Dataclass Support

Throughout the docs, we use Pydantic models in virtually all of our Python examples for validating task inputs and outputs. This is the recommended path, as it provides lots of safety guarantees as you're writing tasks. With that said, Hatchet also supports using dataclasses as both input and output types to tasks. Dataclass support was added in SDK version 1.21.0.

Usage

To use a dataclass instead of a Pydantic model, you'll need to:

1. Provide an input_validator as a parameter to your workflow or task (in the case of a standalone task with hatchet.task).
2. Add return type hints for your tasks.

Example Usage

dataclass validators work exactly like Pydantic models in Hatchet. First, you create the classes:

@dataclassclass Input:    name: str@dataclassclass Output:    message: str

@dataclass
class Input:
    name: str


@dataclass
class Output:
    message: str

And then you provide the classes to your workflow or task:

@hatchet.task(input_validator=Input)def say_hello(input: Input, ctx: Context) -> Output:    return Output(message=f"Hello, {input.name}!")

@hatchet.task(input_validator=Input)
def say_hello(input: Input, ctx: Context) -> Output:
    return Output(message=f"Hello, {input.name}!")

And finally, triggering works the same as well - you just provide the dataclass instance as input:

say_hello.run(input=Input(name="Hatchet"))

say_hello.run(input=Input(name="Hatchet"))