How can I fix the 'bind_tools' AttributeError when creating a React Agent with LangGraph?

Below is an explanation and a solution based on the LangGraph React agent documentation.

---

Issue Explanation

The error

AttributeError: 'OpenAI' object has no attribute 'bind_tools'

indicates that the LLM instance you’re passing (from langchain_openai’s OpenAI class) does not support the .bind_tools method. The create_react_agent function expects an LLM that supports tool binding, as it internally calls .bind_tools to associate your tool(s) with the model.

According to the documentation, you should pass in a model that has the capability to bind tools (typically from the Chat-based models provided in LangChain). This is why you also see deprecation warnings if using outdated agent methods.

---

How to Fix the Error

Solution: Use a chat-based model that supports .bind_tools. For example, replace your import and instantiation of the OpenAI model with the ChatOpenAI model from the appropriate module.

Here is an updated version of your code:

PYTHONCopied
import os
from dotenv import load_dotenv
from langchain.chat_models import ChatOpenAI  # Use ChatOpenAI instead of OpenAI
from langchain_community.utilities import GoogleSerperAPIWrapper
from langgraph.prebuilt import create_react_agent
from langchain.agents import Tool  # Ensure you import Tool from the correct module

load_dotenv()

openai_api_key = os.getenv("OPENAI_API_KEY")
serper_api_key = os.getenv("SERPER_API_KEY")

# Instantiate a chat model that supports bind_tools
llm = ChatOpenAI(temperature=0, openai_api_key=openai_api_key, model="gpt-4")

search = GoogleSerperAPIWrapper()

tools = [
    Tool(
        name="Intermediate Answer",
        func=search.run,
        description="Useful for when you need to ask with search"
    )
]

# Create the agent; create_react_agent will automatically bind tools to the model
agent_executor = create_react_agent(llm, tools)

inputs = {"messages": [("user", "what is the hometown of the reigning men's U.S. Open champion?")]}
response = agent_executor.stream(inputs, stream_mode="values")

print(response)

---

Key Points

• Switch to ChatOpenAI: The ChatOpenAI class (from langchain.chat_models) supports the .bind_tools method, which is required by create_react_agent.
• Tool Binding: When you pass the model (e.g., ChatOpenAI) into create_react_agent, it internally calls .bind_tools so you don’t have to wrap the model manually.
• Deprecation Warnings: If you see warnings related to using deprecated agents, consider consulting the latest LangChain documentation and migrating to the recommended interfaces.

This change should resolve the bind_tools error and allow you to successfully create a React agent using LangGraph.