Gmail Customer Support Auto-Responder with Ollama LLM and Pinecone RAG

This n8n workflow automates the process of responding to incoming customer support emails received via Gmail. It leverages an Ollama Large Language Model (LLM) for generating responses and integrates with a Pinecone Vector Store for Retrieval Augmented Generation (RAG) to provide context-aware and accurate answers.

The workflow aims to streamline customer support operations by automatically drafting intelligent replies, reducing manual effort and improving response times.

What it does

1. Monitors Incoming Emails: Listens for new emails in a specified Gmail inbox.
2. Classifies Email Intent: Uses an AI Text Classifier (powered by Ollama) to categorize the incoming email's intent (e.g., "support inquiry", "sales question", "general feedback"). This helps in tailoring the response strategy.
3. Retrieves Contextual Information (RAG): Based on the classified intent and email content, it queries a Pinecone Vector Store to retrieve relevant information from a knowledge base. This ensures the LLM has access to up-to-date and specific company information.
4. Generates AI Response: An Ollama Chat Model, guided by the retrieved context and the email's content, generates a draft response. A Simple Memory component helps maintain conversational context if needed.
5. Drafts Gmail Reply: Creates a draft reply in Gmail with the AI-generated content, ready for a human agent to review and send.

Prerequisites/Requirements

To use this workflow, you will need:

• n8n Instance: A running n8n instance.
• Gmail Account: A Gmail account configured as a credential in n8n for sending and receiving emails.
• Ollama Installation: An Ollama server running locally or accessible from your n8n instance, with the desired language model(s) downloaded (e.g., llama2, mistral).
• Pinecone Account: A Pinecone account with an existing index containing your knowledge base or relevant customer support documentation.
• n8n Langchain Nodes: Ensure the @n8n/n8n-nodes-langchain package is installed in your n8n instance.

Setup/Usage

1. Import the Workflow:
   • Download the provided JSON file.
   • In your n8n instance, click "New Workflow" and then "Import from JSON".
   • Paste the JSON content or upload the file.
2. Configure Credentials:
   • Gmail Trigger & Gmail Node: Set up your Gmail OAuth2 credentials.
   • Ollama Chat Model & Ollama Model & Embeddings Ollama: Configure the Ollama credential to point to your running Ollama instance (e.g., http://localhost:11434). Specify the model name you have downloaded (e.g., llama2).
   • Pinecone Vector Store: Set up your Pinecone API Key and environment credentials.
3. Customize Nodes:
   • Gmail Trigger: Adjust the "Label" or "Query" to filter which incoming emails trigger the workflow if needed.
   • Text Classifier: Fine-tune the classification categories and prompts to match your specific needs.
   • Pinecone Vector Store: Ensure the "Index Name" and any query parameters are correctly configured to retrieve relevant information from your Pinecone index.
   • AI Agent / Ollama Chat Model: Modify the system prompts and instructions to guide the LLM's response generation according to your brand voice and support policies.
4. Activate the Workflow: Once all configurations are complete, activate the workflow. It will start listening for new emails and process them automatically.