Pydantic fields with an 'alias' do not get passed to LLM silently

[YashalShakti]

Checked other resources

• This is a bug, not a usage question. For questions, please use GitHub Discussions.
• I added a clear and detailed title that summarizes the issue.
• I read what a minimal reproducible example is (https://stackoverflow.com/help/minimal-reproducible-example).
• I included a self-contained, minimal example that demonstrates the issue INCLUDING all the relevant imports. The code run AS IS to reproduce the issue.

Example Code

from typing import Optional, Type

from langchain.callbacks.manager import (
    AsyncCallbackManagerForToolRun,
    CallbackManagerForToolRun,
)


class SearchInput(BaseModel):
    query: str = Field(description="should be a search query")


class CalculatorInput(BaseModel):
    a: int = Field(description="first number", alias="A")
    b: int = Field(description="second number")


class CustomSearchTool(BaseTool):
    name = "custom_search"
    description = "useful for when you need to answer questions about current events"
    args_schema: Type[BaseModel] = SearchInput

    def _run(
        self, query: str, run_manager: Optional[CallbackManagerForToolRun] = None
    ) -> str:
        """Use the tool."""
        return "LangChain"

    async def _arun(
        self, query: str, run_manager: Optional[AsyncCallbackManagerForToolRun] = None
    ) -> str:
        """Use the tool asynchronously."""
        raise NotImplementedError("custom_search does not support async")


class CustomCalculatorTool(BaseTool):
    name = "Calculator"
    description = "useful for when you need to answer questions about math"
    args_schema: Type[BaseModel] = CalculatorInput
    return_direct: bool = True

    def _run(
        self, a: int, b: int, run_manager: Optional[CallbackManagerForToolRun] = None
    ) -> str:
        """Use the tool."""
        return a * b

    async def _arun(
        self,
        a: int,
        b: int,
        run_manager: Optional[AsyncCallbackManagerForToolRun] = None,
    ) -> str:
        """Use the tool asynchronously."""
        raise NotImplementedError("Calculator does not support async")

Error Message and Stack Trace (if applicable)

Description

Fields with an alias do get passed to the LLM silently. This was extremely tough to debug as we were using alias_generators.to_camel and some fields were getting passed were other were not.

https://docs.pydantic.dev/latest/api/config/#pydantic.alias_generators.to_camel

Even if this is not meant to be supported, it should not be silent.

System Info

langchain-anthropic;: Installed. No version info available.
langchain-aws;: Installed. No version info available.
langchain-azure-ai;: Installed. No version info available.
langchain-cohere;: Installed. No version info available.
langchain-community;: Installed. No version info available.
langchain-core<0.4,>=0.3.36: Installed. No version info available.
langchain-core<0.4.0,>=0.3.40: Installed. No version info available.
langchain-core<1.0.0,>=0.3.45: Installed. No version info available.
langchain-core<1.0.0,>=0.3.49: Installed. No version info available.
langchain-core<1.0.0,>=0.3.51: Installed. No version info available.
langchain-core<1.0.0,>=0.3.55: Installed. No version info available.
langchain-deepseek;: Installed. No version info available.
langchain-fireworks;: Installed. No version info available.
langchain-google-genai;: Installed. No version info available.
langchain-google-vertexai;: Installed. No version info available.
langchain-groq;: Installed. No version info available.
langchain-huggingface;: Installed. No version info available.
langchain-mistralai;: Installed. No version info available.
langchain-ollama;: Installed. No version info available.
langchain-openai;: Installed. No version info available.
langchain-perplexity;: Installed. No version info available.
langchain-text-splitters<1.0.0,>=0.3.8: Installed. No version info available.
langchain-together;: Installed. No version info available.
langchain-xai;: Installed. No version info available.
langchain<1.0.0,>=0.3.21: Installed. No version info available.
langgraph: 0.3.34
langgraph-checkpoint: 2.0.23
langgraph-prebuilt<0.2.0,>=0.1.7: Installed. No version info available.
langgraph<0.4.0,>=0.3.5: Installed. No version info available.
langserve[all]>=0.0.51: Installed. No version info available.
langsmith-pyo3: Installed. No version info available.
langsmith<0.4,>=0.1.125: Installed. No version info available.
langsmith<0.4,>=0.1.17: Installed. No version info available.

[hinthornw]

Seems unrelated to LangGraph so moving to the LangChain repo.

[hinthornw]

Confirmed / can reproduce with this:

from typing import Optional, Type
from langchain_core.tools import BaseTool
from pydantic import BaseModel, Field

from langchain.callbacks.manager import (
    AsyncCallbackManagerForToolRun,
    CallbackManagerForToolRun,
)


class SearchInput(BaseModel):
    query: str = Field(description="should be a search query")


class CalculatorInput(BaseModel):
    a: int = Field(description="first number", alias="A")
    b: int = Field(description="second number")


class CustomSearchTool(BaseTool):
    name: str = "custom_search"
    description: str = (
        "useful for when you need to answer questions about current events"
    )
    args_schema: type[BaseModel] = SearchInput

    def _run(
        self, query: str, run_manager: Optional[CallbackManagerForToolRun] = None
    ) -> str:
        """Use the tool."""
        return "LangChain"

