FunctionalFlowStep

[OlaRonning]

Why not use **kwargs here?

llmflows/llmflows/flows/functional_flowstep.py

Line 42 in 3c4ceb7

self, inputs: dict[str, Any]

Then you can name keyword args to the function (i.e., generate), and the remaining will be the arguments for flowstep_fn, so this will go away

llmflows/llmflows/flows/functional_flowstep.py

Lines 58 to 60 in 3c4ceb7

	filtered_inputs = {
	key: value for key, value in inputs.items() if key in fn_args
	}

Rename flowstep_fn to inner_fn or step_fn; it's part of a flowstep class, so the context is already clear.

llmflows/llmflows/flows/functional_flowstep.py

Line 39 in 3c4ceb7

self.required_keys = inspect.getfullargspec(self.flowstep_fn).args

^ is not used.

[stoyan-stoyanov]

Good point. Actually, the generate() method is supposed to be used only by the flow class so it might be also _generate() And looking at it the return type is too general.

Also, the flow class should be already taking care of passing the right inputs so I don't know why I added the filtered_inputs check.

required_keys is used in the flow class:

llmflows/llmflows/flows/flow.py

Line 76 in 3c4ceb7

required_inputs = {key: inputs[key] for key in step.required_keys}

[OlaRonning]

If the two are that tightly interlinked, then you should have a runtime check that you have a flowstep instance.