Self-Hosted AI Models (Quickstart Tutorial)

Learn how to quickly connect self-hosted models to MindStudio for testing & hobbyists

Prerequisites

Before getting started, ensure you have the following installed on your machine:

• Ollama

• Ngrok (for macOS)

---

Step-by-Step Instructions

1. Start an Ngrok Tunnel

Open a terminal window and run the following command to expose port 11434 to the internet:

ngrok http http://localhost:11434

Copy the forwarding HTTPS URL Ngrok provides (e.g. https://xxxx.ngrok.io). You'll need this later.

---

2. Pull and Serve the AI Model

Open a second terminal window:

a. Pull the llama3.2 model:

ollama pull llama3.2

b. Serve the model with the proper environment variables:

Replace the Ngrok URL below with your actual forwarding address from Step 1:

On Mac / Linux:

OLLAMA_HOST=0.0.0.0:11434 OLLAMA_ORIGINS="https://xxxx.ngrok.io/" ollama serve

On Windows:

First

$Env:OLLAMA_HOST   = "0.0.0.0:11434"
$Env:OLLAMA_ORIGINS = "https://xxxx.ngrok.io/" 

then

ollama serve

This command makes your local Ollama instance accessible through the Ngrok tunnel.

---

3. Add the Self-Hosted Model to MindStudio

1. Go to MindStudio Self-Hosted Models
2. Click "Add Model"

3. Use the following fields:

   • API Name: (Ex: llama3.2 )

   • Endpoint: https://xxxx.ngrok.io/v1 (replace xxxx with your actual Ngrok subdomain)

Your self-hosted AI model is now ready to be used in any MindStudio AI Agent.

---

Using the Model in an Agent

Once added, you can:

• Select your AI Model in the Model Settings tab when editing an AI Agent.

• Use it across workflows, prompts, and block-level overrides in your automation.