* feat(agents): use the shell's own dialect; allow 5 s retry backoff
Assisted-by: DeepSeek Harness (deepseek-flash)
* feat(dsh): make Nushell the only shell
Assisted-by: DeepSeek Harness (deepseek-flash)
* docs(agents): wait for CI in the background
Pushes to a PR often start multi-minute CI runs, and an agent that holds
the turn open on a foreground watch reports nothing until it ends. Require
the pushed head and running checks to be reported first, the watch to run
in the background, and the result to be collected after independent work
or after ending the turn.
Assisted-by: DeepSeek Harness (deepseek-flash)
* docs(agents): keep the shell-handoff rule platform-neutral
The global rules carry no Windows assumptions: drop the cmd, pwsh, and
wsl handoff examples and keep the POSIX forms plus the dsh refusal.
Assisted-by: DeepSeek Harness (deepseek-flash)
* docs(dsh): keep the shell-swap docs POSIX-only
This repository manages Linux and Darwin hosts, not Windows: drop the
Windows handoff examples and terminal-backend wording from dsh.md. The home
patch also disabled pwsh-sandbox redundantly because dsh-base already
disables that row on POSIX; keep only the active bash-sandbox disable and
verify the composed rows with --dump-config.
Assisted-by: DeepSeek Harness (deepseek-flash)
* docs(agents): keep the global rules runtime-neutral
Global rules are loaded by every agent CLI, so they must not describe one
runtime's behavior. Drop the dsh-specific sentence and wording while keeping
the expectation to prefer Nushell structured pipelines and Python. Runtime
behavior stays documented in home/base/tui/agents/dsh.md.
Assisted-by: DeepSeek Harness (deepseek-flash)
* docs(agents): default shell work to Nushell
Keep the Nushell preference explicit in the runtime-neutral global rules, and
keep runtime behavior in runtime docs. Bash stays a narrow fallback for tasks
Nushell cannot do.
Assisted-by: DeepSeek Harness (deepseek-flash)
* docs(agents): tighten global-rule scope and CI guidance
Keep global rules runtime- and project-neutral, and state that boundary in the
file. Remove the duplicated CI turn-steering text in favour of the Communication
rule, resolve the watch run before watching instead of relying on
gh pr checks --watch alone, name the working-directory parameter generically,
and distinguish a bounded retry backoff from a fixed completion wait.
Assisted-by: DeepSeek Harness (deepseek-flash)
* docs(agents): never estimate a CI run's duration
A CI run has no predictable length, so do not turn "it looks short" into a
reason to wait. The rule is unconditional: report the head and started checks,
start a background watch, and end the turn; collect the result on completion.
The eval scenario now uses an unknown duration.
Assisted-by: DeepSeek Harness (deepseek-flash)
* docs(agents): move maintenance guidance to the README
The runtime/project boundary is guidance for editing the rules, not a rule
agents need loaded every session. The README's Maintaining global rules
section already owns that audience; the loaded rules keep only behavior.
Assisted-by: DeepSeek Harness (deepseek-flash)
* refactor(dsh): render the home patch from a YAML template
Move the home-layer cordis patch into dsh-cordis.patch.yml, substitute the
bundle and Nushell store paths with replaceVars, and link the rendered file
through home.file.source. This module is the only writer of
~/.dsh/cordis.patch.yml, the current host has no such file, and the profile
patch is a separate layer.
Assisted-by: DeepSeek Harness (deepseek-flash)
* refactor(dsh): install the shell plugin through NUR withPlugins
Replace the local runCommand bundle, peer symlink, and home-layer YAML file
with the plugin package and dsh.withPlugins from ryan4yin/nur-packages#43. The
package wrapper adds a generated --patch overlay after the profile layer, so
dsh no longer writes or manages ~/.dsh/cordis.patch.yml. The NUR input is
temporarily pinned to the PR branch and its commit; point it back at main once
the NUR PR merges.
