Does anyone else have this issue and how to overcome? Since the ability to pass a step output token from an agent task to a service task does not exist, I bound the output to a Maestro process variable. The agent task outputs a string. I bridge it to assign it to the Maestro variable with HITL. I then push that into the RPA workflow. If my in-argument type in the workflow is object I get {} with no parameters although execution logs and orchestrator logs clearly show input. When I change the in-argument type to string I get null input with .length of 0. Does anyone else have problems bringing agent outputs into rpa workflows using Maestro variables? Does not even seem possible if I use the RPA workflow as a tool for the agent.
Set the input type of the argument in RPA to JObject
And then use it within the RPA project like yourJObject("keyName").ToString() or for nested objects yourJObject("parentKey")("childKey").ToString())
![]()
In my example the json object is of type {"sum":2,"product":1}
You can directly pass the json output of Agent as input argument to the RPA flow.
Thanks for the input! Unfortunately, the JSON is dynamic and not deterministic. It can change from source data to source data whether it be xml, JSON, csv, pdf, xlsx so there is not a static defined schema of that type. Sort of like an ETL (Extract, Transform, Load) model. Talked with another MVP last night and got some alternative ideas. I certainly appreciate the input, however. Have a fantastic day!
Thanks @Chris_Bolin, please do share what you come up with, it may be useful for others in future.
