Connect your browser chat to a local Ollama model

If you already run a local LLM with Ollama, you can use that model in Naidan’s chat. This example connects to Ollama already running on your device. Naidan does not start the Ollama process.

First, open Naidan. If first-run setup appears, close it with the × at the top right, then follow step 1.

Have at least one Ollama model available before connecting. You will choose the server address in the connection settings.

1. Open Settings and Connection

Select Settings at the bottom of Naidan’s sidebar. The Settings panel opens on Connection by default; select Connection if you were on another tab. The first scene begins at the sidebar button and ends inside the Connection panel.

The sidebar Settings button opens the Connection settings.

2. Choose Ollama and its local address

In that Connection panel, set API Provider to Ollama. Beside Endpoint URL, choose Ollama (local) to fill in http://localhost:11434, or type your server’s actual address. The preset changes the provider and URL together.

The provider changes to Ollama; the local preset fills the URL in the same panel.

3. Check the connection and model list

Keep that Ollama URL and select Check Connection. The button shows a spinner while Naidan fetches the model list, then briefly says Connected on success. Naidan also checks localhost addresses automatically when the provider or URL changes in Settings. If the check fails, verify the service, address and any browser CORS or origin error before continuing.

Check Connection loads Ollama’s model list, then briefly confirms success.

4. Pick a default model and save

After the successful check, move down to Model Selection. Choose one of the models returned by Ollama in Default Model, then select Save Changes. The default applies to existing and new conversations that inherit it; a chat or group can override the model. This example shows gemma3n:e2b, the sample model named in Naidan’s built-in server guide; choose a model that actually exists on your server.

The selected model and saved state follow the successful connection; the model name is an example.

When the model does not appear

Check that Ollama has a model available, then use Check Connection again. If the page was opened from file://, the browser may send Origin: null and the Ollama service may need to allow that origin. A hosted page can also be limited by its Content Security Policy. The browser’s network error helps identify which boundary blocked the request.