I’ve noticed that AIP Agent Studio agents don’t update the object set context passed to them in Workshop after initial load. This occurs even when the agent is refreshed/reloaded - the LLM will continue to answer based on the initial object set context provided to it when the Workshop loads.
In practice, this means that more complex workflows that involve changing the filtered object set context input to the agent are not supported. The workaround is providing object set property variables as string/string array context to the agent, but this isn’t first class given Foundry’s reliance on Ontology.