Assisted-by: DeepSeek Harness (deepseek-flash)
* refactor(nix): expose the NUR package set as nurPkgs
Pass one per-system NUR package set through specialArgs instead of repeating
nur-ryan4yin.packages.${pkgs.stdenv.hostPlatform.system} at each call site.
The agents, rime, and idols-ruby modules now use the short binding.
Assisted-by: DeepSeek Harness (deepseek-flash)
* refactor(nix): use nurPkgs in package overlays
The overlays receive specialArgs too, so they can use the per-system NUR
package set instead of rebuilding the long input path from prev. This leaves
only the definition in outputs/default.nix.
Assisted-by: DeepSeek Harness (deepseek-flash)
* refactor(nix): name the per-system NUR set myNurPkgs
Leave room for consuming another NUR repository later, and match the existing
mylib/myvars naming.
Assisted-by: DeepSeek Harness (deepseek-flash)
* chore(dsh): consume the merged NUR package set
nur-ryan4yin now points at main after ryan4yin/nur-packages#43 was squash
merged; the lock records the merge commit. No branch ref remains.
Assisted-by: DeepSeek Harness (deepseek-flash)
* fix(dsh): pass the web profile explicitly
The withPlugins launcher prepends --patch, which historically prevented dsh
from auto-expanding "dsh web" into "--profile web". Pass the profile
explicitly in the Just recipe and the verification commands so they work with
both plain dsh and the wrapper.
Assisted-by: DeepSeek Harness (deepseek-flash)
* chore(dsh): pick up the wrapped-profile fix
Update nur-ryan4yin to the squash merge that keeps an implicit profile name
first, so "dsh web" and "just dsh-web" work through the wrapper again.
Assisted-by: DeepSeek Harness (deepseek-flash)
* chore(dsh): pick up the plugin peer-link fix
Update nur-ryan4yin to the squash merge that links the dsh peer tree at
node_modules/@deepseek-ai instead of nesting it one level deeper. This is the
import failure that stopped nushell-executor from activating after deploy.
Assisted-by: DeepSeek Harness (deepseek-flash)
* feat(aquamarine): make the iGPU usable for media and OpenVINO
The R9700s stay reserved for inference: VA-API/QSV transcoding moves to the iGPU media
engine, and `openvino-gpu` works around ROCm's OpenCL loader shadowing the plugin's
`ocl-icd` (WA-038).
Assisted-by: DeepSeek Harness 0.2.1-rc.2 (deepseek-v4.1-flash)
* docs(aquamarine): document OpenVINO on the iGPU
How to run models on the iGPU: the `openvino-gpu` wrapper, the encoder staticization Whisper needs
on `GPU`, compile/steady timings and memory, and the language and cache caveats.
Assisted-by: DeepSeek Harness 0.2.1-rc.2 (deepseek-v4.1-flash)
The goal is speech recognition on the NPU for vinput, and sherpa-onnx cannot
deliver it: every model it loads keeps a dynamic time axis, which the NPU
compiler rejects outright. Qwen3-ASR and SenseVoice both core-dumped in
vpux-compiler ("Got non broadcastable dimensions pair"), and reshaping the ONNX
inputs does not help because the dynamic dimensions sit inside the graph
(WA-037). Whisper is the only family OpenVINO publishes fully static IRs for,
and its fixed 30 s window is exactly what the compiler accepts.
The host now runs an OpenAI-compatible /v1/audio/transcriptions service backed
by whisper-large-v3-turbo int8 on the NPU: 1.5 s for 11 s of audio against
2.1 s on CPU, with the compiled graph cached in /var/cache so only a cold cache
pays the ~60 s NPU compile. vinput on the desktop sends raw PCM to it through a
small provider script.
Only the NPU driver and compiler still come from the unmerged branch in the
`nixpkgs-npu` input (NixOS/nixpkgs#570230), whose nixpkgs also happens to be
the one where openvino-genai 2026.4.0.0 builds at all.
