n8n Gmail Assistant with Full Gmail History RAG using OpenAI

This n8n workflow creates a powerful AI assistant that can answer questions based on your entire Gmail history. It leverages Retrieval Augmented Generation (RAG) by fetching your emails, embedding them, storing them in a vector database (Qdrant), and then using an OpenAI Chat Model to answer queries with context from your emails.

What it does

This workflow automates the following steps:

1. Triggers on Chat Message: The workflow is activated when a chat message is received (likely from an n8n chat interface or a connected chat platform).
2. Fetches Gmail History: It retrieves all emails from your configured Gmail account.
3. Processes Emails in Batches: Emails are processed in batches to manage large datasets efficiently.
4. Extracts and Prepares Email Content: A Code node extracts relevant text content from each email.
5. Splits Text into Chunks: The extracted email content is split into smaller, manageable chunks using a Character Text Splitter, which is crucial for effective RAG.
6. Generates Embeddings: OpenAI Embeddings are created for each text chunk, converting them into numerical vectors.
7. Stores Embeddings in Qdrant: These embeddings, along with their original text, are stored in a Qdrant Vector Store, making them searchable by semantic similarity.
8. Answers Questions with AI Agent: An AI Agent (powered by an OpenAI Chat Model) uses the Qdrant Vector Store as a tool to retrieve relevant email context when answering user questions.
9. Maintains Conversation History: A Simple Memory node helps the AI Agent remember previous turns in the conversation, providing a more coherent chat experience.

Prerequisites/Requirements

To use this workflow, you will need:

• n8n Instance: A running n8n instance.
• Gmail Account: A Google account with Gmail enabled, and an n8n Google OAuth credential configured for Gmail access.
• OpenAI API Key: An OpenAI API key with access to embedding models (e.g., text-embedding-ada-002) and chat models (e.g., gpt-3.5-turbo, gpt-4).
• Qdrant Instance: Access to a Qdrant vector database instance (either self-hosted or cloud-based). You'll need to configure an n8n Qdrant credential.

Setup/Usage

1. Import the workflow: Download the JSON provided and import it into your n8n instance.
2. Configure Credentials:
   • Gmail: Set up a Google OAuth2 credential for Gmail. Ensure it has the necessary permissions to read your emails.
   • OpenAI: Configure an OpenAI API key credential.
   • Qdrant: Set up a Qdrant credential, providing your Qdrant host and API key.
3. Initial Data Loading (Manual Trigger):
   • The workflow includes a "Manual Trigger" and a "Gmail Trigger". The "Gmail Trigger" is for real-time updates, but for the initial loading of your entire email history, you might want to temporarily connect the "Manual Trigger" to the "Gmail" node (Node ID 356) and execute the workflow once. This will fetch and embed your existing emails into Qdrant.
   • Important: After the initial load, reconnect the "Gmail Trigger" to the "Loop Over Items" node (Node ID 39) to process new emails.
4. Configure Nodes:
   • Gmail Trigger (Node ID 824): Ensure this is configured to listen for new emails in your desired inbox.
   • Gmail (Node ID 356): This node is used to fetch the full history. Configure it with your Gmail credential and any specific filters if you don't want to process all emails.
   • Code (Node ID 834): Review the JavaScript code to understand how email content is extracted. You might need to adjust it based on the specific parts of emails you want to use for RAG.
   • Character Text Splitter (Node ID 1189): Adjust chunkSize and chunkOverlap parameters if needed to optimize how your email text is broken down for embeddings.
   • Embeddings OpenAI (Node ID 1141): Select your OpenAI credential and the desired embedding model.
   • Qdrant Vector Store (Node ID 1248): Select your Qdrant credential and specify a collection name. This is where your email embeddings will be stored.
   • OpenAI Chat Model (Node ID 1153): Select your OpenAI credential and the desired chat model (e.g., gpt-4, gpt-3.5-turbo).
   • AI Agent (Node ID 1119): This node orchestrates the RAG process. Ensure it's correctly linked to the "Qdrant Vector Store" as a tool and the "OpenAI Chat Model" as its language model.
5. Activate the Workflow: Once all credentials and nodes are configured, activate the workflow.

Now, when you send a message to the "Chat Trigger", the AI Agent will use your Gmail history stored in Qdrant to provide context-aware answers.