Grain Studios
Homelab Infrastructure
Two-node local network for AI, automation, and media
Project Details
Challenge
Running AI workloads, home automation, media services, and a voice assistant pipeline requires coordinated infrastructure across multiple machines with zero cloud dependency.
Approach
Designed a two-node architecture: Mac Mini M4 'forge' handles AI inference and worker orchestration, while a 2018 Mac Mini 'sentinel' runs Docker containers for Home Assistant, media services, and Frigate. Tailscale mesh VPN connects everything.
Outcome
Full-stack local infrastructure running 24/7. Voice assistant pipeline with Whisper STT, GLaDOS TTS, and Gemma 4 conversation agent. 15+ Docker services coordinated across nodes.
Metrics
2
Nodes
Forge + Sentinel
15+
Services
Docker containers
0
Cloud
Fully local-first
Tech Stack
DockerHome AssistantTailscaleWhisperOllamaMQTT