Set optional input parameter for custom component using component.set_input_types()

[danielescaramuzzi]

If I want to set inputs for a custom component in bulk using component.set_input_types(), I have no way to mark parameters as optional.

Example (a component that receives a combination of int and Optional[int] parameters)

@component
class MyComponent:
    def __init__(self, input_types):
        in_types = {}
        for key, value in input.items():
            if value == "int":
                in_types[key] = int
            elif value == "Optional[int]":
                in_types[key] = Optional[int]
            else:
                in_types[key] = Any
        component.set_input_types(self, **in_types)

MyComponent(input_types={
   "param1": "int",
   "param2": "Optional[int]"
})

Using Optional[int] as type of param2 doesn't seem sufficient to make it optional. Looking at the set_input_types() method in the component class, _empty is used as default value for every input, making all inputs mandatory (it means that the component only runs when both param1 and param2 are available).

Is this behaviour expected?

If so, how can I make an input optional using component.set_input_types()?

[julian-risch]

Hi @danielescaramuzzi Haystack decides when to run a component in a pipeline based on whether the inputs it expects are available. That's why making an input optional can get complex. I suggest two options for what you want to achieve.
First, you could define the input as Variadic, similar to the StringJoiner here: https://github.com/deepset-ai/haystack/blob/main/haystack/components/joiners/string_joiner.py#L46
The second option is to define a default value for your param2, similar to how Haystack handles optional ("non-required") variables in the ChatPromptBuilder here:

haystack/haystack/components/builders/chat_prompt_builder.py

Line 144 in c037052

component.set_input_type(self, var, Any, "")

For more context (from https://github.com/deepset-ai/haystack/blob/main/haystack/core/component/component.py#L384):
Note that if the run() method also specifies some parameters, those will take precedence.

For example:

@component
class MyComponent:

    def __init__(self, value: int):
        component.set_input_types(self, value_1=str, value_2=str)
        ...

    @component.output_types(output_1=int, output_2=str)
    def run(self, value_0: str, value_1: Optional[str] = None, **kwargs):
        return {"output_1": kwargs["value_1"], "output_2": ""}

would add a mandatory value_0 parameter, make the value_1 parameter optional with a default None, and keep the value_2 parameter mandatory as specified in set_input_types.

[danielescaramuzzi]

Hi @julian-risch,

• Regarding Variadic:
  I don't think it's a viable option. If I understand correctly, using Variadic for param2 would mean that the component runs only when param1 is provided and at least one value for param2 is received. However, that's not the behavior I need—the component should execute whenever param1 is available, regardless of whether param2 is present.
  I tried using param2: Variadic[int] and param1: Variadic[Optional[int]], but the behavior remains unchanged.

• On set_input_type() vs. set_input_types():
  Using set_input_type() makes sense, but I’d like to confirm whether the same result can be achieved with set_input_types().

• About https://github.com/deepset-ai/haystack/blob/main/haystack/core/component/component.py#L384:
  I attempted to override param2 in run() with None as the default value, but encountered the following error:

  File "/.../haystack/haystack/core/component/component.py", line 226, in inner raise ComponentError( haystack.core.errors.ComponentError: set_input_types()/set_input_type() cannot override the parameters of the 'run' method

Any insights on this would be appreciated.

Thanks!