TypeError: issubclass() arg 1 must be a class

[aardvarkk]

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

When calling OpenAI.beta.chat.completions.parse in a highly concurrent environment, and providing a class as the response_format, I get the following error:

    response = self.client.beta.chat.completions.parse(
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local_disk0/.ephemeral_nfs/cluster_libraries/python/lib/python3.11/site-packages/openai/resources/beta/chat/completions.py", line 156, in parse
    return self._post(
           ^^^^^^^^^^^
  File "/local_disk0/.ephemeral_nfs/cluster_libraries/python/lib/python3.11/site-packages/openai/_base_client.py", line 1280, in post
    return cast(ResponseT, self.request(cast_to, opts, stream=stream, stream_cls=stream_cls))
                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local_disk0/.ephemeral_nfs/cluster_libraries/python/lib/python3.11/site-packages/openai/_base_client.py", line 957, in request
    return self._request(
           ^^^^^^^^^^^^^^
  File "/local_disk0/.ephemeral_nfs/cluster_libraries/python/lib/python3.11/site-packages/openai/_base_client.py", line 1063, in _request
    return self._process_response(
           ^^^^^^^^^^^^^^^^^^^^^^^
  File "/local_disk0/.ephemeral_nfs/cluster_libraries/python/lib/python3.11/site-packages/openai/_base_client.py", line 1162, in _process_response
    return api_response.parse()
           ^^^^^^^^^^^^^^^^^^^^
  File "/local_disk0/.ephemeral_nfs/cluster_libraries/python/lib/python3.11/site-packages/openai/_response.py", line 319, in parse
    parsed = self._options.post_parser(parsed)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local_disk0/.ephemeral_nfs/cluster_libraries/python/lib/python3.11/site-packages/openai/resources/beta/chat/completions.py", line 150, in parser
    return _parse_chat_completion(
           ^^^^^^^^^^^^^^^^^^^^^^^
  File "/local_disk0/.ephemeral_nfs/cluster_libraries/python/lib/python3.11/site-packages/openai/lib/_parsing/_completions.py", line 122, in parse_chat_completion
    construct_type_unchecked(
  File "/local_disk0/.ephemeral_nfs/cluster_libraries/python/lib/python3.11/site-packages/openai/_models.py", line 445, in construct_type_unchecked
    return cast(_T, construct_type(value=value, type_=type_))
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local_disk0/.ephemeral_nfs/cluster_libraries/python/lib/python3.11/site-packages/openai/_models.py", line 519, in construct_type
    return type_.construct(**value)  # type: ignore[arg-type]
           ^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local_disk0/.ephemeral_nfs/cluster_libraries/python/lib/python3.11/site-packages/openai/_models.py", line 230, in construct
    fields_values[name] = _construct_field(value=values[key], field=field, key=key)
                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local_disk0/.ephemeral_nfs/cluster_libraries/python/lib/python3.11/site-packages/openai/_models.py", line 394, in _construct_field
    return construct_type(value=value, type_=type_)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/local_disk0/.ephemeral_nfs/cluster_libraries/python/lib/python3.11/site-packages/openai/_models.py", line 513, in construct_type
    if not is_literal_type(type_) and (issubclass(origin, BaseModel) or issubclass(origin, GenericModel)):
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<frozen abc>", line 123, in __subclasscheck__
TypeError: issubclass() arg 1 must be a class

This is being run in a multi-threaded environment. When I run with a single thread, I don't see this issue. I'm trying to churn through a bunch of data, so I'm attempting to use about 100 threads to make these API requests in parallel. If I reduce the count to 1 the problem goes away.

To work around this, I believe I can stop using the beta parse method with the provided response_format,

To Reproduce

1. Call OpenAI.beta.chat.completions.parse with 100 threads simultaneously

Code snippets

class Bar(Enum):
    C = "C"

class Qux(Enum):
    D = "D"

class Foo(BaseModel):
    a: Bar
    b: Qux

def main(): # Called by 100 threads concurrently
    OpenAI(api_key="...").beta.chat.completions.parse(
        model="gpt-4o-mini",
        seed=0,
        temperature=0,
        messages=[
            {"role": "system", "content": "..."},
            {"role": "user", "content": "..."},
        ],
        response_format=Foo,
    )

OS

Ubuntu 22.04.4 LTS

Python version

Python v3.11.10

Library version

openai v1.57.2

[Joseelmax-00]

I can confirm I'm having the same issue which is making it impossible for me to use o1-preview. Tried changing versions of pydantic, typing_extensions and other libraries to no avail. Python 3.11.0 and this environment:

Package Version

---

annotated-types 0.7.0
anyio 4.7.0
certifi 2024.12.14
colorama 0.4.6
distro 1.9.0
h11 0.14.0
httpcore 1.0.7
httpx 0.28.1
idna 3.10
jiter 0.8.2
mypy-extensions 1.0.0
openai 1.58.1
pip 24.3.1
pydantic 2.10.4
pydantic_core 2.27.2
python-dotenv 1.0.1
setuptools 65.5.0
sniffio 1.3.1
tqdm 4.67.1
typing_extensions 4.12.2
typing-inspect 0.9.0

File "D:\Trabajo\random2\game of Quatro\openai_player.py", line 172, in select_AI_move
    completion = client.beta.chat.completions.parse(
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\Trabajo\random2\game of Quatro\env\Lib\site-packages\openai\resources\beta\chat\completions.py", line 181, in parse
    "response_format": _type_to_response_format(response_format),
                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\Trabajo\random2\game of Quatro\env\Lib\site-packages\openai\lib\_parsing\_completions.py", line 248, in type_to_response_format_param
    if is_basemodel_type(response_format):
       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\Trabajo\random2\game of Quatro\env\Lib\site-packages\openai\lib\_pydantic.py", line 130, in is_basemodel_type
    return issubclass(typ, pydantic.BaseModel)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<frozen abc>", line 123, in __subclasscheck__
TypeError: issubclass() arg 1 must be a class

[RobertCraigie]

Thanks for the bug report, I haven't been able to reproduce this issue but it should be fixed in the next release as I've added some more inspect.isclass() checks to the places referenced in the stack traces.

#1987