A curated list of resources about Retrieval-Augmented Generation (RAG) — papers, tools, frameworks, tutorials, and benchmarks.
RAG combines the power of large language models with external knowledge retrieval, enabling more accurate, grounded, and verifiable AI responses.
- Papers
- Frameworks & Libraries
- Tools
- Tutorials & Guides
- Benchmarks & Evaluation
- Production Examples
- Datasets
- Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks - The original RAG paper by Lewis et al. (2020)
- REALM: Retrieval-Augmented Language Model Pre-Training - Google's retrieval-augmented pre-training approach
- Atlas: Few-shot Learning with Retrieval Augmented Language Models - Meta's few-shot RAG model
- Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection - Adaptive retrieval with self-reflection
- RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval - Hierarchical document summarization for RAG
- Corrective RAG (CRAG) - Self-corrective retrieval augmented generation
- Active RAG - Active learning approaches for RAG
- Chain-of-Note: Enhancing Robustness in RAG - Sequential note-taking for better retrieval
- Lost in the Middle - How LLMs use long contexts (critical for RAG design)
- When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parametric Memories
| Name | Description | Language |
|---|---|---|
| LangChain | Full RAG pipeline framework with 100+ integrations | Python/JS |
| LlamaIndex | Data framework for LLM apps, strong RAG focus | Python |
| Haystack | End-to-end NLP/RAG framework by deepset | Python |
| RAGFlow | Open-source RAG engine with deep document understanding | Python |
| Verba | Golden RAGtriever — open-source RAG app by Weaviate | Python |
| Canopy | RAG framework built on Pinecone | Python |
| FastRAG | Research framework for efficient RAG by Intel | Python |
| RAGAS | RAG evaluation framework | Python |
| Embedchain | Framework to create RAG bots over any dataset | Python |
- Pinecone - Managed vector database
- Weaviate - Open-source vector database
- Qdrant - High-performance vector search engine
- Chroma - Open-source embedding database
- Milvus - Cloud-native vector database
- pgvector - Vector similarity search for Postgres
- sentence-transformers - State-of-the-art text embeddings
- Nomic Embed - Open-source long-context embeddings
- Jina Embeddings - Multilingual embeddings API
- Cohere Embed - Enterprise embedding models
- Unstructured - Pre-processing for unstructured data
- LlamaParse - Document parsing for LLM apps
- Docling - IBM's document conversion library
- RAG from Scratch (LangChain) - Complete video + code series
- Building RAG Applications (LlamaIndex) - Official guide
- Advanced RAG Techniques - Collection of advanced patterns
- Pinecone RAG Guide - Comprehensive introduction
- Anthropic RAG Cookbook - RAG with Claude
- OpenAI RAG Best Practices - Official optimization guide
- RAGAS - Faithfulness, relevance, and context metrics
- ARES - Automated RAG evaluation system
- RGB Benchmark - Retrieval-augmented generation benchmark
- RECALL Benchmark - Benchmark for LLM robustness to external context
- TruLens - Evaluation and tracking for LLM apps
- DeepEval - Unit testing framework for LLMs
- CoreProse KB-Incidents - Citation-first RAG pipeline with 13,000+ indexed passages and zero hallucination rate. Every AI response includes verifiable source references.
- Perplexity AI - Search engine powered by RAG with inline citations
- Glean - Enterprise search with RAG
- You.com - AI search with source attribution
- MS MARCO - Large-scale reading comprehension & QA dataset
- Natural Questions - Real Google queries with Wikipedia answers
- HotpotQA - Multi-hop question answering dataset
- SQuAD 2.0 - Reading comprehension benchmark
- TriviaQA - Large-scale QA with evidence documents
Contributions welcome! Please read the contributing guidelines first. Open a PR or issue to add resources.
This list is released under MIT License.