That does appear to possibly be what’s going on here @t0bst4r, at least from what I am seeing.
I’ll have to do some testing with the built in openai agent, as I’ve only ever used the extended openai agent.
Switching to Extended Open AI conversation also fixed it for me without changing anything else. Thank you @Rich37804 for the tip!
But if I am right, it should always happen whenever the built in agent is used as primary agent. Actually it should happen everytime when it does not recognize an intent. Even when used as the standalone agent…
Oh, I misread what you said. No, I think it’s specifically an issue with the OpenAI agent that is built in.
The “Turned on the light” would be correct, since there is a request for that afterwards. But for some reason, it’s returned “Unexpected error during intent recognition” rather than an appropriate error message from the fallback agent. Why it’s erroring is also not clear.
1 Like
I created an issue here: Issues detecting error from built in OpenAI Agent · Issue #13 · m50/ha-fallback-conversation · GitHub
Thanks for reporting. I’ll have to do some testing and see if I can figure out what is going on.
Oh yeah, I should have looked better at the screenshots. Just answering in between some meetings during work 
I just assumed it didnt work with that integration. I didnt realize it was an issue…