Assisted-by: DeepSeek Harness 0.2.1-rc.2 (deepseek-v4.1-flash)
On a tmpfs root the flake/fetch cache is empty at every boot, so
`nix run <flake-ref>` re-downloads its input and fails until DNS is up: a reboot
test showed llama-swap's first start failing on that. Other caches write into
RAM meanwhile.
Assisted-by: DeepSeek Harness (deepseek-flash)
The second NVMe and its LUKS volume leave with the Z890 board, so the data-disk disko module is deleted; `ai` must activate this change before the disk is pulled, or its initrd blocks on the missing volume.
Assisted-by: DeepSeek Harness (deepseek-flash)
The Z890 board and its root SSD move here from `ai`, and the retired VM's name and 192.168.5.101 are reused.
Assisted-by: DeepSeek Harness (deepseek-flash)
The worker MicroVMs ran `home-manager-ryan.service`, which failed on two of them
(`/nix/var/nix/temproots` permission denied, and no `/home/ryan`), so
`HostSystemdServiceCrashed` fired for days. Mirror the masters and extend their
regression test to all six nodes.
Assisted-by: DeepSeek Harness 0.2.0-rc.2 (deepseek-flash)
* flake.lock: Update
nixpkgs only. The new revision builds VictoriaMetrics with cgo
(CGO_ENABLED=1); its pure-Go zstd decoder rejected a stored values block, so
every small-part merge panicked on youko.
* fix(colmena): address the VM hosts by name, not by their IP
`~/.ssh/config` pins the passphrase-protected builder key with IdentitiesOnly
for 192.168.*, so IP-addressed ssh cannot authenticate without a terminal. The
name resolves through the aliases in /etc/ssh/ssh_config.
* fix(monitoring): repair two alert rules VictoriaMetrics rejects
`on(uid)` matches on a label kube-state-metrics no longer emits, and
`!= ON(__name__, instance)` treats every pg_settings series as a duplicate.
They now join on (namespace, pod) and compare with `changes(...[5m])`.
* fix(monitoring): keep disk latency alerts off mechanical disks
Both rules averaged per-op latency with no device or sample guard, so the 4 TB
USB HDD volumes paged constantly (>100 ms/write in 26% of the last 30 days).
They now require a non-rotational physical device and >=5 ops/s.
Assisted-by: DeepSeek Harness 0.2.0-rc.2 (deepseek-flash)
No host defines a *-nixos-tests output; the pod firewall drops 5900-5903
from pod sources; the mihomo core keeps geodata and cache in its
StateDirectory; WA-020 was used twice and the cited WA-028 had no row.
`just test` has enforced the suite result since #350: the recipe raises unless
`evalTests` is `true`, so a zero exit already means pass. Update the places that
still described the old raw-`nix eval` behaviour: AGENTS.md, README.md,
SECURITY.md, outputs/README.md, and the nix-config-debug / nix-config-new-host /
nix-config-update skills.
The exit-0 caveat is kept only where the raw `nix eval .#evalTests` command is
documented, because that command genuinely exits 0 even on `false`.
Every remaining pkgs-x64 consumer runs only on x86_64-linux hosts, where the
native pkgs already carries the same allowUnfree config
(modules/nixos/base/nix.nix) and overlays (modules/base/overlays.nix), so
replace the references with pkgs and drop the extra nixpkgs instantiation
(which also removes an eval on aarch64-darwin).
media.nix keeps the native Zoom app on x86_64 hosts only, since zoom-us has
no aarch64 Linux build.
* fix(postgres): require a password for superuser over TCP
* test: assert the postgres loopback auth rules
* fix(darwin): keep root-only agenix secrets root-only
* docs(security): record the local auth boundaries and retire WA-015
* fix(k3s): deny credential-bearing host ports to pod sources
* test: assert the k3s pod firewall drops sensitive ports
* docs(security): record the pod firewall boundary
dsh rewrites a profile's cordis.patch.yml at runtime, so linking the web
profile put GUI state in the checkout and let runtime writes clobber manual
edits. Link the stable policy as the $DSH_HOME home layer instead: dsh applies
it after every profile and only reads it.
