Issue with agent not calling tool calls when it sh...
# supportb
big_doggo_best
07/10/2025, 4:34 PMcall ID: 313a4d90-1820-4839-a8b2-3a7bc59a8ff9
As attached above, it is a known issue that sometimes LLM says it'll use a tool but doesn't do what it says, bottle necking the call. These cases seriously effect the customer experience and want to know if there's a configuration to handle this (such as sending SSE on 3 second time-out after assistant-end token).
We have tried altering confidence threshold both for transcription + EndOfTurnConfidence.
c
Chiranjeet Mishra
07/12/2025, 8:13 AMHey big_doggo_best,
Thank you for reaching out. Please note that our team does not operate over the weekend. We’ll be back in the office on Monday and will dive straight into your issue then to determine the best way forward.
Regards,
Shubham Bajaj
Solutions Engineer
Chiranjeet Mishra
07/12/2025, 8:22 AMHey big_doggo_best,
Thank you for reaching out. Please note that our team does not operate over the weekend. We’ll be back in the office on Monday and will dive straight into your issue then to determine the best way forward.
Regards,
Shubham Bajaj
Solutions Engineer
Chiranjeet Mishra
07/16/2025, 3:00 AMThe system prompt is incredibly large and your maxTokens value is 250, which may truncate the instructions to the assistant or cause unexpected behavior. Increase your maxToken value to 1000-1200 or consolidate your prompt as much as possible, then run some tests.
4 Views