Scribd
Upload
704 views5 pages

Hands-On Guide To Agentic Corrective RAG-1

Uploaded by

luckymishra0734
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
704 views5 pages

Hands-On Guide To Agentic Corrective RAG-1

Uploaded by

luckymishra0734
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Hands-on Guide

to
Agentic
Corrective
RAG Systems

Dipanjan (DJ)
Corrective RAG System

The inspiration for our agentic RAG system will be based on the solution
proposed in the paper, Corrective Retrieval Augmented Generation, Yan et
al. , where they propose a workflow as depicted in the following figure to
enhance a regular RAG system.

The key idea here is to retrieve document chunks from the vector
database as usual and then use an LLM to check if each retrieved
document chunk is relevant to the input question.

If all the retrieved document chunks are relevant, then it goes to the LLM
for a normal response generation like a standard RAG pipeline.

However, if some retrieved documents are not relevant to the input


question, we rephrase the input query, search the web to retrieve new
information, and send it to the LLM to generate a response.
Source: A Comprehensive Guide to Building Agentic RAG Systems with LangGraph Created by: Dipanjan (DJ)
Agentic Corrective RAG
System Workflow

Main Flow
Step 1 - Retrieve Node
Retrieves context documents from the vector database from the input query.
Step 2 - Grade Node
Use an LLM to grade if retrieved documents are relevant to the input question
– yes or no.
Step 3A - Generate Answer Node
If all documents are relevant (all 'yes'), send them to an LLM for response
generation.

Alternative Flow (if any retrieved documents are not relevant)


Step 3B - Rewrite Query Node
If some or all documents are not relevant (at least one 'no'), rephrase the query.
Step 3C - Web Search Node
Search the web to get context information using the rephrased query.
Step 3D - Generate Answer Node
Send the rephrased query and context documents or information to the LLM
for response generation.

Source: A Comprehensive Guide to Building Agentic RAG Systems with LangGraph Created by: Dipanjan (DJ)
Detailed Agentic
Corrective RAG System
Architecture

Query and Context Retrieval


User query sent to vector DB (e.g., Chroma) to retrieve context documents.
If no documents are retrieved, proceed to rephrase the query.
Document Grading
LLM grades retrieved documents as ‘Yes’ (relevant) or ‘No’ (irrelevant).
Decision Node
All Documents Relevant: Follow standard RAG flow; send query and documents to LLM for
response.
Irrelevant/No Documents: Rephrase query using LLM for optimized web search.
Web Search
Use web search tool (e.g., Tavily) to retrieve additional context documents.
Response Generation
Send combined context documents and query to LLM to generate the final response.
Source: A Comprehensive Guide to Building Agentic RAG Systems with LangGraph Created by: Dipanjan (DJ)
Hands-on Guide

Check out the


HANDS-ON GUIDE
here

You might also like