Awesome-RAG is a curated list of resources for Retrieval Augmented Generation (RAG) systems, covering foundational concepts, advanced patterns, tools, and evaluation methods. It targets engineers and researchers building or optimizing LLM applications that leverage external knowledge. The collection aims to provide a comprehensive overview of the RAG landscape, helping users navigate its complexities and improve their system's performance and reliability.

How It Works

The repository organizes information into thematic sections, detailing various aspects of RAG. It explores different retrieval strategies (vector, RAG Fusion, BM25), chunking techniques (semantic, positional), embedding models, and prompt engineering approaches. The content also delves into multi-modal and multi-document RAG, long-context RAG, and the integration of knowledge graphs. This structured approach allows users to deep-dive into specific components or gain a holistic understanding of RAG system design.

Quick Start & Requirements

This repository is a collection of links and concepts, not a runnable codebase. No installation or specific requirements are needed to browse the content.

Highlighted Details

• Extensive coverage of RAG patterns, including Dialogue Routing, FLARE, and Chain-of-Verification.
• Detailed sections on embedding models, vector search, and chunking strategies for RAG.
• Exploration of RAG evaluation metrics, performance optimization, and cost reduction techniques.
• Includes vendor-specific examples and comparisons of popular RAG tools like HayStack, RAGAS, and LlamaIndex.

Maintenance & Community

The repository is maintained by frutik. Further community or maintenance details are not specified in the README.

Licensing & Compatibility

The repository itself is a collection of links and does not have a specific license. The linked resources may have their own licenses.

Limitations & Caveats

The README explicitly states that the outline is in "complete draft form" and "everything is in motion." This indicates the content may be incomplete, unorganized, or subject to significant changes.