Hello Support Team, We’re seeing the assistant stop speaking on very short Japanese acknowledgements (e.g., 「はい」, 「ええ」, 「うん」) even though our stop-speaking configuration is: call id: 14d943e1-4205-4b49-87d2-e061a002a4f8 "stopSpeakingPlan": { "numWords": 0, "voiceSeconds": 0.5, "backoffSeconds": 1, "acknowledgementPhrases": ["はい"] } Issue: Utterances shorter than ~0.5s still interrupt and stop assistant speaking midway. Language: Japanese (ja) STT: Deepgram (nova-2), final transcripts show durations ~0.41s TTS: ElevenLabs Could you advise concrete parameters to prevent these micro-acknowledgements from interrupting? Thank you

we experienced an issue where the some of the status callback did not arrive for all calls, especially for those where the Twilio connection failed or the customer did not answer. Could you please look into what might be causing this? I checked the webhook logs and can confirm that no status callback triggered for those calls. OrgId:fb1c5e62-fa11-4c3d-bc72-bdf22aaa56a5 example callId: 91680ec8-eef8-41e5-80e9-466c77d1529a

I sent an email request as well https://cdn.discordapp.com/attachments/1418067322404601998/1418067322630967346/Screenshot_2025-09-17_225049.gif?ex=68ccc59c&is=68cb741c&hm=0bc60be2e0e244d7d8fe4070a75c70922ec09d2c66acbb53e6ae1b949ddb26cc&

mozartcuidadordigital@gmail.com's Org 1ba0571e-e3df-4ab4-9e6d-cc7d43693fd1 We are currently in an initial pilot phase with one of our clients, and our goal is to grow organically with them over time. For now, however, we need to better manage our daily call consumption. At the moment, our account is set with a concurrency of 10 calls. We kindly ask if you could help us reduce this limit to 5 concurrent calls. This adjustment will allow us to continue testing and evolving the project without exceeding our current budget.

I'm trying to test my agent on the computer and I keep getting this error "Error Key doesn't allow assistantld 'a35610a7-2110-4de0-a2b9- aa8f4bd10c84'." What is going on? It use to work fine before. https://cdn.discordapp.com/attachments/1418054097248194630/1418054097550049411/image.png?ex=68ccb94b&is=68cb67cb&hm=051c8536aedbfe4a0033c770ddd6b62d41578f67e5e35bbef88ef819fbdbbbad&

Suggestion: The error means your API key does not have permission to access the assistant. To fix this, try generating a new public API key in your Vapi dashboard and ensure it is linked to the correct assistant: 1. Go to Vapi Dashboard → API Keys. 2. Create a new public key and use it in your integration. 3. Make sure the assistantId matches the one authorized for your key. If the issue persists, double-check that your assistant and key are in the same organization and environment. Source: - [Vapi API Key Permissions](https://docs.vapi.ai/chat/web-widget#copy-your-public-api-key)

Is it possible to keep recording after transfer? Thanks

I'm a bit confused on how pricing is supposed to work here. We have ONE assistant, configured with Open AI Chat GPT 4o cluster (standard), using our own eleven labs key. 20 second call was: $0.07 ($0.21/min) 39 second call was: $0.10 ($0.15/min) 38 second call was: $0.13 ($0.20/min) 50 second call was: $0.12 ($0.144/min 2 minute 15 second call was: $0.68 ($0.30/min) 3 minute 20 second call was: $0.90 ($0.27/min) 4 minute 57 second call was: $1.48 ($0.29/min) This makes it a bit difficult to chage our clients a fee - if we cannot actually estimate what the actual per minute fee is. When we look at the assistant itself in VAPI it states .19/min, but that's clearly not the case.

I have a node in a workflow with the following

Ask a series of questions in order to gather all the data we need to get a better approximate rate quote. When we have all the data, move on to the next step.

The questions are :
1. Can I have your first name
2. Are there more borrowers on the loan
etc...
etc....
etc...

[conversation flow]

1. Keep the conversation on topic and do not ask questions that we do not need.
2. Only ask one question at a time.
3. Never combine multiple questions.
4. Never reflect back the entire conversation.
5. Never ask for more than 1 piece of information in a question during the conversation.

However it keeps asking multple questions like this :

Assistant
Understood, so you’re a first-time buyer. What do you estimate your combined gross monthly income is for both you and your husband? What are your monthly payment obligations, like car payments, credit cards, or student loans? If you have any deferred

How can I get it to absolutely never combine questions??

