How does it handle poor retrieval results?

It employs a grading node to filter irrelevant content; if the quality threshold isn't met, the system automatically routes to a query transformation node to rephrase the request and try again.

Which orchestration library is used for these patterns?

These patterns are designed for LangGraph, utilizing its StateGraph and conditional routing capabilities to manage complex, cyclic AI workflows.

Does this skill support web search fallbacks?

Yes, it implements Corrective-RAG (CRAG) patterns that can trigger web search tools like Tavily when local vector store documents are graded as irrelevant.

What is the difference between standard RAG and Agentic RAG?

Standard RAG follows a linear retrieve-then-generate path, while Agentic RAG uses LLM-driven decision nodes to grade documents, rewrite queries, or search alternative sources if the initial results are poor.

Agentic RAG Patterns

Name: Agentic RAG Patterns
Author: yonatangross

byyonatangross

•

Data Science & ML

Implements advanced Self-RAG and Corrective-RAG architectures for self-correcting AI retrieval systems.

Agentic RAG Patterns provides a comprehensive framework for building sophisticated retrieval-augmented generation pipelines using LangGraph. It enables Claude to implement adaptive retrieval strategies, including document grading, query rewriting, and web-search fallbacks (CRAG). This skill is essential for developers building production-grade AI agents that need to verify information quality, handle insufficient data gracefully, and reduce hallucinations through multi-step reasoning and recursive feedback loops.

Key Features

01Knowledge Graph integration for entity-rich retrieval

02Corrective-RAG (CRAG) with automated web search fallback

03Self-correcting retrieval workflows using LangGraph 1.0.6

04Query transformation and adaptive re-writing logic

0569 GitHub stars

06Automated document grading for relevance verification

Use Cases

01Developing research agents that fallback to web search when internal data is insufficient

02Implementing complex data pipelines requiring multi-step reasoning and state management

03Building a customer support bot that validates source documents before answering

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yonatangross/orchestkit agentic-rag-patterns

For use in Claude.ai and ChatGPT

Download Skill