To set the stage, I am only starting to poke around Fleet.
I created an Agent in the interface, and then invoked it via Python, twice. When inspecting the output, there were two messages, and the last was not an answer to my trivial question. I ran it twice out of confusion. The output didn’t make sense until I went back to the UI.
In the UI, I have messages in my inbox, where this invocation is resulting a HITL flow to approve the edit of a file.
Because I am invoking the agent programmatically, I expect it to respond based on it’s configuration. I brushed this behavior off in the UI figuring that I was doing something wrong.
When invoking the fleet agent programmatically - say from my app - I never would expect this to be the output of the run.
Is this user error on my end? In case it matters, this agent was created completely within the Fleet Web UI via point/click.