Skip to main content

​
RAG Patterns

Common architectural patterns for building Retrieval-Augmented Generation systems.

​
Basic RAG

Standard retrieve-then-generate pattern. 
from langchain.chains import RetrievalQA
from langchain_openai import ChatOpenAI
from langchain_chroma import Chroma

llm = ChatOpenAI(model="gpt-4o")
vectorstore = Chroma(embedding_function=embeddings)

qa = RetrievalQA.from_chain_type(
    llm=llm,
    retriever=vectorstore.as_retriever(search_kwargs={"k": 4})
)

response = qa.invoke({"query": "What is RAG?"})

​
Agentic RAG

Agent decides when and how to retrieve. 
from agno import Agent, OpenAI
from agno.tools.retrieval import RAGTool

agent = Agent(
    model=OpenAI(id="gpt-4o"),
    tools=[RAGTool(vectorstore=vectorstore)],
    reasoning=True
)

response = agent.run("Analyze the climate data")

​
Corrective RAG (CRAG)

Self-evaluating retrieval with fallback strategies. 
from langgraph.graph import StateGraph

workflow = StateGraph(RAGState)
workflow.add_node("retrieve", retrieve)
workflow.add_node("grade", grade_documents)
workflow.add_node("generate", generate)
workflow.add_node("transform_query", transform_query)
workflow.add_node("web_search", web_search)

workflow.add_conditional_edges(
    "grade",
    decide_to_generate,
    {"transform_query": "transform_query", 
     "generate": "generate",
     "web_search": "web_search"}
)

app = workflow.compile()
Combine vector and keyword search. 
from langchain.retrievers import EnsembleRetriever
from langchain_community.retrievers import BM25Retriever

# Vector retriever
vector_retriever = vectorstore.as_retriever(search_kwargs={"k": 4})

# BM25 retriever
bm25_retriever = BM25Retriever.from_documents(documents)

# Combine
ensemble_retriever = EnsembleRetriever(
    retrievers=[vector_retriever, bm25_retriever],
    weights=[0.5, 0.5]
)

RAG Overview

Learn RAG fundamentals

Advanced Techniques

Explore advanced RAG patterns

Build docs developers (and LLMs) love

Get started for freeTalk to us