    async def _arun(
        self, query: str, run_manager: Optional[AsyncCallbackManagerForToolRun] = None
    ) -> str:
        """Use the tool asynchronously."""
        raise NotImplementedError("custom_search does not support async")


class CustomCalculatorTool(BaseTool):
    name: str = "Calculator"
    description: str = "useful for when you need to answer questions about math"
    args_schema: type[BaseModel] = CalculatorInput
    return_direct: bool = True

    def _run(
        self, a: int, b: int, run_manager: Optional[CallbackManagerForToolRun] = None
    ) -> str:
        """Use the tool."""
        return a * b

    async def _arun(
        self,
        a: int,
        b: int,
        run_manager: Optional[AsyncCallbackManagerForToolRun] = None,
    ) -> str:
        """Use the tool asynchronously."""
        raise NotImplementedError("Calculator does not support async")


tools = [CustomCalculatorTool(), CustomSearchTool()]
from langchain_openai import ChatOpenAI
from pprint import pprint

model = ChatOpenAI(model="gpt-4o")
schemas = model.bind_tools(tools).kwargs["tools"]
pprint(schemas)

[{'function': {'description': 'useful for when you need to answer questions '
                              'about math',
               'name': 'Calculator',
               'parameters': {'properties': {'b': {'description': 'second '
                                                                  'number',
                                                   'type': 'integer'}},
                              'required': ['b'],
                              'type': 'object'}},
  'type': 'function'},
 {'function': {'description': 'useful for when you need to answer questions '
                              'about current events',
               'name': 'custom_search',
               'parameters': {'properties': {'query': {'description': 'should '
                                                                      'be a '
                                                                      'search '
                                                                      'query',
                                                       'type': 'string'}},
                              'required': ['query'],
                              'type': 'object'}},
  'type': 'function'}]

The field should not be dropped

[YashalShakti]

It would be good to add a Warning to https://python.langchain.com/docs/how_to/custom_tools/ meanwhile

[CtrlMj]

The bug stems from line1087:

and name not in fields:
                continue

in the get_all_base_model_annotations method:

def get_all_basemodel_annotations(
    cls: Union[TypeBaseModel, Any], *, default_to_bound: bool = True
) -> dict[str, type]:
    """Get all annotations from a Pydantic BaseModel and its parents.

    Args:
        cls: The Pydantic BaseModel class.
        default_to_bound: Whether to default to the bound of a TypeVar if it exists.
    """
    # cls has no subscript: cls = FooBar
    if isinstance(cls, type):
        annotations: dict[str, type] = {}
        for name, param in inspect.signature(cls).parameters.items():
            # Exclude hidden init args added by pydantic Config. For example if
            # BaseModel(extra="allow") then "extra_data" will part of init sig.
            if (
                fields := getattr(cls, "model_fields", {})  # pydantic v2+
                or getattr(cls, "__fields__", {})  # pydantic v1
            ) and name not in fields:
                continue
            annotations[name] = param.annotation
        orig_bases: tuple = getattr(cls, "__orig_bases__", ())

In this case, fields is the following dictionary:

fields = {
    'a': FieldInfo(annotation=int, required=True, alias='A', alias_priority=2, description='first number'), 
    'b': FieldInfo(annotation=int, required=True, description='second number')
}

However, name in line 1087 is the Alias A which is not the key in the dictionary for the first parameter.
This is because in line 1081 inspect.signature(cls).parameters returns an orderedDictionary:

mappingproxy(OrderedDict([('A', <Parameter "A: int">), ('b', <Parameter "b: int">)]))

For fields with alias like above, inspect sets the alias as the key in the dictionary and skips over the original parameter name.
Therefore I think it's related to the inspect package. I can open an issue there.
However, I can implement a workaround in the meanwhile if we'd prefer that.

One workaround could be to have a dictionary that maps the aliases if they exists to the original variable name and use that in the for loop:

def get_all_basemodel_annotations(
    cls: Union[TypeBaseModel, Any], *, default_to_bound: bool = True
) -> dict[str, type]:
    """Get all annotations from a Pydantic BaseModel and its parents.

    Args:
        cls: The Pydantic BaseModel class.
        default_to_bound: Whether to default to the bound of a TypeVar if it exists.
    """
    # cls has no subscript: cls = FooBar
    if isinstance(cls, type):
        annotations: dict[str, type] = {}
        fields = getattr(cls, "model_fields", {}) or getattr(cls, "__fields__", {}) # pydantic v2+ or pydantic v1
        # create map from aliases to the original variable name (key)
        field_names_map = {} 
        for key, field in fields.items():
            if field.alias:
                field_names_map[field.alias] = key
            else:
                field_names_map[key] = key

        for name, param in inspect.signature(cls).parameters.items():
            # Exclude hidden init args added by pydantic Config. For example if
            # BaseModel(extra="allow") then "extra_data" will part of init sig.
            if field_names_map[name] not in fields: # using the {alias: key} map to check for fields
                continue
            annotations[field_names_map[name]] = param.annotation # using the {alias: key} map to populate annotations

cc: @ccurme

edit1: Added proposed workaround. I have tested and it works
edit2: typo

[ccurme]

@CtrlMj want to PR your workaround? :)

[keenborder786]

@ccurme I have already patched a fix

[CtrlMj]

@ccurme please find it here