Hello,

I would like to use an agent to redact an email. It would iterate calling tools, thinking and calling other tools..
At the end it would formulate the final email.

What the agent outputs would be thinking and not said to any user.
In this use case the rooms or anything voice related are not used. This is text only.

What would be a good way to create such a thing?
I see we can now test the llm part of the agents, maybe this is related?