Incoming SMS are not starting a conversation. The SMS are coming into Twilio, and Twilio has the 10DLC configuration set, but it doesn't seem like anything is happening with webhook to start the conversation. I don't see documentation on the webhook so I'm not sure how to even test the response manually outside of twilio.

Hello, my name is Milan Roušavý, I am from the Czech Republic. I am integrating your platform (VAPI) with a SIP trunk from IPEX, which is a Czech VoIP/SIP provider offering DID numbers in the Czech Republic. My provider (IPEX) assigned me the DID number: +420 910 107 712 In VAPI I created an inbound phone number 910107712 (non-E164) and attached it to my assistant. The SIP trunk is configured as static, so I connect only by IP address – no registration is used. On the IPEX side the trunk status shows as Reachable. For IP access, I followed your documentation and whitelisted these signaling IPs in IPEX: 44.229.228.186/32 44.238.177.138/32 📋 Problem description When I call my DID from an external mobile number, IPEX log shows that the call is routed to the trunk muj_demo and the number 910107712. However, the call never appears in VAPI call logs – there is no trace of the inbound INVITE in VAPI. The caller only hears a busy tone. 📑 Log excerpt from IPEX 2025-09-17 22:58:49:498 Volající číslo: 773985625 Volané číslo: 910107712 Vstupní směrování VoIPEX vyhledávací vzor: 910107712 Volám přesměrování na Trunk - muj_demo 2025-09-17 22:58:49:507 Příchozí hovor z čísla 773985625 byl ukončen ❓ My questions Why is the inbound call not visible at all in VAPI call logs, even though IPEX confirms it forwarded the INVITE to the trunk? Does VAPI use any additional signaling IP addresses or subnets that must also be whitelisted on the provider side (IPEX)? Can you provide the complete list of SIP signaling IPs and RTP/media IPs that must be allowed for inbound calls to work? For a static SIP trunk (IP-based only, no registration), is it sufficient to configure only the provider’s IP (212.71.129.36, sip.ipex.cz) on the VAPI side, or should we also configure domain/realm values? Could the issue be related to the number format (E.164 +420910107712 vs. national 910107712) or codec negotiation? Thank you very much for your support and guidance. Best regards, Milan

The voicemail detection seems to work for assistants but not workflows. How can I configure it to work for workflows?

hey folks, does anyone have a beginner guide how to send a post call webhook to a make automation? Thank you! I have the sdk but i have no idea what to update.

Cant log in!! Critical impact. Please resolve so we can log in. https://cdn.discordapp.com/attachments/1417939325626355824/1417939325881942248/Screenshot_2025-09-17_at_11.18.35_AM.png?ex=68cc4e67&is=68cafce7&hm=b8c7dcac45bf4b3e7a2a1ca5b94826553ddd9d12948f015e0d79ee3a99d6a058&

Hi Team, I'm looking for the best solution to send the summarization of a tech support AI agent call via email using. I'm new to Vapi and I would like to know what tool, integration or other feature I could use to make this happen. Thank you all!

There seems to be quite a cumbersome issue related to the agent creation SDK. We are trying to set up all our agent creation requests with the 'Background Denoising' and 'Use End Call Function' toggles enabled. However, it seems to be a hit or miss and this is causing some of our users to question the performance of our app since we now have to manually ensure these are activated whenever they create agents. PLEASE HELP!

I'm building a financial commitment system that requires a two-step authentication flow: Tool 1: Get OAuth token (POST to /oauth/token) → returns access_token Tool 2: Use token to call business API (POST to /api/resource) → requires Authorization: Bearer {token} My current setup: Tool 1 (get_auth_token): Works correctly Returns valid access_token Configured as Function with API Request Tool 2 (call_business_api): Header configured as: Authorization: Bearer {{get_auth_token.access_token}} Fails with "Invalid token" because header arrives empty LLM doesn't automatically pass token from Tool 1 What I've tried: Automatic variables: {{tool_name.variable}} in headers - Doesn't work Explicit parameter: Added auth_token as parameter to Tool 2, configured system prompt for LLM to pass token - Doesn't work consistently Workflows: Tried workflows with variables {{access_token}} between nodes - Workflows have bugs (doesn't start conversation) My question: What's the correct way to pass a tool's result as a parameter to another tool in Vapi? This is a very common pattern in enterprise applications where you need to: Get authentication token Use it immediately in business operations Maintain separation of concerns between tools Technical info: Platform: Voice calls via assistant Token type: JWT with short expiration (fixed tokens not viable) Endpoints: Corporate REST APIs with OAuth 2.0 Use case: Financial system requiring authentication for regulatory compliance Has anyone successfully implemented this type of flow? Is there a specific configuration I'm missing? Any guidance would be greatly appreciated. Thanks in advance.

