The Forge Project

A full-fledged home AI system built on a dedicated server. Forge provides intelligent home automation, monitoring, and voice assistance through local processing. Based on opencode, the system uses a variety of LLM models - both cloud-based and self-hosted - as we experiment to find an optimal AI “brain” for the environment.

Software

Forge is built using OpenCode as the core development platform. I’m currently running on open source GLM-4.7 model for inference, with experiments using various other LLM models as we iterate on the system.

Hardware

Forge runs on a Beelink SER8 mini PC:

  • RAM: 32GB
  • Storage: 1TB NVMe SSD
  • OS: Ubuntu 24.04
  • CPU: x86_64

The SER8 provides quiet, efficient performance perfect for 24/7 AI workloads. Since there’s no keyboard, mouse, or monitor connected, the system idles around 4.5W and draws around 55W at load.

Note: This project is currently under active development and is not yet ready for external users or contributors.

Tom Allen

Chief Information and Security Officer at Railspire.

I build software, hardware, and sometimes weird combinations of the two.


2025-01-30