How to Intermediate · 3 min read

How to use ReActAgent in LlamaIndex

Q: How to use ReActAgent in LlamaIndex

Use ReActAgent in LlamaIndex by importing it from llama_index.agents.react and initializing it with a language model and a LLMRetriever. This agent combines reasoning and acting capabilities, allowing you to query your indexed documents interactively with reasoning steps.

Quick answer

Use ReActAgent in LlamaIndex by importing it from llama_index.agents.react and initializing it with a language model and a LLMRetriever. This agent combines reasoning and acting capabilities, allowing you to query your indexed documents interactively with reasoning steps.

PREREQUISITES

Python 3.8+
OpenAI API key (free tier works)
pip install llama-index openai>=1.0

Setup

Install the required packages and set your OpenAI API key as an environment variable.

bash

pip install llama-index openai>=1.0

Step by step

This example shows how to create a simple ReActAgent with LlamaIndex using the OpenAI gpt-4o model and a retriever from an index.

python

import os
from llama_index import GPTVectorStoreIndex, SimpleDirectoryReader, LLMRetriever
from llama_index.agents.react import ReActAgent
from openai import OpenAI

# Set up OpenAI client
client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

# Load documents from a directory
documents = SimpleDirectoryReader("./data").load_data()

# Create a vector store index
index = GPTVectorStoreIndex.from_documents(documents, client=client, model="gpt-4o")

# Create a retriever from the index
retriever = LLMRetriever(index=index, client=client, model="gpt-4o")

# Initialize the ReActAgent with the retriever and language model
agent = ReActAgent(
    retriever=retriever,
    llm=client,
    model="gpt-4o"
)

# Run a query with reasoning and actions
query = "Explain the main topics covered in the documents and provide references."
response = agent.query(query)
print(response)

output

Explain the main topics covered in the documents and provide references.

[Agent's detailed reasoning and references based on indexed documents]

Common variations

Use gpt-4o-mini or other OpenAI models for faster or cheaper responses.
Use async calls if integrating into async frameworks.
Customize the ReActAgent with different prompt templates or memory modules.

python

import asyncio

async def async_query():
    response = await agent.aquery("Summarize the documents asynchronously.")
    print(response)

asyncio.run(async_query())

output

[Agent's asynchronous response summarizing documents]

Troubleshooting

If you get authentication errors, verify your OPENAI_API_KEY environment variable is set correctly.
If the agent returns irrelevant answers, ensure your documents are properly loaded and indexed.
Check your model name matches current supported models like gpt-4o.

✅

Key Takeaways

Use ReActAgent with a retriever and OpenAI client for reasoning over indexed documents.
Initialize ReActAgent with current models like gpt-4o for best results.
Customize the agent with async calls or different prompt templates for flexibility.

Verified 2026-04 · gpt-4o

Verify ↗