Autopilot - Running addon with local OLLAMA LLM

Addons
1

I wanted to share my working configuration for running the Autopilot addon with a local Ollama instance instead of OpenAI API. This might be useful for those who want to use local LLMs for privacy or cost reasons.

Environment:

  • Cockpit CMS Pro with Autopilot addon

  • Ollama server running locally

Step 1: Configure config/config.php

Add the following to your config file:

'autopilot' => [
    'openAI' => [
        'apiKey' => 'ollama',
        'chatModel' => 'gemma3:latest',
        'useTools' => false,
        'baseUri'=> 'http://your-ollama-server-ip:port/v1'
    ]
],

Notes:

  • apiKey can be any non-empty string (Ollama doesn’t require authentication)

  • chatModel - I’m using gemma3:latest, but you can use any model available in your Ollama instance

  • useTools must be set to false (tool calling is not supported by most local models)

  • baseUri - point this to your Ollama server’s OpenAI-compatible endpoint

Step 2: Modify addons/Autopilot/Controller/Autopilot.php

The addon has useTools hardcoded to true by default. To make it configurable, modify line 52:

Before:

$useTools = $this->param('useTools', true);

After:

$configUseTools = $this->app->retrieve('autopilot/openAI/useTools', true);
$useTools = $this->param('useTools', $configUseTools);

This change allows the useTools setting from the config file to be respected, while still maintaining backward compatibility (defaults to true if not specified).

Result: The Autopilot addon now works with my local Ollama instance running Gemma 3. Response times depend on your hardware, but it’s a great solution for local development or privacy-focused deployments.

image
image1096×658 46 KB

1 Like