I’ve had most of my clients complain about the tools not finishing it’s output. And I keep experiencing the same thing. It’s not a connectivity issue as I’ve made sure of that. It’s like it takes a long time to think and then it spits out some output but doesn’t finish it’s thought. And when prompted to keep going, it will sometimes but sometimes it starts all over and does the same loop over and over.
Just checking to ensure you’ve got sufficient output tokens set in the configure tab of the builder? That’s the number one reason we see things failing to finish, though if those are set it’d be helpful to know what models you’ve been experiencing this with specifically?
1 Like
Thanks my prompts/tools create pretty long outputs (beyond 3000 words) so I am updating all my prompts with limits and then will configure the outputs accordingly. I am using Haiku right now.
That’s indeed one of the main reasons this issue occurs with many users. However, It can easily fly by the user that they should adjust the output.
It will be super useful to add a error message in red informing users that the output stopped due to this specific factor - and informs them where and how to adjust the output token slider.
Maybe this error message can be placed below the chat text input box similar to other error messages that appear.
1 Like