#Hi all, I'm pretty sure our
1 messages · Page 1 of 1 (latest)
looking at the logs now
well, the agent didn't attempt calling the tool at all, did you change anything in brain logic?
the Conv I see is only 22s long
I just asked for the one question we need to test this: "Do you have the email address for Edward", this should give back my test email address from Sheets.
This is a dev project, not production. Once it is working well we'll connect Twillio to it
I appreciate you looking into it. If it is something on our end I'll dig in, figured since it worked earlier this week, ther might be an issue on your end
Which helps if anybody mentions it 🙂
the thing about LLMs is they are inherently nondeterministic and may sometimes not work exactly the same way
especially with providers like OpenAI constantly updating the models
Not to mention their neverending error/issue reports of today
anyway, please look at the prompting guides. Looking at the agent config of yours, I see it would be great to mention in a more specific detail when the tool can be used
Learn how to engineer lifelike, engaging Conversational AI voice agents
Yeah thanks. We just wanted to send ALL requests to the webhook and use the tools descriptions there.
But I'll update on your end too.
yeah, but sometimes LLM brain just decides not to call the tool... but improving the prompt would definitely help 🙂
it doesn't seem like there is a tool outage of some sort, but I'll keep an eye, and please msg me if this issue persists
Thanks for checking, I'll make some adjustments
I will
louis put his soul to it