We are evaluating Vapi as a provider for a new product, but our legal team has some questions that we want to discuss before moving on. I tried to submit the form to get in touch with the sales team, but I have not had any feedback yet. Is there a way we could set up a call with the sales team and/or a solutions engineer to go through those security/legal concerns we might have, and also to understand some privacy aspects of the contract?

Hi, we would like to install oin Prem for us to dev/test and then for our customer. We have devs available to do the install. How do we continue from here?

Hello guy's We know that STT for non-english AI Voice are hard and lack accuracy. We have developed a tech that allow us to create industry centrics STT with minimum 30% better WER than any other model on the market and average latency of 350ms. I think we could discuss about it with Vapi team, with whom can I discuss ?

Hi Everyone, I'm having issues with Tools sections of the ai agents. I have set up a tool that sends a request n8n to receive some data . the request is sent successfully but the response sent back doesnt show in vapi even though i'm certain a response is sent. the response is sent back in following way : [ { "results": [ { "toolCallId": "", "result": { "exists": true, "first_name": "", "last_name": "", "address": "", "unit_number": "" } } ] } ] I would apperciate some help. Thank you 🙂

Chat ID: c64d3f0e-0683-4ddd-9c9d-7f2dbf0a26ba Time stamp: Sep 17, 2025 22:03 The response I expected is the chat bot able to retrieve the correct information from the KB and provide it back to the user. The response I received looks like it is a message 'Success' with no other details. I have uploaded a JSON file and added it to query tool so that it can look up information in the file by matching the provided ID to the same ID in the KB. I was debugging on the calls but it seemed to provide more information on the chats and I could see the function calls. I have even asked the AI assistant questions on other chats and this is what is said - "You are right to ask for clarification. The actual response from the query_triumph_physio_business_data call was simply "Success." without any detailed data included in the message. This means I did not receive the specific business, appointment type, or practitioner details from that call." https://cdn.discordapp.com/attachments/1417845520587686000/1417845520940011580/image.png?ex=68cbf70a&is=68caa58a&hm=1343b8b0a9cd74bc976c7360d23220bf940cf0bd75a8386811116f5b4b7b6431&

Hi, I’m trying to use the Blind Transfer with Summary in SIP Header feature, but I noticed the summary comes through as a REFER. Is there any way to make it work with an INVITE instead? I’m running Asterisk 20 and 11 (currently testing on 11). The problem is, I can’t really get REFER working at all. In our setup we don’t usually use REFER for transfers. We rely on the PBX’s built-in * transfer, so having this available via INVITE would be super helpful. Thanks a lot in advance for any pointers!

Need to connect with VAPI executive to discuss the feasability of integrating VAPI voice in our voice bot

Does anybody know how to make it so your workflow on n8n automatically starts as soon as your call is answered from the agent

I want to use VAPI but can’t find any provider offering an Indian number without issues.

As per title, we cannot trigger call from the VAPI dashboard. https://cdn.discordapp.com/attachments/1417723944856850582/1417723945309962310/Screenshot_2025-09-17_at_2.08.22_PM.png?ex=68cb85d0&is=68ca3450&hm=55fda552cc2dcaa4b03eed9f1e97004550ef6320a86916bd0af261956ff8a86c&

Why does when the user calls in, we get a value in {{customer.number}} sometimes and not all the times? Call ID: b519b898-0c97-4209-b339-f90cdfd397ac

I've been using vapi for a while but haven't logged in for about a month and when I try to login today it asked me to do a identity verification which I did. Now the screen is still verifying it. I've revefiried and still its stuck on the verification screen https://cdn.discordapp.com/attachments/1417687754497589329/1417687754908762173/StaticShot_17-09-2025_11-40-08.png?ex=68cb641c&is=68ca129c&hm=be55be5c191d4129336da89c285c38b33e9d477084cea21fe17787e4efc4bcb2&

I need to pass a specific callerid so AE knows its a transfer from ai, but it doesnt work. https://cdn.discordapp.com/attachments/1417634366871306371/1417634367223889950/image.png?ex=68cb3263&is=68c9e0e3&hm=be7a62eef886ed60f4cff8992af740b681db704e15fd2b7f542ac64c7c8ec770&