langchain-ai/langchain · error · TypeError

Invalid args_schema: expected BaseModel or dict, got {args_s

Error message

Invalid args_schema: expected BaseModel or dict, got {args_schema}

What it means

StructuredTool.from_function accepts args_schema only as a pydantic BaseModel class or a JSON-schema-ish dict. Any other type (a TypedDict, an instance instead of a class, a string, a dataclass) raises TypeError with the offending value.

Source

Thrown at libs/core/langchain_core/tools/structured.py:232

            description_ = source_function.__doc__ or None
        if description_ is None and args_schema:
            if isinstance(args_schema, type) and is_basemodel_subclass(args_schema):
                description_ = args_schema.__doc__
                if (
                    description_
                    and "A base class for creating Pydantic models" in description_
                ):
                    description_ = ""
                elif not description_:
                    description_ = None
            elif isinstance(args_schema, dict):
                description_ = args_schema.get("description")
            else:
                msg = (
                    "Invalid args_schema: expected BaseModel or dict, "
                    f"got {args_schema}"
                )
                raise TypeError(msg)
        if description_ is None:
            msg = "Function must have a docstring if description not provided."
            raise ValueError(msg)
        if description is None:
            # Only apply if using the function's docstring
            description_ = textwrap.dedent(description_).strip()

        # Description example:
        # search_api(query: str) - Searches the API for the query.
        description_ = f"{description_.strip()}"
        return cls(
            name=name,
            func=func,
            coroutine=coroutine,
            args_schema=args_schema,
            description=description_,
            return_direct=return_direct,
            response_format=response_format,

View on GitHub (pinned to e32fa9a52e)

Solutions

1. Pass a pydantic BaseModel subclass: args_schema=MyArgs (class, not instance)
2. Or pass a dict schema: args_schema={'title': 'MyArgs', 'type': 'object', 'properties': {...}, 'required': [...]}
3. Convert TypedDicts: args_schema=create_model from the TypedDict's annotations, or redeclare as a pydantic model

Example fix

# before
class MyArgs(TypedDict):
    query: str

tool = StructuredTool.from_function(fn, args_schema=MyArgs)  # TypeError

# after
from pydantic import BaseModel, Field

class MyArgs(BaseModel):
    query: str = Field(description="Search query")

tool = StructuredTool.from_function(fn, args_schema=MyArgs)

from pydantic import BaseModel

def is_valid_args_schema(schema: object) -> bool:
    return (
        (isinstance(schema, type) and issubclass(schema, BaseModel))
        or isinstance(schema, dict)
    )

# use before construction:
assert is_valid_args_schema(args_schema), f"bad args_schema: {args_schema!r}"

from pydantic import BaseModel

def is_pydantic_schema_or_dict(x: object) -> bool:
    if isinstance(x, type):
        return issubclass(x, BaseModel)
    return isinstance(x, dict)

try:
    t = StructuredTool.from_function(fn, args_schema=schema, name="t")
except TypeError as e:
    if "Invalid args_schema" in str(e):
        from pydantic import create_model
        fields = {k: (v, ...) for k, v in schema.__annotations__.items()}  # TypedDict case
        t = StructuredTool.from_function(
            fn, args_schema=create_model("t_args", **fields), name="t"
        )
    else:
        raise

When it happens

Trigger: StructuredTool.from_function(fn, args_schema=MyTypedDict); passing an instantiated model args_schema=MyModel(...); passing a JSON schema string.

Common situations: Teams using TypedDicts for tool schemas (works with @tool type inference, not here); passing schema instances rather than classes; migrating schemas from JSON strings.

Related errors

• Runnable must have an object schema.
• Tool input must be str or dict. If dict, dict arguments must
• Too many arguments to single-input tool {self.name}.
• Expected a Pydantic model. Got {model}
• Function must have either a docstring or description when in

AI-assisted analysis of langchain-ai/langchain@e32fa9a52e (2026-08-14). Data as JSON: /api/errors/b395d2f08b2c7de0. Report an issue: GitHub.