How do I set the URL for using Ollama in Agent Zero?

[factor3]

I would like to use Agent Zero (A0) with some Ollama models, and I am having a problem telling A0 where to go to get the models.

According to everyone I have looked at on the Internet, I am supposed to perform the following steps to make A0 use Ollama:

Open the Settings panel in the Agent Zero Web UI.
Select Ollama as the provider for your Chat, Utility, and Embedding models.
Set the Base URL to http://host.docker.internal:11434.
Enter the exact name of your downloaded model.

There are a few problems with these instructions:

1. There are no settings for the Chat, model. There is a "Main" model
2. In order to edit the models, I must click on an Edit Presets button.
3. Once I have clicked on the Edit Presets, yes, I can select Ollama. Unfortunately, there is no place to put in a Base URL.

Being able to put in a Base URL is important, because the machine that is running my Ollama is not the same machine that I am using to run A0. I cannot utilize Agent Zero without pointing it to the Ollama machine.

Of course, without being able tyo point A0 to the Ollama machine, I cannot access the models that are on it.

So: considering that the instructions for using Ollama with A0 are incorrect (out of date?), how do I tell A) where to get its Ollama models?

[sknoyon324234234234]

Yeah, those instructions are from before the presets UI. The field still exists, it's just hidden now:

1. Edit Presets → pick your preset
2. In the Main Model section, set provider to Ollama and enter the model name (e.g. qwen2.5:7b)
3. Right under that, click Advanced Settings (it's collapsed by default)
4. The first field there is API base URL. Put in your Ollama machine, e.g. http://192.168.1.50:11434
5. Do the same for Utility Model and Embedding Model if they're on Ollama too, then save

Set it on each one. An empty API base means "use the provider default", and for Ollama that's http://host.docker.internal:11434, which is exactly the address that won't work for you.

Two things that usually bite with a remote Ollama box:

• By default Ollama only listens on 127.0.0.1, so other machines can't reach it. On the Ollama machine set OLLAMA_HOST=0.0.0.0 (on Linux: systemctl edit ollama → Environment="OLLAMA_HOST=0.0.0.0", then restart) and make sure port 11434 isn't firewalled.
• Quick test from the A0 machine (or inside the A0 container): curl http://192.168.1.50:11434/api/tags. If that lists your models, A0 will see them too.

[factor3]

Thank you for your information abut the UI. I can use A0 with Ollama now.

As for Ollama's setup, I already had it set up that way. A0 is not the first application that I am using with Ollama. Several applications -- including applications that I have written -- use Ollama fr accessing AI.