Gareth D

I've an n8n workflow that is triggered from VAPI as a tool call. The hits an N8n workflow and makes its way to the response node. This sends the data response back to VAPI. But.....the agent does not get the data. The VAPI logs shows the response has arrived there with a 200 + the message body as attached. It leaves n8n's response node like the 2nd screenshot. Has anyone come across this issue before / have any idea what the issue is? As its showing 200 and the response data on the Vapi Call Logs am a bit confused.

New comment Jan 11

Mutiso Muathime

Dec '24 •

Support Needed 💻

Vapi AI Tool Calling Issue

Hi guys, hope you're well. So I'm having issues with my Vapi Assistant. The assisant is for a lawn mowing service enabling users to book in appointments. I'm using make.com for the booking. The issue is the vapi assistant doesn't wait long enough to get a response from make.com, if there is no delay from make.com it works just fine and is able to tell the user that their appointment has been booked, but if there is a delay, the Vapi assistant tells the user it can't book in the appointment, despite succesful booking of the appointment from Make.com, I haven't turned on the async feature on vapi, therefore the assistant is meant to wait for a response from the tool call. Here's a loom video, on the same, any help would be greatly appreciated: https://www.loom.com/share/7baa003c22e742d58a6c690a09f4c06f?sid=db3f3a47-2c54-4571-bfc3-34d469ec2470

New comment Jul 15

Gareth D

0 likes • Dec '24

What is happening in the make automation? I know you said it is booking successfully, but does the webhook response module in make generate the correct json to send back to VAPI each time its triggered? You mention the agent isn't able to process, but is VAPI receiving anything as a response from make?

Gareth D

0 likes • Dec '24

Also have you tried extending the 20 second timeout on Vapi for the webhook server request?

Gareth D

Dec '24 •

Support Needed 💻

Setting up Squads on ⚡Vapi?⚡

Question for anyone who has set up squads: Is it possible to create a top level "manager" voice agent, which can then route the caller through to 4-5 different niche voice agents (each with access to different tools)? And run the agentic system from 1 phone number?

New comment Jul 15

Gareth D

1 like • Dec '24

@Sam Rexford My experience is that one Ai with lots of tools = Lots of things that could potentially break/conflict. The AI can get confused about what its meant to be doing and if one element fails, the whole process can fail. Adding or removing one sentence in the prompt can sometimes affect whether the AI runs its tool - it can be a bit fragile. A modular system with a manager that routes to 5 agents each with just 1 tool with 1 defined task would be more efficient to set up and test. Can swap and upgrade the tool for each individual agent, without being worried about breaking the whole process. That's the theory anyway 😊

Brendan Jowett

Dec '24 •

Announcements

How To Build ElevenLabs Conversational AI Agents

ElevenLabs is the most popular AI voice model provider on the market, and they have just released their own AI agent creation playground. This means we can now build AI voice agents directly from ElevenLabs without any middleman providers. In this video, I will dive straight into building an AI voice agent using their platform. Please leave your comments on the video!

New comment Dec '24

Gareth D

0 likes • Dec '24

@Thys van Zyl They have added a new phone number feature so you can import your number, so assume using ngrok is no longer needed.

Gareth D

1 like • Dec '24

@Charles Amponsah Agreed 👍 they need to be transparent about pricing and list the cost of each call.

1-8 of 8