nixos-config/machines/kusanagi/default.nix, branch main

kusanagi: constrain ollama resources

2026-08-04T23:36:13Z

laguna-xs-2.1 is large enough to nearly fill the workstation's display GPU at the previous 64k context and 1h keep-alive. That let Ollama keep a very large resident set around after requests and could destabilize the no-swap desktop session. Unload models quicker, limit parallelism/queueing, reserve VRAM for the compositor, and cap the service cgroup so runaway requests fail before the desktop does. Assisted-by: pi:gpt-5.5

networking: expose Ollama and mDNS on Wi-Fi LAN

2026-08-04T22:14:09Z

Serve Ollama on kusanagi's Wi-Fi LAN while removing the blanket tailscale0 firewall trust so tailnet traffic follows explicit host firewall rules instead of inheriting every listening service. Add shared LAN networking for both hosts: allow ping, publish and resolve .local names with Avahi, and scope mDNS to each machine's declared Wi-Fi interface. Assisted-by: OpenAI:gpt-5

kusanagi/ollama: double default context, add 1h keep-alive

2026-08-04T21:05:20Z

The 12B model at 32k context uses only 7.7 GB of the 7900 XT's 20 GB, so there is ample VRAM headroom for 64k with the q8_0 KV cache — agentic coding tools are context-hungry and 32k is the practical bottleneck. The keep-alive stops the default 5-minute idle unload from adding a reload stall to every resumed session on a single-user workstation. Update flake to get newer packages. Assisted-by: Claude Code:claude-fable-5

kusanagi/ollama: fix boot race with amdgpu, drop misdiagnosed HSA override

2026-08-04T03:51:26Z

The recurring "ollama runs 100% on CPU" regressions were a boot-order race, not a gfx-target mismatch: ollama.service only ordered after network.target, and at boot it started between /dev/kfd and the 7900 XT's render node appearing. ollama probes GPUs once at startup, found none, and silently fell back to CPU until the next manual restart — which is why the previous HSA_OVERRIDE_GFX_VERSION commit "worked": the deploy restarted the service on a long-running system, and the restart was the actual cure. The card is a 7900 XT (Navi 31), natively gfx1100 — verified that discovery with no overrides finds it fine and drops only the unsupported Raphael iGPU (gfx1036). Remove HSA_OVERRIDE_GFX_VERSION and HIP_VISIBLE_DEVICES (ollama itself warns overriding visible devices can break discovery). Real fix: systemd-tag the kfd/renderD* char devices via udev (they get no device units otherwise) and order ollama after them, so discovery only runs once the GPU exists. Assisted-by: claude-code:claude-fable-5

profiles: extract shared sandbox profile for enzo + kusanagi

2026-08-04T03:12:32Z

Move the yokai agent-sandbox account (user, ikai group, /srv/git drop dir, scoped sudo de-escalation, and the local-clones workflow docs) out of machines/enzo into a shared profiles/sandbox.nix that both machines import. The only per-machine value is the human owner allowed to drop into the sandbox, exposed as a `sandbox.owner` string option so the profile stays username-agnostic (enzo: hwebs, kusanagi: henz). kusanagi gains the sandbox it didn't have before; enzo's inline block collapses to one setting. Assisted-by: claude-code:claude-opus-4-8

kusanagi/ollama: fix ROCm GPU fallback to CPU for Gemma

2026-08-04T02:39:51Z

The RDNA3 card reports as gfx1101/gfx1102/gfx11-generic, which the ROCm build bundled with ollama-rocm has no kernels for, so ollama found no usable device and ran 100% on CPU after the nixpkgs bump. Force it to present as gfx1100 via HSA_OVERRIDE_GFX_VERSION. Also set OLLAMA_FLASH_ATTENTION=1 so the existing q8_0 KV cache actually engages instead of silently falling back to f16 and inflating VRAM use. Assisted-by: Claude-code:Opus-4.8

migration away from x-clip

2026-08-03T18:01:21Z

- replace x-clip with wl-clipboard for wayland machines - add TODOs I noticed while editing Assisted-by: Claude Code:claude-opus-4-8

kusanagi: remove non-working hiberation

2026-08-01T15:18:25Z

Assisted-by: Claude:Opus-4.8

enzo/kusanagi: idle suspend via swayidle

2026-07-31T03:47:58Z

Bare niri has no power daemon watching for inactivity, so neither box slept on its own. Add a user swayidle service bound to graphical-session.target that fires after 15 min of no input: - enzo: suspend-then-hibernate (matches the existing lid-switch policy) - kusanagi: plain suspend-to-RAM, plus Wake-on-LAN on enp9s0 so it can be woken remotely while the Kavita/Immich/Ollama/NFS servers sleep Assisted-by: Claude:claude-opus-4-8

refactor: shared desktop profile for enzo + kusanagi

2026-07-31T00:14:09Z

profiles/desktop.nix becomes a username-agnostic shared base (niri, dconf, the dark-theme portal fix, greetd, pipewire base, gpg agent, graphics, Firefox with blank-home/no-bookmarks prefs, common desktop tools, font base). Both hosts now import it. Per-host desktop config splits into machines//desktop.nix: - enzo: Adwaita cursor + monospace, HiDPI console font, yambar, hwebs groups. - kusanagi (new file): GNOME/X11, printing/avahi/udisks2, pro-audio (rtkit/jack/32-bit), gpg ssh support, richer fonts, Bibata cursor, henz desktop groups, waybar/nm-applet. Shared profile is username-agnostic (enzo=hwebs, kusanagi=henz): shared desktop tools moved to environment.systemPackages, per-user groups stay in machine files. Firefox now owned by the shared profile, so its redundant enable is dropped from profiles/apps.nix. Dark theme + Firefox tweaks now apply to both hosts. Both configurations build. Assisted-by: Claude:claude-opus-4-8