The host-plane compaction-basic row is disabled on the web surface and each
agent preset owns its own copy, so the top-level override was inert. Drop it;
the model contextWindow/maxTokens are the effective knobs.
- add home/base/tui/agents/dsh/cordis.patch.yml, linked from default.nix
- drop the ~/.dsh/profiles/web link and the tracked web/ profile
- keep runtime state and private/local model routes in the profile patch
- update the eval test and README
Vinput's streaming Zipformer models have unbounded dynamic shapes, so
OpenVINO cannot compile them for the GPU or the NPU; the `openvino`
provider and ai's Intel NPU enablement could never run on the NPU (noted
as unverified in WORKAROUNDS WA-017). Remove the backend option, the
daemon's OpenVINO library path, and the intel-npu module enable, and
update the eval test, README and WORKAROUNDS (drop WA-016 and WA-017).
Keep Vinput's stock package and its bundled sherpa-onnx (1.13.4): the
fork-based and nixpkgs-master sherpa overrides existed only to provide the
`openvino` provider, so drop package.nix with them.
Assisted-by: opencode (DeepSeek V4.1 Flash)
* fix(desktop): tie ai's link DNS takeover to mihomo's lifecycle
sing-tun normally points systemd-resolved at the TUN DNS by running
resolvectl (dns/domain/default-route). Under nixpkgs' hardened
DynamicUser mihomo unit that hand-off fails silently: the unit's
RestrictAddressFamilies blocks resolvectl's D-Bus socket and resolve1
requires privileged authorization, so resolved keeps querying the
physical-link DNS, fake-ip is bypassed, and domain rules degrade to
MATCH.
Do the hand-off as '+' ExecStartPost/ExecStopPost hooks instead: '+'
commands run outside the sandbox (systemd.exec: address-family
restrictions do not apply to them) and ExecStopPost reverts even on
crashes, so a dead mihomo leaves DNS usable.
Signed-off-by: Ryan Yin <xiaoyin_c@qq.com>
* docs(mihomo): record sing-tun's resolvectl hand-off as the leak cause
Signed-off-by: Ryan Yin <xiaoyin_c@qq.com>
* refactor(desktop): keep mihomo only on ai and shoukei
ruby/kana never ran the core (service inactive live), and their
resolved + TUN combo would reintroduce the resolver leak once enabled.
Drop the enable and the explicit module import from computer-use; the
eval test now pins mihomo=false for both guests.
---------
Signed-off-by: Ryan Yin <xiaoyin_c@qq.com>
* refactor: consume computer-use packages from NUR
* fix: expose flake inputs to package overlays
Package overlays are imported from two places with different argument
shapes: outputs/default.nix passes a set containing `inputs`, while
modules/base/overlays.nix passes the NixOS module args, where genSpecialArgs
spreads each flake input as a top-level argument. The new NUR-backed
overlays read `inputs.nur-ryan4yin`, so evaluation failed on the module
path. Also expose the whole flake input set as `inputs` in genSpecialArgs.
* chore: bump nur-ryan4yin input
Update the locked nur-packages revision to the current main (nixpkgs
26.05, unused packages removed).
Add SECURITY.md covering the threat model, trust boundaries,
proportionate kernel/process hardening, the update/reboot routine and
remaining prioritized work. Record the dated read-only audit findings
(listeners, privileged workloads, effective RBAC, secret metadata and
sandbox/SSH boundaries) separately from planned controls and runtime
verification.
Define agent-first diagnosis: scoped root read-only access on homelab
hosts, and normal-user diagnostics on core desktops with no new sudo
grants. Add canonical domain-based k3s API endpoint regression tests,
record the unverified Darwin secret-ownership concern as WA-015, and
correct comments claiming mode 0000 blocks privileged root.
