@Marvin what function does prefect use to serialize metadata of flow iputs

Prefect uses a

result_serializer

for serializing metadata of flow inputs. This is part of the flow configuration, where you can specify how results, including inputs, are serialized. You can find more details in the Prefect source code for flows here. To learn more about handling inputs in Prefect, you might want to explore the documentation on flow inputs here.

@Marvin would it use

result_serializer

for serializing flow's input values?

The

result_serializer

in Prefect is primarily used for serializing the results of tasks and flows, not specifically for input values. Flow inputs and outputs are managed separately, and inputs can be handled using Prefect's input management features, such as pausing a flow to wait for input or using the

wait_for_input

parameter. For more details on handling inputs in Prefect, you can refer to the flow inputs documentation.

hi @Ji-Oh Yoo - what are you trying to do?

Hi @Nate we are using prefect in local mode, with concurrent task runner, using ephemeral server to talk to postgres DB. We found a flow that's passing a huge dataframe, and would like to skip storing the dataframe to DB

usually i’d recommend passing a reference to the data instead of the data itself, so some s3 / path on disk wherever you have it saved

Yes, that would be the right solution, but we need to provide a hot fix for a customer, so trying to get any workaround for it