chore(wiki): weekly services table refresh
Update AI stack diagram for MemPalace
Update Home.md — enhance descriptions and clarify software references
chore(wiki): update Home.md to reflect recent changes
Update AGENTS.md — Troubleshooting_guides removed, points to Notion KB Drops the deleted directory from Directory Intent, the now-dead Troubleshooting Page Pattern section, and dangling links into the removed guides. Adds a pointer so future agents write new fix guides to the Notion KB instead of recreating this directory.
chore(wiki): remove obsolete Dotfile-management and KDE-shortcuts documentation
Remove Troubleshooting_guides — fully migrated to Notion KB All 20 guides confirmed present in the Notion AI Knowledgebase (bulk migration 2026-07-02, plus Fix-NVMe-APST-Latency-Spikes.md added 2026-08-26 as it postdated the bulk migration). Wiki is no longer the source of truth for this content.
chore(wiki): drop leftover LMSHUB mermaid class on Home.md
chore(wiki): drop leftover llmster/Piper-on-ai-hub claims Home.md LM Studio row and ai-stack.mmd still advertised local llama.cpp/llmster/Piper on ai-hub. Live units are masked.
chore(wiki): drop ai-hub llama.cpp/Piper from stack mermaid Local llama.cpp/llmster and Piper are masked on ai-hub. Keep NAS llama.cpp only.
docs: drop outdated Synology troubleshooting guides Remove shell-change and VSCode-on-Synology pages that no longer match the lab.
docs: YubiKey setup covers polkit GUI dialogs and linutil module Document Arch-style /etc override for polkit and point at Security → YubiKey PAM.
docs: AI-stack diagram matches oliver's rebuilt hierarchy sketch Channels: Conduit, Hermes desktop app, Signal, Official Android apps, VS Code via Tailscale. Frontend&agents: explicit OWU<->Hermes, Hermes<->CLI bidirectional, CLI-can-control-OWU. Added NAS (Intel N100) as third self-hosted llama.cpp provider. pgvector marked planned, not live. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015MPLeBjzHFWAvTYhAQ84xb
docs: remove redundant standalone chatbot-apps box from AI-stack diagram Its content was fully absorbed by the "Official standard chatbot apps" Channels pill added in the previous commit — a disconnected box duplicating that name added no information. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015MPLeBjzHFWAvTYhAQ84xb
docs: shrink AI-stack Channels box — consolidate 7 pills into 4 Merged Browser+Conduit into one "Open WebUI access" pill, and Signal+Tailscale SSH+Hermes dashboard into one "Hermes access" pill (labels preserved on the edge). Renamed "Claude · Android" to "Official standard chatbot apps" (it covers launching any first-party CLI-agent app, not just Claude). Removed Lumo from the reference apps box. Layout unaffected — same band structure. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015MPLeBjzHFWAvTYhAQ84xb
docs: AI-stack layout — middle band (MCP/providers/memory/OWU services) side by side Dropped the inter-box vertical chain; each band box now ranks directly below the app layer so dagre places them in one horizontal row, sandbox below via invisible links. Added the APP -> shared-memory dashed edge back. SVG regenerated. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015MPLeBjzHFWAvTYhAQ84xb
docs: consolidate MCP layer into two gateway nodes; regenerate ai-stack.svg Individual MCP-server nodes made the diagram 6000px tall. Server lists moved back into the two gateway labels (MCPJungle self-hosted / mcpo remote), stacked with an invisible link. ai-stack.svg regenerated from source via headless Chrome + mermaid 11 (XMLSerializer, pure-SVG labels — valid XML). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015MPLeBjzHFWAvTYhAQ84xb
docs: fix AI-stack vertical stacking — invisible rank links instead of arrows Grok's directed chains (baikal --> dbhub --> ...) drew false data-flow arrows inside the MCP/provider/memory boxes. Replaced with ~~~ invisible links: same vertical stacking, no fake semantics. Added missing chains for providers and OWU services. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015MPLeBjzHFWAvTYhAQ84xb
docs: sync AI stack diagram in Home.md with ai-stack.mmd (MIDDLE wrapper + connections for layout; directions already correct per goal)
docs: sync ai-stack.mmd (AI stack diagram source with required direction LR for channels+frontend/agents, TB for mcp+providers+memory per spec)
docs: fix AI-stack diagram layout - vertical stacking inside boxes, narrower columns, official apps section, centered HDASH, code-server to APP box
docs: rework AI-stack diagram (shared memory/knowledge, RPi4 sandbox, channels) + prose
docs: rework network topology diagram Two-layer security model (Cloudflare L1 / Zoraxy L2), Entry Points grouping (Cloudflare + Tailscale), filtered Mullvad/Cloudflare DNS, recruiter-facing service lists, RPi4 as Folding@home + agent sandbox, all client devices in a bottom row wired to the router. Prose rewritten to match. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015MPLeBjzHFWAvTYhAQ84xb
docs: update service list in Home.md for clarity and consistency
docs: create AGENTS.md to outline guidelines for AI coding agents
docs: add troubleshooting guide for NVMe APST latency spikes
docs: sync Home.md with live infra — rebuilt services table (58 rows), regrouped AI-stack diagram (channels/agents/MCP/providers layers), zone-style network diagram, no private IPs, ollama removed, Conduit+Signal channels documented
docs(llm): add LM Studio headless setup guide, add hosted-services row Fixed broken aur/lmstudio-beta lms CLI, wired as a system-level systemd service (avoiding the known linger/gamescope conflict), and shared ~/.lmstudio across both os93-pc/os93-steam-pc CachyOS instances via symlink to shared SSD storage.