This repository provides a comprehensive collection of advanced and agentic Retrieval-Augmented Generation (RAG) techniques for researchers and developers. It simplifies the implementation and evaluation of complex RAG systems, offering ready-to-use notebooks that progress from naive RAG to sophisticated agentic approaches, thereby improving LLM accuracy and relevance with external data.

How It Works

The project implements RAG by breaking down documents into chunks, creating embeddings, and storing them in a vector store. A retriever then identifies relevant documents based on user queries. The retrieved context is augmented with the user's query into a prompt for an LLM, which generates the final response. This approach allows LLMs to access up-to-date, accurate information, mitigating issues like hallucinations and outdated knowledge.

Quick Start & Requirements

• Install by cloning the repository: git clone https://github.com/athina-ai/rag-cookbooks.git
• Navigate to the project directory: cd rag-cookbooks
• Follow detailed implementation guides within the cloned repository.
• Dependencies include LangChain, various vector stores (Pinecone, ChromaDB, Weaviate, Qdrant, FAISS), Athina AI, Unstructured, and LangGraph.
• A demo is available at: https://github.com/user-attachments/assets/c6f17961-40a1-4cca-ab1f-2c8fa3d71a7a

Highlighted Details

• Covers 10 advanced RAG techniques including Hybrid RAG, Hyde RAG, Parent Document Retriever, RAG fusion, Contextual RAG, Rewrite Retrieve Read, and Unstructured RAG.
• Includes 5 agentic RAG techniques: Basic Agentic RAG, Corrective RAG, Self RAG, Adaptive RAG, and ReAct RAG.
• Each technique is accompanied by research paper references for further exploration.
• Provides end-to-end RAG implementation and evaluation using Athina AI.

Maintenance & Community

• Open to community contributions for new techniques or improvements.
• No specific contributor, sponsorship, or roadmap information is detailed in the README.

Licensing & Compatibility

• Licensed under the MIT License.
• Permissive for commercial use and closed-source linking.

Limitations & Caveats

The repository focuses on advanced and agentic RAG techniques, assuming a foundational understanding of RAG and its core components. Specific hardware or computational resource requirements for running the notebooks are not detailed.