* fix(colmena): expose colmenaHive for the new evaluator
nixpkgs now ships the nix-community Colmena rewrite, whose default
evaluator reads an evaluated colmenaHive flake output instead of the raw
colmena hive. Without it every colmena apply recipe failed with
"does not provide attribute colmenaHive".
Add a colmena input pinned to the same v0.5.0 as nixpkgs and expose
colmenaHive = colmena.lib.makeHive <raw hive>, with the raw hive kept as
an internal binding rather than a flake output of its own.
* chore: clean up stale colmena and aquamarine references
- point the Colmena links at nix-community/colmena after the move
- rename the security-aquamarine-metrics test to youko-metrics (it
already tested youko) and update the outputs README
- drop the commented aquamarine remote-builder example
nixpkgs-blender tracked the nixpkgs master branch, so every `just up` re-bumped it and started a full Blender/CUDA source rebuild even though Blender stayed at 5.2.2 LTS. Pin it to a fixed rev so bumps are deliberate.
The OIDN cmake override is obsolete: nixpkgs#545542 made setupCudaHook add nvcc to CUDAToolkit_ROOT, fixing the CMake 4.3+ split-root failure (nixpkgs#544701). Dropping it also stops Blender's derivation from depending on nixpkgs-stable.
Hosts used to trust the router as the only boundary (servers ran with the
firewall off; service ports were opened globally), so any LAN device could
reach every host service. Move filtering to the hosts as defence in depth
behind the router, which keeps its firewall on and uses a port mapping for
the few ports we actually need.
- base: nftables backend; deny inbound by default; trust the home LAN by
source and the tailnet by interface (restrict the tailnet with ACLs);
suppress per-service `openFirewall` global ports; allow node_exporter only
from the monitoring host. The local podman bridge, the tailnet and the pod
CIDRs are accepted only after the node_exporter drop.
- servers: stop disabling the firewall; drop redundant global port opens
(caddy 80/443, NFS 2049, Jellyfin, localsend/test ports); delete the
redundant desktop firewall module; aarch64/riscv64 import the firewall
explicitly (akane was missing it).
- k3s: Cilium pods (10.0.0.0/8, fd00::/8) may reach host services; drop the
flannel-era cni0/flannel.1 trust.
- youko: postgres listens on localhost + the LAN with pg_hba limited to the
LAN; the NFS export stays LAN-scoped; remove the unauthenticated v2ray
proxy.
- tests: `security-firewall` asserts every host runs the shared firewall, has
the LAN CIDR rule and no global ports.
Deploy RustFS in place of the retired MinIO and move the S3 consumers over.
- flake.nix / flake.lock / outputs/default.nix: add a nixpkgs-small
(nixos-unstable-small) input for the rustfs 1.0.0 package (with the web
console), instantiated as pkgs-small in genSpecialArgs alongside the other
nixpkgs instances; drop the rustfs flake input. The rest of the system stays
on the default nixpkgs.
- hosts/12kingdoms-youko/homelab-services/rustfs.nix: use nixpkgs' own
services.rustfs module, with package = pkgs-small.rustfs and
environmentFile = the agenix rustfs.env secret.
- caddy.nix: s3.writefor.fun (S3 API) and s3-console.writefor.fun (console,
redirects / to /rustfs/console/).
- secrets/nixos.nix: declare a single rustfs.env (dotenv) secret holding
RUSTFS_ACCESS_KEY / RUSTFS_SECRET_KEY, replacing the separate key files.
- infra/minio -> infra/rustfs: buckets are managed with the hashicorp/aws
provider against the RustFS S3 endpoint; IAM/ILM are applied with rc.
Keep the six node configuration intact while removing the failing user activation from the three control-plane MicroVMs. Update the regression test and current K3s configuration notes.
0.156.1 still lacks the codex-chatgpt recursion_limit attribute, so keep the postPatch that adds it. The vendored Cargo.lock is unchanged, so the cargoHash stays the same.