UnicodeEncodeError when running OpenAI SDK test code in Jupyter Notebook

[taless474]

Confirm this is an issue with the Python library and not an underlying OpenAI API

• This is an issue with the Python library

Describe the bug

Running the official OpenAI SDK test code in a Jupyter Notebook causes the following error:

UnicodeEncodeError: 'ascii' codec can't encode character '\u2014' in position 110: ordinal not in range(128)

The test script appears to fail in Jupyter due to UnicodeEncodeError from an em dash \u2014, and Jupyter's sys.stdout is not reconfigurable. I think this makes the default SDK examples unusable in Jupyter by default.

To Reproduce

Launch a Jupyter Notebook from terminal

Create a new Python 3 notebook

Paste and run the following code cell

Code snippets

from openai import OpenAI
import base64

client = OpenAI()

response = client.responses.create(
    model="gpt-4.1-mini",
    input="Generate an image of gray tabby cat hugging an otter with an orange scarf",
    tools=[{"type": "image_generation"}],
)

I got it form this image generation example
The notebook crashes with a UnicodeEncodeError, and Jupyter’s I/O wrappers don’t allow easy reconfiguration of encoding.

OS

macOS

Python version

Python 3.13.4

Library version

openai 1.85.0

[RobertCraigie]

I don't think there's anything we can do here, if the API is returning unicode and your environment is configured to just support ascii then you'll need to update your environment to support unicode.

[taless474]

Thanks for the reply. I understand your point, but this issue occurs when using the latest Python and Jupyter on macOS, both of which support UTF-8 by default.

The official example you’re sharing throws encoding-related errors (UnicodeEncodeError) when running in a Jupyter notebook environment, which is a standard development tool for data science and ML workflows.

If the SDK can’t handle that cleanly or needs specific encoding configuration, I think that should be documented or built into the example code.

Would you be open to updating the example or adding a note in the docs so other users don’t run into the same confusion?

[taless474]

Update: The issue exists also in Google Colab

I’ve done additional testing in both Jupyter Notebook and Google Colab, and I now believe this is an issue with how the SDK handles Unicode in logging or internal errors.

Setup Details

OpenAI SDK version: 1.84.0
Python version: 3.11.13
Environment: Google Colab
Encoding: sys.stdout.encoding shows 'UTF-8'

Behavior Recap

from openai import OpenAI
client = OpenAI()
client.models.list()

crashes with

UnicodeEncodeError: 'ascii' codec can't encode character '\u2014' in position 110: ordinal not in range(128)

I would appreciate a workaround if you decide to close this issue.

[RobertCraigie]

Thanks for diving deeper, can you share a full stack trace?

[taless474]

Thanks! Here's the minimal code that triggers the issue in Google Colab:

import os
from openai import OpenAI

os.environ["OPENAI_API_KEY"] = "sk-..."  # My real key was used here although should not be on colab
client = OpenAI()
response = client.models.list()

full stack trace from Google Colab:

---------------------------------------------------------------------------
UnicodeEncodeError                        Traceback (most recent call last)
[<ipython-input-2-4052322613>](https://localhost:8080/#) in <cell line: 0>()
      4 os.environ["OPENAI_API_KEY"] = "sk-..."
      5 client = OpenAI()
----> 6 response = client.models.list()

7 frames
[/usr/local/lib/python3.11/dist-packages/httpx/_models.py](https://localhost:8080/#) in _normalize_header_value(value, encoding)
     80     if not isinstance(value, str):
     81         raise TypeError(f"Header value must be str or bytes, not {type(value)}")
---> 82     return value.encode(encoding or "ascii")
     83 
     84 

UnicodeEncodeError: 'ascii' codec can't encode character '\u2014' in position 110: ordinal not in range(128)

[RobertCraigie]

Thanks, could you also enable debug logging? https://github.com/openai/openai-python?tab=readme-ov-file#logging.

It looks like one of the headers that's getting set is non-ascii.

[taless474]

Sure thing!

