Wondering about why some of my calls today didn't ...
# support
2 of our calls ended because of: silence-timed-out 1 of them because of: pipeline-error-eleven-labs-voice-failed 1 of them ended the call because it didn't respond 1) pipeline-error-eleven-labs-voice-failed f8103f07-9e06-4224-81e0-2a0c8f4a0485 2) silence-timed-out (although caller did respond) f063b00c-9597-44a1-ad83-0bfbd18e2795 947d68b9-5485-4cc7-977c-883593f199e2 af5f87fa-4a5d-4a33-8aac-6a127ebebb57 3) customer-ended-call (AI didn't respond for 15 seconds) a33bf7d4-904a-41c9-8a2c-ed2921228aea These are 4 phone calls that weren't fulfilled, would like to know what caused and if there's anything I can do to prevent this from happening?
Hi @ZeroBot.ai, I'm an AI assistant for Vapi. While a team member reviews this question, I'll look through relevant sources and see if I can provide suggestions. Please tag @Vapi Support with any follow-up questions.
----------------------------- **While I generate an answer, here are some sources that might be helpful**: * [Get Call]() * [Outbound Calls from Python]() * [Vapi Python SDK]() * [Set up your Server URL]() -----------------------------
### Answer The reasons for the calls not being fulfilled as you described are related to specific error conditions and a customer action: 1. **Pipeline-error-eleven-labs-voice-failed**: This signifies a failure in the Eleven Labs voice service, which could be due to various issues such as service downtime, configuration errors, or quota limits exceeded. It's recommended to check the service status and configurations for Eleven Labs voice service [(1)]() . 2. **Silence-timed-out**: This indicates that the call was ended due to a prolonged period of silence being detected. Even if the caller responded, it's possible the system didn't detect the speech. Adjusting silence detection settings or ensuring better audio quality might help in preventing this issue. 3. **Customer-ended-call (AI didn't respond for 15 seconds)**: This occurred because the AI did not respond within 15 seconds, leading the customer to end the call. This could be due to processing delays or errors in fetching or generating the AI's response. Monitoring system performance and optimizing response handling can help mitigate such situations. To prevent these issues, you might want to review and adjust your setup, especially around the voice service configurations, silence detection thresholds, and system performance optimizations. Additionally, ensuring fallback measures or error handling strategies are in place can help improve the user experience even when issues occur.
1) 11labs threw: "We are sorry, the system is experiencing heavy traffic, please try again. Higher subscriptions have higher priority." Are you using your own 11labs key or ours? 2) Please don't use "general"
Copy code
"transcriber": {
    "provider": "deepgram",
    "model": "general",
    "keywords": [],
    "language": "en"
Use the latest nova-2 model for web calls and nova-2-phonecall for phone calls. we do this by default so maybe don't even send us transcriber. 3) Likely the same thing where Deepgram didn't transcribe or send us anything Send more call ids if still an issue after transcriber fix
1) I am currently using my own 11labs key, do you recommend using 11labs without the direct api key? 2) I've updated the settings from the Vapi interface to use "Nova 2 Phonecall" just now We had a call before this update which resulted in a silence-timed-out This one's from today: 83c742cf-9c0a-495b-b547-2f51ac469c03 Excited for the results!
2. This occurred because Deepgram didn't send us anything for 10 seconds. To fix it, you can increase the silence timeout to 30 seconds.
1) Yes, our account is enterprise and is safer to use 2) On the call id you sent, I still see
Copy code
šŸ”µ 14:04:09:873 Deepgram Model "general". Using Default Client...
This is pretty neat! For the Deepgram "general", its probably because that sample was before I updated the model. Since updating it, I haven't had any new silence-timed-outs šŸ™‚