A hands-on walkthrough showing how to run Ubuntu Core 26 in a Multipass virtual machine and turn it into a local AI inference appliance. Using the gemma4 snap, you can install an OpenAI-compatible inference server and WebUI inside the VM, expose the endpoints to your host machine, and interact with a local LLM via browser or API. The post also explains how this VM-based experiment maps to a real production Ubuntu Core image, covering snap-based application delivery, transactional updates, rollback support, and fleet management via Landscape.
Table of contents
Why start with Ubuntu Core in a VM?Turning the VM into a local AI applianceChecking the inference endpointWhat this demonstratesFrom local experiment to production imageManaging appliances over timeWhat’s next?117 Impressions