Voice-Driven AI Assistant using VAPI and GPT-4.1-mini with Memory

Send VAPI voice requests into n8n with memory and OpenAI for conversational automation

This template shows how to capture voice interactions from VAPI (Voice AI Platform), send them into n8n via a webhook, process them with OpenAI, and maintain context with memory. The result is a conversational AI agent that responds back to VAPI with short, business-focused answers.

✅ What this template does Listens for POST requests from VAPI containing the session ID and user query
Extracts session ID and query for consistent conversation context
Uses OpenAI (GPT-4.1-mini) to generate conversational replies
Adds Memory Buffer Window so each VAPI session maintains history
Returns results to VAPI in the correct JSON response format

👤 Who’s it for Developers and consultants building voice-driven assistants
Businesses wanting to connect VAPI calls into automation workflows
Anyone who needs a scalable voice → AI → automation pipeline

⚙️ How it works Webhook node catches incoming VAPI requests
Set node extracts session_id and user_query from the request body
OpenAI Agent generates short, conversational replies with your business context
Memory node keeps conversation history across turns
Respond to Webhook sends results back to VAPI in the required JSON schema

🔧 Setup instructions

Step 1: Create Function Tool in VAPI In your VAPI dashboard, create a new Function Tool
Name: send_to_n8n
Description: Send user query and session data to n8n workflow
Parameters:
session_id (string, required) – Unique session identifier
user_query (string, required) – The user’s question
Server URL: https://your-n8n-instance.com/webhook/vapi-endpoint

Step 2: Configure Webhook in n8n Add a Webhook node
Set HTTP method to POST
Path: /webhook/vapi-endpoint
Save, activate, and copy the webhook URL
Use this URL in your VAPI Function Tool configuration

Step 3: Create VAPI Assistant In VAPI, create a new Assistant
Add the send_to_n8n Function Tool
Configure the assistant to call this tool on user requests
Test by making a voice query — you should see n8n respond

📦 Requirements An OpenAI API key stored in n8n credentials
A VAPI account with access to Function Tools
A self-hosted or cloud n8n instance with webhook access

🎛 Customization Update the system prompt in the OpenAI Agent node to reflect your brand’s voice
Swap GPT-4.1-mini for another OpenAI model if you need longer or cheaper responses
Extend the workflow by connecting to CRMs, Slack, or databases

📬 Contact
Need help customizing this (e.g., filtering by campaign, connecting to CRMs, or formatting reports)?

📧 rbreen@ynteractive.com
🔗 https://www.linkedin.com/in/robert-breen-29429625/
🌐 https://ynteractive.com