How to Intermediate · 3 min read

How to use create_retrieval_chain in LangChain

Q: How to use create_retrieval_chain in LangChain

Use create_retrieval_chain from langchain.chains to quickly build a retrieval-augmented generation chain by combining a vectorstore retriever and an LLM. Pass your retriever and LLM instance to the function, then call the chain with a query to get context-aware answers.

Quick answer

Use create_retrieval_chain from langchain.chains to quickly build a retrieval-augmented generation chain by combining a vectorstore retriever and an LLM. Pass your retriever and LLM instance to the function, then call the chain with a query to get context-aware answers.

PREREQUISITES

Python 3.8+
OpenAI API key (free tier works)
pip install langchain openai faiss-cpu

Setup

Install the required packages and set your OpenAI API key in the environment variables.

bash

pip install langchain openai faiss-cpu

Step by step

This example shows how to create a simple retrieval chain using a FAISS vectorstore and OpenAI's GPT-4o model for answering questions based on documents.

python

import os
from langchain_openai import ChatOpenAI
from langchain_community.vectorstores import FAISS
from langchain.chains import create_retrieval_chain
from langchain_community.document_loaders import TextLoader
from langchain_openai import OpenAIEmbeddings

# Load documents
loader = TextLoader("example_docs.txt")
docs = loader.load()

# Create embeddings
embeddings = OpenAIEmbeddings(openai_api_key=os.environ["OPENAI_API_KEY"])

# Build FAISS vectorstore
vectorstore = FAISS.from_documents(docs, embeddings)

# Create retriever
retriever = vectorstore.as_retriever()

# Initialize LLM
llm = ChatOpenAI(model="gpt-4o", temperature=0, openai_api_key=os.environ["OPENAI_API_KEY"])

# Create retrieval chain
chain = create_retrieval_chain(llm=llm, retriever=retriever)

# Query the chain
query = "What is LangChain used for?"
result = chain.run(query)
print("Answer:", result)

output

Answer: LangChain is used to build applications that combine language models with external data sources, enabling retrieval-augmented generation and complex workflows.

Common variations

Use different LLMs like gpt-4o-mini or Anthropic's claude-3-5-sonnet-20241022 by swapping the LLM instance.
Use async calls by importing and using async versions of the LLM and chain.
Swap vectorstores like Chroma or Weaviate instead of FAISS.
Customize the chain by passing additional parameters like return_source_documents=True to get context documents.

Troubleshooting

If you get empty answers, verify your documents are loaded and embedded correctly.
Ensure your OpenAI API key is set in os.environ["OPENAI_API_KEY"].
Check that the vectorstore contains vectors by inspecting vectorstore.index.ntotal.
For slow responses, reduce max_tokens or use smaller models like gpt-4o-mini.

✅

Key Takeaways

Use create_retrieval_chain to combine an LLM and retriever for retrieval-augmented QA.
Pass a vectorstore retriever and an LLM instance to build the chain quickly.
Customize with different models, vectorstores, and chain parameters for your use case.

Verified 2026-04 · gpt-4o, gpt-4o-mini, claude-3-5-sonnet-20241022

Verify ↗