AttributeError: module 'openai' has no attribute 'DefaultHttpxClient' when running LangGraph example in Chainlit backend

[sh-js]

I encountered an error when trying to run the LangGraph example inside the Chainlit backend. Specifically, I cloned the Chainlit repository and added the LangGraph example code in backend/chainlit/hello.py.

However, when I ran the code, I got the following error:

from langchain_openai.chat_models import AzureChatOpenAI, ChatOpenAI
  File "/home/chaimaa/.cache/pypoetry/virtualenvs/chatbot-backend-a0QcPf-j-py3.12/lib/python3.12/site-packages/langchain_openai/chat_models/__init__.py", line 1, in <module>
    from langchain_openai.chat_models.azure import AzureChatOpenAI
  File "/home/chaimaa/.cache/pypoetry/virtualenvs/chatbot-backend-a0QcPf-j-py3.12/lib/python3.12/site-packages/langchain_openai/chat_models/azure.py", line 21, in <module>
    from langchain_openai.chat_models.base import BaseChatOpenAI
  File "/home/chaimaa/.cache/pypoetry/virtualenvs/chatbot-backend-a0QcPf-j-py3.12/lib/python3.12/site-packages/langchain_openai/chat_models/base.py", line 105, in <module>
    from langchain_openai.chat_models._client_utils import (
  File "/home/chaimaa/.cache/pypoetry/virtualenvs/chatbot-backend-a0QcPf-j-py3.12/lib/python3.12/site-packages/langchain_openai/chat_models/_client_utils.py", line 17, in <module>
    class _SyncHttpxClientWrapper(openai.DefaultHttpxClient):
                                  ^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: module 'openai' has no attribute 'DefaultHttpxClient'

Steps to Reproduce:

Clone the Chainlit repo.

Add the LangGraph example ( https://docs.chainlit.io/integrations/langchain#with-langgraph ) code to backend/chainlit/hello.py.

Run the backend.

Environment:

Python version: 3.12

Poetry-managed virtualenv

Latest versions of chainlit(2.6.0), langchain (0.3.26), langchain_openai (0.3.27), and openai (1.93.3)

[tituslhy]

It's an import error faced on LangChain's side - you're basically failing to even import the ChatOpenAI model. I tried and didn't receive the same failure message. I've this feeling it's because you're running on a default kernel.

If you're using Mac and assuming you've already activated your .venv - run which python to see if it's calling your actual python venv environment's dependencies. If not, then just run unalias python and it'll point to the correct environment.

Here's proof that I didn't face the error. I navigated to the source script of the issue in my .venv folder:
.venv / libs / langchain_openai / chat_models / _client_utils.py and added a "#%%" to the top of the script so that I can execute it in Jupyter mode.

When you click "run cell" it opens a Jupyter window and runs the entire script. I faced no errors. Good luck!