Hi, I'm havint trouble with the bot interrupting. I have tried to adjust the response and llm delays but that only worsens the latency and does not improve the interruptions. I also tried prompting it to avoid interruptions but saw no improvement either. Any ideas on how to fix this issue? These are my settings: https://cdn.discordapp.com/attachments/1265769149590667285/1265769150165549228/image.png?ex=66a2b6e2&is=66a16562&hm=d29f01f7d1bbe5e0721c629249ae95a6b5ff856028227a34c046538b2dd54ea3&

@rodrimeza I'm here to help you with the interruption issue you're experiencing. To better assist you, could you please provide me with the current settings you're using for

numWordsToInterruptAssistant

and

maxDurationSeconds

? Once I have those details, I'll be able to offer more specific advice or adjustments to improve the performance of the bot's interruptions.

@Vapi the numWordsToInterruptAssistant is set to 2 and the maxDurationSeconds is set to 900

@rodrimeza Thank you for providing the settings. If the bot is interrupting too often and adjusting

numWordsToInterruptAssistant

to 2 hasn't helped, you might want to increase this number slightly. This will give the user more time to speak before the bot considers interrupting. As for

maxDurationSeconds

, a setting of 900 seconds (15 minutes) is quite long for most interactions. Unless you have a specific need for the bot to wait that long before timing out, you could consider reducing this to a more reasonable duration, such as 60-120 seconds, which is typically ample for most use cases. Would you like to try adjusting the

numWordsToInterruptAssistant

to a higher value, such as 3 or 4, to see if that reduces unwanted interruptions? If so, I can guide you through the process.

@Vapi increasing the numWordsToInterruptAssistant to 4 did not help either

@rodrimeza I understand that increasing

numWordsToInterruptAssistant

to 4 did not resolve the issue. In that case, it might be beneficial to look into other factors that could be causing the interruptions. These could include the sensitivity of the voice recognition system to background noise, the clarity of the user's speech, or other settings that might affect the bot's behavior. If you've already tried adjusting the relevant settings without success, it may be helpful to review the bot's logs to identify any patterns or specific instances that lead to unwanted interruptions. Additionally, if there's an option to adjust the sensitivity of the voice recognition or to add more robust noise cancellation, that might also help. Since the issue persists despite the adjustments you've made, it might be best to reach out to the support team for a more in-depth analysis. They could provide insights or updates that could address this issue more effectively. Would you like me to provide you with the contact information for further support, or is there anything else I can assist you with at this moment?

@Vapi yes, id like to reach the support team, thank you

@rodrimeza Certainly! You can reach out to the support team by sending an email to [support@vapi.ai](support@vapi.ai). Please include a detailed description of the issue you're facing, the settings you've tried, and any other relevant information that could help the support team understand and resolve your problem. If you need any further assistance or have additional questions, feel free to ask. I'm here to help!

can you please share the call id.

@Shubham Bajaj yes, a good example of the problem we are having is this call: 54cd65a4-a90e-4429-8a19-145a8709b2f1

timestamp as well.

july 23 at 8:43 pm ET

set the LLM Request Delay to 0.4 seconds using API if can't do using UI.

not really, it still interrupts, i baited it by pausing mid thought and it interrupted, but it also interrupted while i was in the middle of longer thoughts. the call id is: 0c85fc53-84cf-4da2-aac5-63ba31dd3be2 and it is from today july 24 at 5:17pm ET

i checked it is working as you stop bot starts speaking and in one interrupted you better is to aslo set Response Delay to 0.4 seconds and try now.

that works a bit better thank you!

closing this ticket for now.