How to Set Up Supabase for Local AI Agents: A Step-by-Step Guide

Introduction

Supabase has quickly become one of the most popular database solutions for AI applications. Built on PostgreSQL, it offers powerful features like PGVector for vector storage, authentication, and object storage making it a perfect choice for AI agents and Retrieval-Augmented Generation (RAG) workflows.

In this guide show you how to set up Supabase locally using Docker and integrate it into your Local AI Package for a fully self hosted AI agent experience. Let’s dive in!

Why Supabase for Local AI?

Supabase is an open source Firebase alternative, and it’s gaining traction as a goto database for AI applications. Here’s why:

• PostgreSQL Under the Hood – Advanced SQL support with JSONB storage.
• PGVector for RAG – Turn your database into a vector store for embedding based retrieval.
• Built-in Authentication – Easily manage user authentication without extra setup.
• Object Storage – Store and retrieve files efficiently.
• Docker Self-Hosting – Run it locally for privacy, control, and performance.

By integrating Supabase into your Local AI Package, you can have a complete AI-powered environment running entirely on your machine. No cloud dependencies just 100% local AI.

Prerequisites

Before we start, ensure you have the following installed:

1. Python (For running the setup script)
2. Git or GitHub Desktop (For cloning the repository)
3. Docker & Docker Compose (For containerized AI services)

If you don’t have these installed yet, grab them from their official sources:

• Python
• Git
• Docker

Step 1: Clone the Local AI Package Repository

Run the following command to clone the Local AI Package repository:

git clone https://github.com/your-repo/local-ai-package.git
cd local-ai-package        

This repository contains all the configuration files needed to launch a fully integrated AI stack using Supabase, n8n, Ollama, and Flowise.

Step 2: Configure Environment Variables

Supabase requires several environment variables. Let’s create an .env file:

cp .env.example .env
nano .env  # Or use any text editor        

Set the following required variables:

POSTGRES_PASSWORD=yourpassword
SUPABASE_DASHBOARD_USERNAME=admin
SUPABASE_DASHBOARD_PASSWORD=admin123
POOLER_TENANT_ID=1        

To generate JWT secrets for authentication, use Supabase’s tool:

• Generate JWT Keys

Copy the generated keys into your .env file:

JWT_SECRET=your_generated_secret
SUPABASE_ANON_KEY=your_generated_anon_key
SUPABASE_SERVICE_ROLE_KEY=your_generated_service_role_key        

Save and close the file.

Step 3: Start Supabase and AI Stack with Docker

Run the following command to launch all services:

bash start_services.sh --profile=gpu # For NVIDIA GPUs
bash start_services.sh --profile=cpu # For CPU-based setup        

This script will:

• Clone the Supabase repository (if not already present)
• Start all containers: n8n, Flowise, Ollama, Supabase
• Set up networking between services

Check running containers with:

docker ps        

If everything is working, you should see multiple containers running.

Step 4: Access Your Local AI Services

Now that everything is running, let’s check out the different services:

• Supabase Dashboard: https://localhost:8000 (Use the admin credentials from your .env file)
• n8n (Automation Platform): https://localhost:5678
• Flowise (Low Code AI Builder): https://localhost:3001
• Ollama (LLM Runner): Accessible via API on port 11434

You now have a fully functional local AI stack!

Step 5: Setting Up a RAG AI Agent

To build an AI Agent with RAG, follow these steps:

1. Create a Vector Table in Supabase:
2. Ingest Data Using n8n:
3. Query Supabase for Relevant Context:

Example Query for Vector Search:

SELECT content FROM documents
ORDER BY embedding <-> 'your_vector_here'
LIMIT 5;        

And that’s it! You now have a fully local AI agent powered by Supabase + Ollama!

Troubleshooting

If you run into issues, check the Docker logs:

docker logs supabase-db  # Check database logs
docker logs local-ai-package  # Check AI services        

Common fixes:

• Port Conflicts? Ensure ports 8000, 5678, and 11434 are free.
• Containers Not Starting? Run docker compose down and restart services.
• Vector Queries Failing? Make sure PGVector is installed in Supabase.

Conclusion

By integrating Supabase into your Local AI Package, you now have a robust AI infrastructure running entirely on your machine. Whether you’re building chatbots, AI-powered search, or automated workflows, Supabase makes it seamless to manage both structured and vector data.

Next Steps:

• Add user authentication for secure access.
• Deploy on-premise or to a private cloud.
• Extend with more AI models and workflow automation.

video tutorial