Repository navigation
Feature: support Pydantic Model as function param and return type #1066
Description
** Please make sure you read the contribution guide and file the issues in the right place. **
Contribution guide.
Is your feature request related to a problem? Please describe.
When defining function tool using python function, we need to allow simple pyndatic model as params and return types.
E.g.
class Param1(BaseModel);
id: str
query: str
class Param2(BaseModel);
id: str
log: str
class ReturnType(BaseModel):
result: str
def my_tool(param_1: Param1, param_2: Param2) -> ReturnType:
...
Describe the solution you'd like
- Supporting to convert PydanticModel to json schema then to FunctionDeclaration for the usage of GenAI SDK.
Describe alternatives you've considered
N/A
Additional context
N/A
Reacted by Illia Kaliuzhnyi, Michael Jones, Donghyuk Kang, mwoods-precis, Kevin O'Hara, Nicolas Ramy and Baptiste-ms
Activity
hangfei commentedon May 30, 2025
@Jacksunwei is driving this work.
If you have any requirements or thoughts, please share with him. Thanks!
mikeedjones commentedon Sep 3, 2025
Ideally it would work with complex pydantic models! It would also be good to be able to mark some of the arguments as to be excluded from the LM-generation like FastMCP's exclude_args.
lucasnobre212 commentedon Feb 13, 2026
@Jacksunwei Any updates on this?
armude commentedon Apr 20, 2026
Are you going to add this anytime soon?
xuanyang15 commentedon Jun 26, 2026
This has been supported with the feature JSON_SCHEMA_FOR_FUNC_DECL being enabled by default.