Feature Requests

You've made 90% of the job by intergrating Gladia solaria transcriber model. Let's just add some configurations parameters, especially custom vocabulary : https://docs.gladia.io/chapters/live-stt/features#custom-vocabulary Custom vocabulary is a must have to compete with Deepgram nova. Custom spelling would be a nice to have.

In the transfercall tool destinations array of objects (CreateTransferCallToolDTO, TransferDestinationNumber), the parameter callerId enables to choose between setting the customer.number or phoneNumber.number. This only works for TWILIO and not VONAGE. A calls B B forwards to C Currently C sees B number by default (phoneNumber.number), I would like this to be a choice.

Want long term conversational memory to be built in.

Add dual mode support for widgets Chat AND Voice Mode, not having to choose either of these.

Environment: Vapi Voice Agent Portal – Transfer Call Tool Description: When multiple numbers are configured under different departments using the Transfer Call Tool and assigned to a specific agent, the system does not respect the department selected during call transfer. Instead, Vapi always initiates the call to the first number configured, regardless of the department selection. Steps to Reproduce: 1.Navigate to the Transfer Call Tool in the Vapi dashboard. 2.Add 2–3 contact numbers, each tagged under different departments (e.g., Sales, Support, Finance). 3.Attach the configured transfer options to a specific voice agent. 4.Trigger a transfer call from the agent to one of the departments (e.g., Support). 5.Observe which number Vapi dials. Expected Result: The call should be transferred to the number associated with the department selected at runtime. Actual Result: The call always goes to the first configured number in the Transfer Call Tool, ignoring the selected department.

Currently, there’s no way to add background sound during a call that uses workflow, which makes the conversation feel unrealistic and the pauses awkward. It would be great to have the same level of voice customization in workflows as we have with agents.

Getting lead from crm workflow & making personalized outbound call. let me explain in bit detail: when lead replies with phone number on my workflow crm. i want to take that leads number, name & city. then make personalized call automatically lead drops via mail on my workflow-Make & n8n is not supporting this simple thing still. Ps: my use case specifically is email back & forth thread. lead catches after pitch mail(i have opener & pitch mail) & sends to webhook like wise.

I am experiencing a bug with Hebrew transcribers specifically on workflows. When I am using the Google and ElevenLabs Hebrew transcribers on workflows, they dont transcribe the call at all (not the ai voice and not mine) and also you can see the transcription window staying empty (and because of that, the workflow is stuck on the first block). However when I use Azure's transcriber it works just fine. Needless to say I tested the Google\ElevenLabs transcribers on the assistants menu when I created a demo assistant and they worked perfectly fine. I prefer to use Google's transcriber over Azure and seems weird that it works on the singular assistant prompt but doesn't work in the workflows. I would really appreciate if you would fix it. (Just for clarification I did the exact same prompt and exact same workflow and changed only the transcriber provider so I am certain it has nothing to do with configurations from my side)

As per your earlier instructions, we have added a dedicated number intended for Vapi to handle customer interactions. However, we're currently facing a routing issue. When a customer directly calls this number, instead of the call being handled by the Vapi agent as expected, it is being immediately routed to GoTo, completely bypassing Vapi. Our intended flow is as follows: Customer calls the configured number. The call should be processed first by the Vapi agent (for IVR, knowledge base interaction, etc.). Only after agent processing, the call should be transferred to the respective department via GoTo. This current behavior is blocking our production flow, as the Vapi agent is not getting triggered at all. Please assist us in ensuring that Vapi is the primary entry point for all inbound calls, and GoTo is only used for post-agent call transfers. Looking forward to your urgent support to resolve this.

I'm currently building a Voice AI Assistant in Spanish, and I’ve run into a challenge with voice formatting, which is enabled by default and currently cannot be toggled off via the dashboard. The existing voice formatting logic appears to be optimized for English, and in Spanish, it often produces outputs that are only compatible with English, leading to awkward or incorrect formatting, especially for numbers, currencies, and other language-specific constructs. While I understand that this behavior can be controlled via the API using voice.chunkPlan.enabled , having a dashboard-level toggle (ideally configurable at the assistant level) would significantly improve usability. Why this matters: Speeds up testing by allowing quick toggling without touching code Makes the setting accessible to non-technical team members Improves support for non-English use cases like Spanish