Sometimes my prompt works perfectly for
# general-english
k
Sometimes my prompt works perfectly for 3 cases but 2 no when i try to fix the 2 last ones others problems appear on the others cases, it doesn’t call tools when needed
k
Use multi_tool_use.parallel, removing outdated markers like ## Function Call, and ensuring tool settings like async/sync are correctly configured..
k
can i show you my prompt and give me some advices on it ?
it has little bugs like sometimes it doesn't call the tool or address someone informally instead of formally
or it says two times in a row wait please
i didn't configure messages on my tool specifically and i use this vapi agent is linked to my n8n workflow
how do i use multi_tool_use.parallel or parallel_tool_calls
k
Your prompt is solid but needs clearer structure, simplified logic, and tool call examples to avoid missed triggers, informal language, or repeated messages, define tool usage clearly, avoid overlapping conditions, test step-by-step, and ensure your webhook returns valid JSON with exact
toolCallId
.
Call multi_tool_use.parallel with a tool_uses array listing each tool and its parameters but be aware GPT‑4o may bug with malformed calls, so disable parallel calls parallel_tool_calls=False if needed..
k
Okay thanks
can i send you a log exemple to see the problems ? i tried to changed it but tbh it's the best that i made for the moment i feel like when he calls 2 tools at a row he uses an automatic response but it's informally so it's bothering
for webhook responses i have any problem, i just have a problem sometimes it doesn't call the tool when needed (somtimes it does), i don't know what's the factor that makes it work or no, and the automatic response from vapi