AI-Powered Zendesk Support Responses with RAG, OpenAI, and Supabase Knowledge Base

This n8n workflow automates the generation of intelligent support responses for Zendesk tickets using a Retrieval Augmented Generation (RAG) approach. It leverages OpenAI for AI agent capabilities and a Supabase Vector Store as a knowledge base to provide contextually relevant answers.

What it does

This workflow streamlines the process of responding to support tickets by:

1. Receiving a Zendesk Ticket: It listens for incoming Zendesk ticket data via a webhook.
2. Preparing Context: It extracts relevant information from the incoming ticket (e.g., the user's question or issue).
3. Applying AI Agent: It uses an OpenAI-powered AI agent to process the ticket information.
4. Retrieving Knowledge: The AI agent queries a Supabase Vector Store (acting as a knowledge base) to find relevant articles or information.
5. Generating Response: Based on the ticket context and retrieved knowledge, the AI agent generates a comprehensive and helpful support response.
6. Formatting Output: It formats the generated response for further use (e.g., posting back to Zendesk).

Prerequisites/Requirements

To use this workflow, you will need:

• n8n Instance: A running n8n instance.
• Zendesk Account: A Zendesk account to trigger the webhook and potentially post responses.
• OpenAI Account & API Key: For the AI Agent and Embeddings.
• Supabase Project: A Supabase project configured with a Vector Store to act as your knowledge base. This will store your documentation or support articles as embeddings.
• PostgreSQL Database (for Chat Memory): A PostgreSQL database (can be part of your Supabase project) to store chat memory for the AI agent (optional, but recommended for conversational context).

Setup/Usage

1. Import the Workflow:
   • Download the provided JSON file.
   • In your n8n instance, go to "Workflows" and click "New".
   • Click the three dots in the top right corner and select "Import from JSON".
   • Paste the workflow JSON or upload the file.
2. Configure Webhook:
   • Locate the "Webhook" node.
   • Copy the "Webhook URL".
   • In your Zendesk account, set up a webhook that triggers when a new ticket is created or updated, sending the ticket data to this n8n Webhook URL. Ensure the payload includes the ticket subject and description.
3. Configure Credentials:
   • OpenAI Credentials: Configure your OpenAI API key in the "Embeddings OpenAI" and "AI Agent" nodes.
   • Supabase Credentials: Configure your Supabase API URL and API Key in the "Supabase Vector Store" node.
   • Postgres Credentials: If using chat memory, configure your Postgres database credentials in the "Postgres Chat Memory" node.
4. Configure AI Agent:
   • In the "AI Agent" node, review and adjust the prompt and tools as needed to fit your specific support scenarios.
   • Ensure the "Answer questions with a vector store" tool is correctly configured to use your Supabase Vector Store.
5. Configure Supabase Vector Store:
   • In the "Supabase Vector Store" node, specify the table name and column names where your embeddings are stored in Supabase.
   • Ensure your Supabase database has a table with your knowledge base content already embedded and stored as vectors.
6. Activate the Workflow:
   • Once all credentials and configurations are set, activate the workflow.

Now, whenever a new Zendesk ticket triggers the webhook, n8n will process it, retrieve relevant information from your Supabase knowledge base, and generate an AI-powered response. You can then extend the workflow to automatically post this response back to Zendesk or notify a support agent.