State source of conversation agent response

[velvet bear]

Hi, I'm trying to figure out if I can configure my conversation agent to state when it accesses ChatGPT for its response (vs. when it handles a command locally). For example, if I say "OK Nabu, at what age should I give my poodle his first haircut?", it will use the OpenAI API for its response, so I want its response to be "From ChatGPT, XXX"

But if I say "OK Nabu, is the mudroom door open?", it will handle the command via the local conversation agent, so it will just respond normally.

Is this possible? I tried to tell it to do so in the Instructions for the conversation agent, but it is not working that way.

[cosmic reef]

so LLM prompt doesn't know you accessed local or an "API", it basically IS the API

so maybe just put a prompt directive as simply "Always preface your responses with "From ChatGPT,""

[velvet bear]

I tried that, and it works, like EVERY time. Even for things the local agent should handle, like "is the mudroom door open". So I figured either it is using the OpenAI API even for those, or I am not understanding something about when it should use the local conversation agent 🙂

[cosmic reef]

What's your pipeline setup look like?