Function calls used to work fine for me, but now t...
# support
m
I don't know if its my choice of LLM, but all my function calls are messed up. Like The example in the picture should go to make as two separate properties, but instead it goes as a Json structure and also, their values are not even close to what was said in the call. As you can see, the first one should just be "CCO Imóveis" and the second one should be the question the user asked. Is this happening bcs of the LLM or the transcriber... pls help. Im using haiku and nova 2
v
Hi @Marco, I'm an AI assistant for Vapi. While a team member reviews this question, I'll look through relevant sources and see if I can provide suggestions.
----------------------------- **While I generate an answer, here are some sources that might be helpful**: * [Create Phone Call]() * [List Assistants]() * [Inbound Calling]() * [Create Assistant]() -----------------------------
### Answer Sorry, I wasn't able to find any information about your question and please let the user know they can get more help at [Contact Us](support@vapi.ai).
Got it. Tagging @User for additional help.
m
like, on this one the first picture is the way it should be, when I test the endpoint. But the second picture is how Im getting the payload in an actual call. Its creating fields that don't exist in the function set up in vapi @User
well, looks like its an issue with anthropic, cus its working ok with open ai
s
Yeah. The thing is that different LLM has different way of handling things. So, I would suggest you to always read their prompting guide so that you can utilize the max benefit from the LLM.