This repository provides a comprehensive, hands-on implementation of various Retrieval-Augmented Generation (RAG) techniques, targeting developers and researchers seeking to understand and apply advanced RAG methods. It offers clear, runnable Python code for each technique, demystifying RAG by focusing on fundamental implementations without relying on heavy frameworks like LangChain.

How It Works

The project breaks down complex RAG strategies into digestible, step-by-step Python notebooks. Each notebook details a specific RAG enhancement, such as semantic chunking, context-enriched retrieval, query transformation, reranking, and fusion retrieval. The core approach emphasizes building these techniques from scratch using common libraries like openai, numpy, and pymupdf, allowing for deeper comprehension and easier modification.

Quick Start & Requirements

• Installation: Primarily involves cloning the repository and installing dependencies via pip.
• Prerequisites: Requires Python, openai API key, and potentially PDF documents for processing. Specific CUDA versions or hardware are not explicitly mandated for core functionality.
• Resources: Links to detailed explanations and related GitHub repositories are provided in the README.

Highlighted Details

• Implements 18 distinct RAG techniques, covering a wide spectrum from basic RAG to advanced methods like Self-RAG, CRAG, and Graph RAG.
• Focuses on foundational implementations, avoiding reliance on high-level orchestration frameworks.
• Includes techniques for improving chunking, retrieval, reranking, and query processing.
• Demonstrates methods for integrating knowledge graphs and hierarchical indexing.

Maintenance & Community

Information on maintainers, community channels (like Discord/Slack), or a public roadmap is not detailed in the README. The project appears to be a personal or small-team effort focused on educational implementation.

Licensing & Compatibility

The repository's license is not explicitly stated in the provided README. Users should verify licensing for commercial use or integration into closed-source projects.

Limitations & Caveats

The project is currently text-based and does not include implementations for multimodal RAG. While the code is presented as runnable, the complexity of integrating and tuning each RAG technique may require significant effort and domain expertise.