Connecting Supabase and n8n to Build a Vector based RAG AI Chatbot

Introduction

This approach, leveraging Supabase and n8n, offers significant scalability for enterprise use. By efficiently handling large volumes of data through vectorized storage and leveraging AI agents for intelligent querying, it can be expanded to accommodate more complex data sets, multiple sources, and higher user demands. As businesses grow, the system can be easily scaled to manage increasing data complexity, ensuring high performance and accuracy while maintaining flexibility in managing workflows and automating processes. This makes it an ideal solution for enterprises looking to enhance their knowledge management and data-driven decision-making at scale.

• Overview: This tutorial walks you through integrating Supabase and n8n to manage data for a custom knowledge base application. The goal is to demonstrate how to use Supabase for storing vectorized data, creating a search system, and utilizing n8n to manage workflows.
• Prerequisites: Ensure you have n8n installed locally and a Supabase account created. The example will demonstrate extracting data from the Valve Employee Handbook and storing it in a Supabase Vector database.

Step 1: Set Up Supabase

1.1 Create a Supabase Account

• Sign up for a free Supabase account at Supabase.
• Confirm your email and log in.

1.2 Create a New Project

• After logging in, click "New Project" and enter the project details: Project name: Supabase Vector Project Password: Choose a secure password and copy it for future use.

1.3 Access API and Database Credentials

• Once your project is created, navigate to Settings > Database.
• Click on Connect and copy the following: Host (e.g., AWS region address) Username and Password (stored during project setup) Port (usually 5432 for Postgres, but confirm from the transaction pooler section).

1.4 Create Vector Store in Supabase

• Go to the SQL Editor in Supabase, paste the provided SQL script to create a vector store, and run it.
• This will create a documents table for storing your vectorized data.

Step 2: Set Up n8n Workflow

2.1 Install n8n Locally

• Follow the n8n installation guide to set up n8n on your machine.

2.2 Configure n8n with Supabase

• Open the n8n dashboard (localhost:5678).
• Under Credentials, create a new Supabase connection: Enter the Host, Database, Username, Password, and Port (from Step 1.3).

2.3 Set Up PostgreSQL Memory in n8n

• Add the Postgres Chat Memory node to store conversation history.
• Configure the credentials to connect to the Postgres database created in Supabase.

2.4 Add Supabase Vector Store Node

• Add the Supabase Vector Store node for storing embeddings.
• Connect this to your Postgres Chat Memory node.

Step 3: Add Google Drive for Data Input

3.1 Connect Google Drive Node

• Add a Google Drive node to download the Valve Employee Handbook.
• Authenticate your Google Drive account in n8n.
• Select the file from Google Drive to use as your input data (the Valve Employee Handbook).

3.2 Text Splitter Node

• Add a Text Splitter node to divide the document into smaller, manageable chunks.
• Configure the text splitter to handle the structure of the Valve Employee Handbook.

3.3 Embed Data Using OpenAI

• Add the OpenAI Embedding node to generate vector embeddings of each chunk of text.
• Use these embeddings to store the document data in the Supabase Vector Store.

Step 4: Test the Workflow

4.1 Trigger AI Agent with Chat Node

• Add an AI Agent to respond to queries about the Valve Employee Handbook.
• Configure the agent to use both Postgres chat memory and the Supabase vector store.

4.2 Test the AI Agent

• Test the AI Agent by querying it about sections of the Valve Employee Handbook (e.g., "What are the employee benefits?").
• Verify that the agent retrieves accurate responses by referencing the stored vectorized data.

4.3 Inspect Supabase for Data Storage

• Go to the Table Editor in Supabase and inspect the documents table for stored vector data.
• Verify that the vectorized document data is stored correctly.

Step 5: Advanced Customizations (Optional)

5.1 Customizing the Agent’s Role

• Modify the agent’s role in n8n to refine its responses (e.g., tell it to focus on employee benefits or policies).

5.2 Adding More Data

• Add additional documents or sections to the Valve Employee Handbook and re-run the workflow to incorporate the new information into the vector database.

5.3 Scalability Considerations

• Explore options for scaling your vector database, including self-hosting Supabase or considering other vector stores.

Conclusion

• Recap of the key steps in integrating Supabase with n8n for a knowledge base search system.
• How to use this setup for creating AI-powered agents with memory and vector search capabilities.

Next Steps

• Experiment with different types of documents and data inputs.
• Explore deeper customizations for your AI agent's behavior in n8n.

This tutorial gives you a comprehensive understanding of how to work with Supabase and n8n for AI-powered knowledge base management. You can now build more complex workflows, add multiple data sources, and create sophisticated querying systems using vectorized data.

Paul Hankin is the author of:

AI Adoption: A Practical Guide for Business

and

AI and Law: Navigating the Future