String run_input for resume_flow_run

[gsakkis]

According to the docs, the wait-for-input parameter of pause_flow_run and suspend_flow_run can be a built-in type like int or str. However resume_flow_run's run_input should be "a dictionary of inputs to provide to the flow run"; trying to pass a string raises an exception. So how is resume_flow_run supposed to be called if pause_flow_run expects a string?

[desertaxle]

Hey @gsakkis!

When pause_flow_run(wait_for_input=str) is called with a built-in type like str, Prefect internally wraps it in a dynamically generated AutomaticRunInput Pydantic model that has a single field named "value". So the flow run is actually waiting for a model like {"value": <your string>}.

To resume it, pass the dict with that "value" key:

resume_flow_run(flow_run_id, run_input={"value": "your string here"})

So the pattern is:

# Pausing side (inside your flow):
user_input = pause_flow_run(wait_for_input=str)


# Resuming side (external caller):
resume_flow_run(flow_run_id, run_input={"value": "hello"})

The same applies to any other primitive type (int, float, etc.) — always wrap it in {"value": ...}. If you use a custom RunInput subclass or a Pydantic model with named fields, then those field names are used directly instead.

I'll submit a PR to clarify this behavior in the docs.

[gsakkis]

Thanks @desertaxle! If this is documented somewhere I didn't spot it.

Btw is the user_input returned by pause_flow_run the wrapped AutomaticRunInput object (or dict) or it's the unwrapped value?