026: 3,593 stars24 Oct 2026: 3,596 stars31 Oct 2026: 3,606 stars7 Nov 2026: 3,612 stars14 Nov 2026: 3,614 stars

3,422 on 25 Apr 20263,614 on 14 Nov 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

FlashRAG is a Python library for RAG research. It provides retrievers, rerankers, generators and compressors that can be assembled into pipelines, with 36 pre-processed benchmark datasets and 23 implemented RAG algorithms, 7 of them reasoning-based. It includes index-building scripts for Faiss, BM25s, Pyserini and SPLADE (Seismic), plus a UI for configuring and evaluating methods.

Who it is for: Researchers benchmarking or prototyping RAG methods in Python

Strengths

  • 36 pre-processed RAG benchmark datasets and 23 algorithms with reported results
  • Dense (Faiss), BM25 (bm25s or Pyserini) and SPLADE retrieval via Seismic
  • Custom pipelines by subclassing BasicPipeline, configured through YAML or dict
  • Optional vLLM for faster generation; multimodal RAG and OpenAI generators supported

Weaknesses

  • Research-oriented; not a ready-made document Q&A app for end users
  • Faiss has to be installed through conda because the pip package is incompatible
  • faiss-gpu is Linux x86_64 only; latest Faiss unavailable on some systems
  • Roadmap still lists API-based retriever (vLLM server) support as not done

What it needs

  • GPU optional
  • Needs Faiss, PyTorch, vLLM (optional), Pyserini or bm25s, sentence-transformers (optional), Serper API key (web search retriever)
  • Models: OpenAI models, vLLM, FastChat, Hugging Face embedding models (E5, BGE), Llava, Qwen, InternVL (multimodal)

What changed

From the release notes. No flag means the note does not say so, not that the upgrade is safe.

Also in RAG, knowledge and search

See all 67
Also in RAG, knowledge and search
RankProjectScore
1Open NotebookSelf-hosted NotebookLM alternative for research notebooks, chat, and podcasts42.1k starsMIT92 out of 100
2LightRAGGraph-based RAG framework with a REST server and WebUI40.8k starsMIT92 out of 100
3RAGFlowRAG engine with document parsing, agentic retrieval and citations Live demo ↗ (opens in a new tab)94.4k starsApache-2.090 out of 100
#4deeptutorTutoring workspace with knowledge bases, guided learning, memory and quizzes60.2k starsApache-2.090 out of 100
#5MaxKBEnterprise knowledge-base agent platform with RAG, workflows and MCP tools23.2k starsGPL-3.089 out of 100