{
  "schema_version": 1,
  "title": "Local AI Infrastructure Lab",
  "author": "Deshimaru Sakaguchi",
  "canonical": "https://deshimarusakaguchi.com/lab/local-ai-cluster/",
  "language": "en",
  "baseline_date": "2026-09-23",
  "baseline_timezone": "Asia/Tokyo",
  "source_revision": "1.0",
  "site_build": "1.0.0",
  "documents": [
    {
      "part": 0,
      "title": "Local AI Infrastructure Lab",
      "description": "A dated field record of a three-node Strix Halo local AI cluster: architecture, memory, context, persistent services, and future workload studies.",
      "html_url": "https://deshimarusakaguchi.com/lab/local-ai-cluster/",
      "markdown_url": "https://deshimarusakaguchi.com/lab/local-ai-cluster/index.md",
      "markdown_sha256": "f3f72ebd19f55046941ceaedb9e8c40e779ecda9260fd33a759e7d2d7ff829ec"
    },
    {
      "part": 1,
      "title": "Building a Fully Local 120B / 72B / 31B AI Cluster with Three 128GB Strix Halo Mini PCs",
      "description": "Architecture, model roles, software versions, and memory observations from three 128GB Strix Halo nodes hosting ten local AI services.",
      "html_url": "https://deshimarusakaguchi.com/lab/local-ai-cluster/architecture/",
      "markdown_url": "https://deshimarusakaguchi.com/lab/local-ai-cluster/architecture/index.md",
      "markdown_sha256": "877046577839e268c7e9f0da987b191543359c773ce9996e304010f88a6ee7c9"
    },
    {
      "part": 2,
      "title": "Memory, Vulkan, and AMD Strix Halo",
      "description": "A 96GB UMA field record: Vulkan/RADV deployment, TTM/GTT differences, memory snapshots, historical ROCm failures, and micro-batch observations.",
      "html_url": "https://deshimarusakaguchi.com/lab/local-ai-cluster/memory-vulkan-strix-halo/",
      "markdown_url": "https://deshimarusakaguchi.com/lab/local-ai-cluster/memory-vulkan-strix-halo/index.md",
      "markdown_sha256": "8d4b41d5a828512df456b577c9a1a5a5ee820e3a894a07ba7e9ea79a2434e7cf"
    },
    {
      "part": 3,
      "title": "Model Context vs. Runtime Context",
      "description": "GGUF declarations, runtime-recognized limits, requested context, effective slots, and captured request rejections in a local llama.cpp cluster.",
      "html_url": "https://deshimarusakaguchi.com/lab/local-ai-cluster/context-runtime/",
      "markdown_url": "https://deshimarusakaguchi.com/lab/local-ai-cluster/context-runtime/index.md",
      "markdown_sha256": "1aa3eabaf5fc6c1db11e7bf712b15593e3cf2baa1d0774d0bd7ecb688938829c"
    },
    {
      "part": 4,
      "title": "Persistent AI Services and Operations",
      "description": "Ten local model endpoints: the actual systemd supervision boundary, API contracts, Dify registration, remote management, and operational limits.",
      "html_url": "https://deshimarusakaguchi.com/lab/local-ai-cluster/persistent-services/",
      "markdown_url": "https://deshimarusakaguchi.com/lab/local-ai-cluster/persistent-services/index.md",
      "markdown_sha256": "070faf42a52b72c38893098a5645291593afb6670ccdbcb6297e25e00ffbbbdf"
    }
  ],
  "source_policy": "The approved Markdown is retained byte-for-byte. HTML is a presentation of the same content. Private raw logs are not included.",
  "scope": "Phase 1 documents a dated construction and deployment baseline; Phase 2 workload measurements are not included."
}