Configuring a local model in APEX Running on VirtualBox VM

Now that APEX 26.1 is here, I'd like to test a lot of things using an agent to generate Applications, but it's always nice to save some coins on the way. So I want to install a couple of local models for that.

I'm going to use Ollama, so I went to Download Ollama and got the version for my the OS I'm using (Windows).

Installation is really simple, using a wizard. After that, you just can go to a browser and check Ollama servers is running in localhost under port 11434.

I'm stalling 2 small models that seems to be good enough to start and can be executed within my hardware, and doing so takes just these lines:

ollama pull deepseek-r1:8b

ollama pull llama3.1:8b

and then using Ollama list we can see all the models installed

Now, I want to use it inside the VM, so I checked the network configurations in VirtualBox and there is an IP address from the Virtual Host-Only ethernet adapter that is manually configured, so that one will be static.

Then in the WM browser, if we insert that IP and the Ollama server port, it works.

After testing that, then let's go to APEX. Under APP Builder -> Workspace Utilities. Go to Generative AI section, and then click the "Create" button on the upper right.

In the Select list, pick Ollama as service provider.

Fill the info in the form like this, and click "Test Connection"

Name: Ollama

Used by App Builder: Enabled (You can change this later)

Default for New Apps: Enabled (You can change this later)

Base URL:HTTP://YOUR#VB#IP:11434

Credential: keep the option - Create New -

API Key: Ollama

Static ID: Automatically filled (ollama)

AI Model: llama3.1:8b

Then click "Test Connection". After a successful test, click "Create"

Now we have a Generative AI Service with a local model attached to your APEX instance