I am using the open ai GPT 4.0. Ideally, this llm should know the current date and time but through Vapi it doesn’t work I am using a dynamic transient-based assistant which I am returning through a webhook in make.com. I am providing the current date there as well using the formatDate(now; YYYY-MM-DD) function. I have checked the response and the date is present in the prompt but the assistant still struggles to understand the current date. Some tips and tricks on this would be appreciated. Thanks! For context, this is the assistant that I am returning as a dynamic transient-based assistant and I am having a lot of issues with the date. I have reduced the temperature as recommended and am using gpt40 which is supposed to be good but you can see how bad it is in the following recordings - 1) call id - 5479b977-1808-47b4-af76-5e81cce5c959 -> 1 minute onwards where it repeats some details and gets the date wrong in the end 2) call id - a59ffa21-1666-4754-91cf-69a63bef5b44 -> last 1 minute I would appreciate some help with making the assistant more reliable. If there is some issue with my prompt please let me know. Or if any workarounds are working for other people, I would love to know https://cdn.discordapp.com/attachments/1264979923161055265/1264979923488215142/message.txt?ex=669fd7db&is=669e865b&hm=2e40985c9c8044c7b8d8014f8d6aa7b03559f10b5cf963260117f9c334b28bb4&

@r7sh7 'yes the pre defined dynamica variables are not working as expected please use assistant overrides for date and time and to instruct the AI not to change any parameters add the following.

Hi Shubham thanks for replying. I am using dynamic date and time variables while returning the transient-based assistant from make.com to give the AI a context of the current date and time. I am not using the built-in {{date}} variable of VAPI. The assistant is also asking questions multiple times. I am not sure what the issue is. I tried the same prompt in GPT4o in the chat and it worked really well. Seems to be a problem with Vapi.

actually your variable was never replaced from make and for above assistant id is used.

Why is this happening? When I use the same prompt in the dashboard, it works fine but the dates are hardcoded in there. When I am retrieving the assistant via an API call and providing the date dynamically based on the current timestamp, it does not work. Can I know why? I have attached the json returned by the make.com rest API endpoint with the latest date and time in the "content" key. When I paste this same prompt after removing all the formating in the vapi dashboard it works fine. https://cdn.discordapp.com/attachments/1264979923161055265/1266012444430569534/message.json?ex=66a39978&is=66a247f8&hm=ba7e38b3c10d7a3395a50609bba0eb90f7ebd68795483dd8e1dfde2e7ee18fdc&

@r7sh7 can you please share the call ids with timestamp.

1) call id - 5479b977-1808-47b4-af76-5e81cce5c959 -> 1 minute onwards where it repeats some details and gets the date wrong in the end 2) call id - a59ffa21-1666-4754-91cf-69a63bef5b44 -> last 1 minute