n8n Clinic WhatsApp Support with GPT-4o Mini, Pinecone RAG, and Gmail Alerts

This n8n workflow automates a clinic's WhatsApp support system by leveraging AI for intelligent responses, a vector database for knowledge retrieval, and email for human intervention when needed. It acts as a smart front-line for patient inquiries, providing immediate answers and escalating complex cases.

What it does

This workflow is designed to:

1. Receive WhatsApp Messages: It listens for incoming WhatsApp messages via a webhook, acting as the entry point for patient queries.
2. Process Messages with AI: It uses an AI Agent (powered by an OpenAI Chat Model) to understand the user's query and generate a response.
3. Retrieve Information from Pinecone: The AI Agent is equipped with a tool to query a Pinecone Vector Store, enabling it to retrieve relevant information from a knowledge base (e.g., clinic FAQs, service details) for accurate answers.
4. Manage Conversation History: A simple memory component helps the AI maintain context within a conversation, leading to more coherent and helpful interactions.
5. Prepare Data for Response: It processes the AI's response and formats it appropriately.
6. Send Gmail Alerts (Conditional): An "If" node likely checks for specific conditions (e.g., if the AI couldn't answer, or if a human intervention is requested) to trigger a Gmail notification to a staff member. This ensures that critical or complex queries are not missed and can be handled by a human.
7. Respond to WhatsApp: It sends the AI-generated or escalated response back to the user via the webhook.
8. Knowledge Base Management (Manual Trigger): A separate branch, manually triggered, allows for updating the Pinecone vector store. This involves:
   • Loading documents (e.g., from Google Docs).
   • Splitting the text into smaller chunks for embedding.
   • Generating embeddings using OpenAI.
   • Storing these embeddings in Pinecone.
   • (Potentially) logging this update to Google Sheets.

Prerequisites/Requirements

To use this workflow, you will need:

• n8n Instance: A running n8n instance.
• WhatsApp Webhook: A configured WhatsApp integration that can send messages to an n8n webhook.
• OpenAI API Key: For the AI Chat Model (GPT-4o Mini is suggested by the directory name, but the JSON uses a generic OpenAI Chat Model) and Embeddings.
• Pinecone Account & API Key: To set up and interact with your vector store for RAG (Retrieval Augmented Generation).
• Google Account: For Google Docs (to manage your knowledge base) and Gmail (for sending alerts).
• Credentials: Appropriate n8n credentials configured for OpenAI, Pinecone, Google Sheets, and Gmail.

Setup/Usage

1. Import the Workflow: Download the provided JSON and import it into your n8n instance.
2. Configure Credentials:
   • Set up your OpenAI API Key credential.
   • Set up your Pinecone API Key and environment credential.
   • Set up your Google OAuth2 credentials for Google Sheets and Gmail.
3. Configure Webhooks:
   • Copy the URL from the "Webhook" trigger node and configure your WhatsApp integration to send incoming messages to this URL.
4. Set up Pinecone Vector Store:
   • Ensure your Pinecone index is created and correctly configured.
   • In the "Pinecone Vector Store" node, specify your index name and other relevant details.
5. Manage Knowledge Base:
   • In the "Google Docs" node, specify the ID of the Google Doc(s) containing your clinic's knowledge base.
   • Manually execute the workflow starting from the "Manual Trigger" node connected to the "Google Docs" branch to populate or update your Pinecone vector store with your clinic's information.
6. Configure Gmail Alerts:
   • In the "Gmail" node, specify the recipient email address for alerts and customize the email content as needed.
   • Adjust the conditions in the "If" node to determine when an alert should be sent.
7. Activate the Workflow: Once all configurations are complete, activate the main workflow to start processing WhatsApp messages.

This workflow provides a robust and intelligent system for handling clinic support, reducing manual workload and improving patient response times.