VAPI not responding on calls... why?
# support
b
Examples: CallID bb87aa68-de49-49a6-9356-406e7f501714 CallID 697df4b2-88e3-468e-83e5-d08eee1bf08c VAPI dials numbers, waits for speaker to stop speaking, but does not answer. Other calls (such as ca433994-2691-4607-8b6c-92999c558308) seem to work correctly. Any idea what's going on? Same parameters on calls with slightly different prompts (one asks for business hours, one asks for list of services). Thanks!
v
@Brian @ Testbot.app LemurLabs.ai It sounds like you're experiencing inconsistent behavior with the VAPI calls. Since the issue is not consistent across all calls, it could be related to the specific prompts or the phoneCallProviderBypassEnabled setting. If the bypass is enabled, VAPI won't initiate calls directly, and any issues might be related to the external call handling setup. To assist you better, I would need to check the configuration for each CallID you've mentioned, especially the phoneCallProviderBypassEnabled setting and the phoneCallProviderDetails. Unfortunately, I don't have direct access to call logs or configurations, but I can guide you on what to check: 1. Verify if
phoneCallProviderBypassEnabled
is set to true for the calls that are not working, and compare it with the ones that are working. 2. Check the
phoneCallProviderDetails
for any discrepancies between the calls that might be causing the issue. 3. Review the prompts used in the calls to ensure they are correctly formatted and do not contain any elements that might cause the VAPI to wait indefinitely. If you've already checked these settings and everything seems to be in order, it might be best to reach out to support with the specific CallIDs and details of your configuration for further investigation. Please ask support for more info on this.
You can continue this conversation by mentioning me in the message.
b
I ran another 2 test calls, one failed again ca433994-2691-4607-8b6c-92999c558308 and one worked 873f4c39-f187-4aab-9091-b18fff0520ee
s
@Brian @ Testbot.app LemurLabs.ai can you be more clear here what exactly is not working for you for this call id ca433994-2691-4607-8b6c-92999c558308?
b
We place the outbound call with VAPI. The call is answered and the initial greeting is spoken. After that, the VAPI agent never responds and the call ends up timing out. It does not always happen, sometimes it works correctly. The example call ids show the one-sided call.
I've changed the model and transcriber to new providers and models, but I'm still seeing these dead-response calls. General Info Test Call: Failing Example: 2cc79add-8604-43f8-91b2-802950d797f2 Working Example: 388a2789-5bc4-433b-9c5b-5688ba294b4a Get Office Hours Test Call: Failing Example: 8a30534d-e93c-4d03-b941-a92b6d2a4161 Working Example: 9239ff6c-e30e-45d6-9d0f-64c6d3483dd7
I have verified that the assistant is identical in both Failing and Working cases.
Is there a problem if I’m starting multiple calls at the same time (2 calls to Start Call API). I thought VAPI could handle 10 calls at once, but still trying to figure out how to duplicate or fix regularly.
Thanks!
s
@Brian @ Testbot.app LemurLabs.ai the model is hallucinating sometimes, it is better to change the model to openai gpt-4o-mini/gpt-4o.
b
I already changed it to 4o and same thing
s
can you share the call id for it?
r
@Brian @ Testbot.app LemurLabs.ai it's probably your mic
Try on some other devices where the mic is working fine
b
There is no mix, it’s one bot calling another. I missed a place where the model was still the old version and that seems to have resolved that issue, but I’m seeing issues where it looks like that analysis doesn’t run and returns n/a… does that mean a true false test failed or it was unable to determine the state? Example: call ID 1bd82ec3-91e3-4b9f-9e96-30fa09f773bf.
Success evaluation= true
Any idea why we’re seeing these inconsistent evaluation results?
Call ID eb5e50f8-d533-4523-8cd6-df986be70db9
s
@Brian @ Testbot.app LemurLabs.ai can you please use new startSpeakingPlan and stopSpeakingPLan endpointing model because of old endpointing model used in your assistant user was considered speaking continuously.
logs
🔵 20:05:22:559 [user CHECKPOINT] User speech possibly stopping 🔵 20:05:13:967 Talkscriber
user
Still Pushing Chunks...
Because of no response closing this ticket.
30 Views