Hi,
I am using Azure Foundry as the agent service provider and memory / conversation manager.
I have tried both python-1.0.0b251204 and python-1.0.0b251120 tag versions.
I have defined a workflow with two agent executors and two custom executor without agentic capabilities. The workflow takes human inputs in certain cases for example requirement human feedbacks and decisions. For the two agent executors, always new conversations were created for each beforehand.
This error happens when the workflow took inputs, and resumed to the next executor that is an agent executor. I don't see other patterns. Sometimes the workflow can iterate many times (ie., executors are moving on and on), but sometimes it fails after taking first human inputs.