LangChain pydantic v1 vs v2 compatibility error fix
pydantic v2, causing compatibility errors if your code uses pydantic v1 syntax or imports. Fix this by updating your imports to use langchain_core and adjusting model initialization to the new pydantic v2 style, such as replacing BaseModel.Config with model_config.code_error langchain_core and refactor your pydantic models to use model_config instead of Config for compatibility with pydantic v2.Why this happens
LangChain v0.2+ migrated from pydantic v1 to pydantic v2, which introduced breaking changes in model configuration and validation. Legacy code using from pydantic import BaseModel with class Config inside models or relying on old LangChain import paths (from langchain.llms import OpenAI) will raise errors like TypeError or AttributeError.
Typical error output includes:
TypeError: __init__() got an unexpected keyword argument 'arbitrary_types_allowed'
or
AttributeError: module 'langchain.llms' has no attribute 'OpenAI'
This happens because pydantic v2 replaced Config with model_config and LangChain reorganized its package structure.
from langchain.llms import OpenAI
class MyModel(BaseModel):
name: str
class Config:
arbitrary_types_allowed = True
model = MyModel(name='test')AttributeError: module 'langchain.llms' has no attribute 'OpenAI'
The fix
Update your imports to the new LangChain v0.2+ structure and refactor pydantic models to use model_config. Replace:
from langchain.llms import OpenAIwithfrom langchain_openai import ChatOpenAI- Inside pydantic models, replace
class Configwithmodel_config = { ... }
This aligns with pydantic v2 expectations and LangChain's new modular design.
from langchain_openai import ChatOpenAI
from pydantic import BaseModel
class MyModel(BaseModel):
name: str
model_config = {
"arbitrary_types_allowed": True
}
model = MyModel(name='test')
llm = ChatOpenAI(model="gpt-4o", temperature=0)
print(llm)<langchain_openai.chat.ChatOpenAI object at 0x7f8c2b3d1d60>
Preventing it in production
Use dependency management tools like pip-tools or poetry to lock LangChain and pydantic versions together. Validate your environment with pip freeze to ensure pydantic>=2.0 and LangChain v0.2+ are installed. Add automated tests that instantiate your models and LangChain clients to catch compatibility issues early. Consider adding retry logic around API calls to handle transient errors gracefully.
Key Takeaways
- LangChain v0.2+ requires pydantic v2-compatible model definitions using model_config.
- Update imports from deprecated langchain.llms to langchain_openai and other new packages.
- Lock your dependencies and test model instantiation to catch compatibility issues early.