This is the code I tried:

import os
import logging
import openai
from openai import OpenAI

openai.log = "debug"

logging.basicConfig(level=logging.DEBUG)

os.environ["OPENAI_API_KEY"] = "sk-..." 

client = OpenAI()  

response = client.models.list()

print(response)

Attempt 1 - Google Colab

I didn't get the debugging information

UnicodeEncodeError                        Traceback (most recent call last)
[<ipython-input-1-2546941339>](https://localhost:8080/#) in <cell line: 0>()
     12 client = OpenAI()
     13 
---> 14 response = client.models.list()
     15 
     16 print(response)

7 frames
[/usr/local/lib/python3.11/dist-packages/httpx/_models.py](https://localhost:8080/#) in _normalize_header_value(value, encoding)
     80     if not isinstance(value, str):
     81         raise TypeError(f"Header value must be str or bytes, not {type(value)}")
---> 82     return value.encode(encoding or "ascii")
     83 
     84 

UnicodeEncodeError: 'ascii' codec can't encode character '\u2014' in position 110: ordinal not in range(128)

Attempt 2 - Terminal (zsh, Python 3.13.4, OpenAI 1.86.0)

so I ran it as a python code in terminal and got this:

DEBUG:openai._base_client:Request options: {'method': 'get', 'url': '/models', 'post_parser': <function SyncAPIClient._request_api_list.<locals>._parser at 0x107021c60>, 'json_data': None}
Traceback (most recent call last):
  File "/Users/Unick/Desktop/Coursera/testUnicode2.py", line 14, in <module>
    response = client.models.list()
  File "/Users/Unick/Desktop/Coursera/openai-env/lib/python3.13/site-packages/openai/resources/models.py", line 91, in list
    return self._get_api_list(
           ~~~~~~~~~~~~~~~~~~^
        "/models",
        ^^^^^^^^^^
    ...<4 lines>...
        model=Model,
        ^^^^^^^^^^^^
    )
    ^
  File "/Users/Unick/Desktop/Coursera/openai-env/lib/python3.13/site-packages/openai/_base_client.py", line 1291, in get_api_list
    return self._request_api_list(model, page, opts)
           ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^
  File "/Users/Unick/Desktop/Coursera/openai-env/lib/python3.13/site-packages/openai/_base_client.py", line 1142, in _request_api_list
    return self.request(page, options, stream=False)
           ~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/Unick/Desktop/Coursera/openai-env/lib/python3.13/site-packages/openai/_base_client.py", line 958, in request
    request = self._build_request(options, retries_taken=retries_taken)
  File "/Users/Unick/Desktop/Coursera/openai-env/lib/python3.13/site-packages/openai/_base_client.py", line 495, in _build_request
    headers = self._build_headers(options, retries_taken=retries_taken)
  File "/Users/Unick/Desktop/Coursera/openai-env/lib/python3.13/site-packages/openai/_base_client.py", line 439, in _build_headers
    headers = httpx.Headers(headers_dict)
  File "/Users/Unick/Desktop/Coursera/openai-env/lib/python3.13/site-packages/httpx/_models.py", line 156, in __init__
    bytes_value = _normalize_header_value(v, encoding)
  File "/Users/Unick/Desktop/Coursera/openai-env/lib/python3.13/site-packages/httpx/_models.py", line 82, in _normalize_header_value
    return value.encode(encoding or "ascii")
           ~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^
UnicodeEncodeError: 'ascii' codec can't encode character '\u2014' in position 110: ordinal not in range(128)

Please let me know if I could help with anything.

[taless474]

Update

Just to add more context:

I’m seeing the UnicodeEncodeError even when running from a clean virtual environment (Python 3.11.13, macOS, OpenAI SDK 1.86.0) with UTF-8 properly set via locale and PYTHONIOENCODING.

The error seems to happen before it reaches my try/except block, so I suspect it might be related to how the SDK formats or logs errors internally, possibly involving a non-ASCII character like an em dash (\u2014).

I also tried redirecting output to a file and using logging/debug mode, but the error still occurs.

Hope this helps pinpoint the issue. Let me know if I can test anything further.