When using convert_to_openai_function, if a type hint of field is an Enum, its description will be lost.

[gbaian10]

Checked other resources

• I added a very descriptive title to this issue.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.
• I am sure that this is a bug in LangChain rather than my code.
• The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).

Example Code

from enum import StrEnum

from langchain_core.utils.function_calling import convert_to_openai_function
from pydantic import BaseModel, Field


class WithDesciptionModel(BaseModel):
    foo: str = Field(..., description="This is a foo field")


class MyEnum(StrEnum):
    NAME = "name"
    AGE = "age"


class WithoutDesciptionModel(BaseModel):
    foo: MyEnum = Field(..., description="This is a foo field")


# The description of foo will be kept.
print(convert_to_openai_function(WithDesciptionModel))
# The description of foo will be lost.
print(convert_to_openai_function(WithoutDesciptionModel))


# print(WithDesciptionModel.model_json_schema())
# print(WithoutDesciptionModel.model_json_schema())

Error Message and Stack Trace (if applicable)

{'name': 'WithDesciptionModel', 'description': '', 'parameters': {'properties': {'foo': {'description': 'This is a foo field', 'type': 'string'}}, 'required': ['foo'], 'type': 'object'}}
{'name': 'WithoutDesciptionModel', 'description': '', 'parameters': {'properties': {'foo': {'enum': ['name', 'age'], 'type': 'string'}}, 'required': ['foo'], 'type': 'object'}}

Description

If the type hint in a BaseModel includes an Enum, its description will be lost.

System Info

langchain-core==0.3.5
pydantic==2.9.2

python-version==3.12.6

[eyurtsev]

@gbaian10 do you use this together with the open ai chat model? did this arise while using with_structured_output

[gbaian10]

@eyurtsev Using with_structured_output also has this issue.
Although it's currently known that modifying the __doc__ of the Enum class itself can solve it, this doesn't seem like a good approach.

from enum import StrEnum, auto

from dotenv import load_dotenv
from langchain_openai import ChatOpenAI
from pydantic import BaseModel, Field

load_dotenv()


class NameEnum(StrEnum):
    ALICE = auto()
    BOB = auto()


class NewNameEnum(StrEnum):
    CHARLIE = auto()
    DAVID = auto()


description = "Name of the person"
NewNameEnum.__doc__ = description


class EnumModel(BaseModel):
    name: NameEnum = Field(..., description=description)


class StrModel(BaseModel):
    name: str = Field(..., description=description)


class FixEnumModel(BaseModel):
    name: NewNameEnum


inputs = "Hello, I am Alice."
model = ChatOpenAI(model="gpt-4o-mini")
model.with_structured_output(EnumModel).invoke(inputs)
model.with_structured_output(StrModel).invoke(inputs)
model.with_structured_output(FixEnumModel).invoke(inputs)