Danielskry/Awesome-RAG

What it solves

Large language models have a fixed knowledge cutoff and cannot access external information on their own, which makes them prone to outdated or hallucinated answers. Retrieval-Augmented Generation (RAG) fixes this by dynamically retrieving relevant context from external knowledge sources during generation, letting models pull in up-to-date, domain-specific, or proprietary information. This repository collects the best resources for learning about and building such RAG systems, so a developer does not have to hunt across the web for tutorials, frameworks, and best practices.

How it works

The repository itself is an organized catalog, not a piece of software. It is structured into sections covering general RAG concepts, architecture patterns (from naive retrieve-then-generate pipelines to agentic and graph-based approaches), advanced techniques, frameworks that facilitate RAG, the Python ecosystem, chunking and embedding methods, metrics and evaluation, databases, production considerations, and best practices. Each entry links to authoritative sources, tutorials, runnable notebooks, papers, or implementations. The README also explains the core RAG idea with a concrete example: a system querying about renovation materials for a specific house can retrieve blueprints, material specs, and local building codes that a general LLM would not know. It is additionally available as an agent plugin for VS Code, GitHub Copilot CLI, and Claude Code, and can be queried via DeepWiki.

Who it's for

Developers, engineers, and researchers who want to explore or build RAG applications. Whether someone is starting from scratch, planning to adopt an existing framework, choosing chunking or embedding strategies, or preparing a RAG system for production, the catalog points them to relevant tools and materials. It also benefits anyone comparing advanced RAG variants or looking for evaluation and metric guidance.

Highlights

• Curated resource map covering the full RAG ecosystem: architecture patterns, advanced approaches, frameworks, techniques, metrics, databases, and production guidance
• Links to authoritative sources including runnable Python notebooks, tutorials for LangChain, LlamaIndex, and Haystack, papers, and YouTube explainers
• Covers a wide spectrum of advanced approaches, from Corrective RAG, RAG Fusion, and Self-Reflective RAG to GraphRAG, multimodal RAG, video RAG, and reasoning-based frameworks
• Includes detailed guidance on practical techniques such as chunking strategies, embeddings, and prompting, with pros and cons for each
• Available as an agent plugin for VS Code, GitHub Copilot CLI, and Claude Code, and queryable via DeepWiki