I hope you’re doing well. I’m reaching out because I’m experiencing an issue with our VAPI agent integrated with n8n. Specifically, when the agent is waiting for a response from an n8n webhook (for example, checking table availability for a restaurant reservation), the agent sometimes skips the actual webhook response if the user says something like “okay” or interrupts in any way. Instead of waiting for the n8n data, the agent just moves on and gives a premature response (like confirming a reservation) without the actual availability data. We’ve tried adjusting the System Prompt and adding delays, but we’re still facing this issue. Is there any recommended approach or best practice to ensure the agent always waits for the webhook response before proceeding? Thank you in advance for your help! Best regards, David

To prevent the agent from skipping webhook responses due to user interruptions, configure a more conservative

stopSpeakingPlan

in your voice pipeline. For example:

json
"stopSpeakingPlan": {
  "numWords": 2,
  "voiceSeconds": 0.4
}

This setup waits for at least two transcribed words before treating an interruption as valid, reducing false triggers from short interjections like "okay" during webhook waits. Apply this in your assistant's configuration to ensure the agent reliably waits for the webhook response before proceeding. For more details, see the [Voice Pipeline Configuration documentation](https://docs.vapi.ai/customization/voice-pipeline-configuration). Source: - [Voice Pipeline Configuration documentation](https://docs.vapi.ai/customization/voice-pipeline-configuration)

I am experiencing a consistent issue with my voice assistant integrated with n8n. When the assistant checks table availability (restaurant use case) and says something like: “Give me a moment while I check availability,” the webhook runs normally. If the customer stays silent, the assistant waits and uses the RespondToWebhook output correctly. But if the customer says anything during the wait — for example: “Okay,” “Alright,” “Listo, dale,” “I’m waiting,” etc. — the assistant skips the webhook response and hallucinates a reservation confirmation (e.g., “I’ll book the table for you”), without ever reading the actual availability returned by n8n. To reduce false interruptions, I already increased: "numWords": 7, "voiceSeconds": 1 But the issue persists consistently. I need guidance on how to ensure the assistant always waits for the webhook response, even if the user speaks briefly during the tool-call. Thank you, David

Hi, could you please provide a call id where you experienced this issue? We would like to review the logs to identify the cause.