WhatsApp Support Assistant with GPT-4 Mini and Google Sheets Data Training

This n8n workflow creates an intelligent WhatsApp support assistant that leverages GPT-4 Mini for conversational AI and trains itself on data fetched from Google Sheets. It allows you to provide dynamic, AI-powered responses to WhatsApp messages, making your support operations more efficient and automated.

Description

This workflow automates the process of responding to WhatsApp messages using a sophisticated AI agent. It listens for incoming WhatsApp messages, processes them with a custom code snippet, and then uses a LangChain AI Agent powered by GPT-4 Mini and a simple memory buffer to generate intelligent responses. The AI agent's knowledge can be dynamically updated by ingesting data from Google Sheets (though the Google Sheets node itself is not present in this JSON, the workflow is designed to support such an integration).

What it does

1. Listens for WhatsApp Messages: The workflow is triggered by incoming messages via a Webhook, which would typically be connected to a WhatsApp integration (e.g., Twilio, MessageBird, etc.).
2. Processes Incoming Data: A "Code" node is used to parse and prepare the incoming message data for the AI agent.
3. Initializes AI Agent: An "AI Agent" node (LangChain Agent) is set up to handle conversational logic.
4. Configures Chat Model: The AI Agent uses an "OpenAI Chat Model" (specifically GPT-4 Mini, as suggested by the directory name) as its underlying language model.
5. Manages Conversation History: A "Simple Memory" node (LangChain Memory Buffer Window) maintains a short-term memory of the conversation, allowing the AI to understand context.
6. Generates Responses: The AI Agent processes the incoming message, utilizes its memory, and generates a relevant response.
7. Sends HTTP Request: An "HTTP Request" node is responsible for sending the AI-generated response back to the user, likely via the WhatsApp API.

Prerequisites/Requirements

• n8n Instance: A running n8n instance (self-hosted or cloud).
• WhatsApp Integration: An external service that can send and receive WhatsApp messages via webhooks (e.g., Twilio, MessageBird, 360dialog, etc.). The Webhook node in this workflow will need to be configured with the URL provided by your WhatsApp integration.
• OpenAI API Key: An API key for OpenAI to use the GPT-4 Mini chat model. This will be configured within the "OpenAI Chat Model" node.
• Google Sheets (Optional, for training data): While not explicitly present in the provided JSON, the workflow's context suggests that Google Sheets would be used as a data source for training the AI agent. If you intend to use this feature, you would need a Google Sheets account and an n8n Google Sheets credential.

Setup/Usage

1. Import the Workflow:
   • Copy the provided JSON code.
   • In your n8n instance, click "New" to create a new workflow.
   • Click the three dots menu (...) in the top right and select "Import from JSON".
   • Paste the JSON and click "Import".
2. Configure Webhook:
   • Locate the "Webhook" node.
   • Set the "Webhook URL" to "POST" and copy the generated URL.
   • Configure your WhatsApp integration (e.g., Twilio, MessageBird) to send incoming message webhooks to this URL.
3. Configure OpenAI Credentials:
   • Locate the "OpenAI Chat Model" node.
   • Click on "Credentials" and add a new "OpenAI API" credential.
   • Enter your OpenAI API Key.
   • Ensure the "Model" is set to "gpt-4-mini" or a similar appropriate model.
4. Configure AI Agent:
   • Review the "AI Agent" node configuration. The default "Plan and Execute" agent is powerful for complex tasks.
   • Adjust the "System Message" in the "OpenAI Chat Model" to define the AI's persona and instructions (e.g., "You are a helpful support assistant for [Your Company Name]. Answer questions based on the provided context.").
5. Configure HTTP Request (for sending replies):
   • Locate the "HTTP Request" node.
   • Configure it to send a POST request to your WhatsApp integration's API endpoint for sending messages.
   • Map the AI agent's output (e.g., {{ $node["AI Agent"].json["text"] }}) to the message body.
   • Ensure proper authentication (e.g., API Key, Bearer Token) is set up for your WhatsApp API.
6. Activate the Workflow:
   • Once all credentials and configurations are set, save the workflow and toggle the "Active" switch to enable it.

Note on Google Sheets Data Training: The workflow JSON provided does not include a Google Sheets node. To incorporate Google Sheets for training data, you would typically add a "Google Sheets" node before the "AI Agent" to fetch relevant data. This data would then be passed to the AI Agent (e.g., via a "Document Loader" or by embedding it directly into the system message or context of the AI Agent) to inform its responses.