GPU Documentation

This folder records GPU roles, Windows development evidence, and the experimental Linux/Proxmox compute and telemetry work.

✅ Working — Windows development environment

Vulkan inference and GPU telemetry have been exercised on the RX 6800 XT.

⚠️ Not yet validated — target host

Live Linux GPU telemetry and ROCm/HIP inference on the intended Proxmox host still require validation. Concurrent AI profiles remain planned.

Guide

• GPU Integration Plan — planned hardware roles, Windows Vulkan and DirectML testing, Linux telemetry, and ROCm/Proxmox compatibility gates.

Status and boundaries

Windows Vulkan inference and Windows GPU telemetry have been exercised. The Linux amdgpu telemetry provider is implemented and fixture-tested, but live Linux readings and ROCm/HIP inference on the intended Proxmox host remain unverified. A resident Utility AI plus Household/Maintenance profile models is the proposed single-GPU architecture, but concurrent residency and model routing are not implemented or validated. Windows GPU results do not establish Linux or Proxmox support.