Hello 👋🏻 team. We added a Telynx number through the dashboard and so far the assistant response time during the call has become soo long, almost 10seconds. But the error isn't visible in the call log. In the call log - it shows as if the caller always reply late and that is not the case during the call as it was the assistant that always replied late. Here is the call id. 482dbe2f-6742-441c-bddb-521cc82c5ab8

Hi there! 👋 Thanks for reaching out about this issue with your Telynx number. I'm unable to explain exactly why the call log incorrectly shows the caller as replying late instead of the assistant. However, I can definitely help with troubleshooting the slow assistant response time you're experiencing! Based on the information I've found, there are several potential causes for the slow response time after adding a Telynx number: 1. **Configuration Settings**: Your

startSpeakingPlan

settings might need adjustment. Setting

waitSeconds

to 0.8 and adjusting Deepgram's endpointing value to 300 has helped others with similar issues. According to [troubleshooting information](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1344177033625997322), this combination can significantly improve response times. 2. **Model-generated responses**: If your assistant is configured to generate the first message using a language model (LLM), this can introduce a delay of approximately three seconds. The configuration

firstMessageMode: assistant-speaks-first-with-model-generated-message

might be causing delays as mentioned in [this discussion](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1339303957675245669). 3. **Using your own key**: If you're using your own key instead of Vapi's, it can add extra latency according to [this source](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1339303957675245669). Here are some troubleshooting steps you can try: 1. **Adjust your configuration**:

json
   "start_speaking_plan": {
     "waitSeconds": 0.8,
     "smartEndpointingEnabled": false,
     "transcriptionEndpointingPlan": {
       "onPunctuationSeconds": 0.5
     }
   }

2. Consider using a static first message instead of a model-generated one to reduce initial response time.

Thank you for reaching out to us. We've received your message and will get back to you as soon as possible.

This isn't resolved yet.

Hey it was because of drift detected in call logs and because of this responses were delayed. This is coming from telnyx side, as its pretty new tech you can expect few hiccups sometime.

Okay sir. So what do you suggest sir? We should keep using it that way?

Yes, you can use it as long as drips are less than 10% of your total calls.

I don't understand, what is drips?

Sorry, there is a typo. It's not drips, instead, drifts.