AI Infrastructure — What I'm Running

My current AI stack: Ollama running Hermes3-Mythos:70b as primary inference model on Node 2 (whitebox server, NVMe SSD), ChromaDB for vector memory, ARIA agent daemon on a 60-second cycle loop, and a task queue system for asynchronous agent work. All inference is local — no Anthropic API charges for routine operations, no data leaving the building for internal tasks. Claude Code handles complex multi-step tasks via Max plan; local models handle the routine 80%.