- promote.sh: fail fast before mutating the public tree if the tag was already
promoted (avoids a half-commit left on the runner on re-dispatch).
- ci-staging.yml: pass STAGING_PROJECT_DIR via env + :? guard instead of a
${{ }} expression in run: (clears the template-injection lint, fails loudly
on misconfig).
- README: PUBLIC_REPO_URL was never read by any workflow — keep only the slug.
- set-repository-url.test.sh: chmod +x for mode parity with promote.test.sh.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Promotion workflows for the release pipeline. promote-public publishes a tag to
the public repo's beta channel; promote-stable re-tags a finalised beta as a
stable vX.Y.Z. Both bound to the 'production' environment (required reviewer),
reading PUBLIC_REPO_SLUG/PUBLIC_REPO_TOKEN. Tree copy via scripts/promote.sh
excludes .git/.github so private CI never reaches the public repo.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Stable re-tags the same source commit as the last beta, so the extracted tree
is byte-identical and `git commit` would abort on "nothing to commit" under
set -euo pipefail, leaving the stable tag uncreated. Use --allow-empty and add
a promote.test.sh case that promotes a second identical-tree tag into the same
pub repo to guard the regression.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Restore the plan's if-grep / replace / else-append skeleton in the
CLUSEV_REPOSITORY write-back. The replace branch deletes existing
line(s) then appends once so duplicates collapse to a single entry,
keeping scripts/set-repository-url.test.sh green (the plan's verbatim
in-place sed leaves two identical lines and fails the duplicate-line
assertion).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Capture the maintainer's release workflow: dev-dashboard buttons drive the
pipeline (Deploy to Staging / Deploy to Public / Promote to Stable). Beta =
vX.Y.Z-betaN on staging, Stable = vX.Y.Z on public; public serves both channels.
Public promotion triggered via the GitHub API with a deploy token in the
encrypted vault. Decomposed: Phase A = git/CI/CD + config-URL infra (buildable
now, URL-agnostic); Phase B = the dashboard release-control UI (own spec, after
A + the repos exist).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Per the maintainer: dev/staging may hold any URL (in their gitignored .env);
the public repo's tracked files carry only the public URL. So config/clusev.php
defaults to the public URL, private (Gitea/GitHub-private) URLs live solely in
gitignored .env, and install.sh derives CLUSEV_REPOSITORY from the clone origin.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Design for promoting Clusev dev → staging → public across three repos with
GitHub Actions CI, a manual gate before public, and a clean public history.
Core: no repository URL in tracked files — derived from the clone origin, kept
in gitignored .env per environment, so the public repo never exposes the
private/Gitea URLs and the URLs plug in later when the GitHub repos exist.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- A friendly HTML comment in the page source (both layouts, before <head>) —
ASCII "CLUSEV", what it is, and the project link — for anyone who View-Sources.
- A styled console banner printed on load (every page, incl. login): the CLUSEV
wordmark, a one-liner, the repo link, AND a self-XSS security warning. The
warning earns its place: a Linux-server control panel is a prime target for
"paste this into your console" social engineering, so the notice sits exactly
where that attack happens. console styling is best-effort and never throws.
Verified: /login 200, the console logs the brand + warning + link, the source
carries the comment, no page errors; 374 tests pass; Vite build clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- File manager: on tablet widths the lg table grid squeezed the name column
(minmax(0,1fr)) to near-zero behind the perms/size/modified columns + wide
action buttons, so the filename was unreadable/overlapped. The rows now render
as a stacked card below xl — name on its own full-width line, a folded
perms·size·modified line and the actions beneath — and only switch to the
table grid at xl (tightened columns). Mobile tidied up to match.
- Restore the pointer cursor: Tailwind v4's preflight makes buttons use the
default cursor, so clickable controls (folder names, every button) didn't show
a hand. Added a base rule giving enabled button/[role=button]/summary/label
the pointer cursor.
Verified the row at desktop / tablet / mobile widths (name readable at every
breakpoint, table only when wide). 374 tests pass (the Livewire test renders the
blade), Tailwind build clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Opening a binary / non-UTF-8 file in the file manager threw "undefined is not
valid JSON". The editor bound the raw file bytes to its public $content
property; Livewire JSON-encodes the component snapshot, and invalid UTF-8 makes
PHP's json_encode return false, so the client received an empty response and
JSON.parse("undefined") failed (the trace in mergeNewSnapshot/deepClone).
Fixed at both layers: FleetService::readFile blanks the content when it detects
binary/non-UTF-8 (returns the binary flag instead), and FileEditor::load only
binds content for non-binary, non-oversize files. Binary/oversize files show
their existing notice; the snapshot is always valid UTF-8.
Tests: load blanks binary content (keeps the snapshot valid) and keeps valid
text content. 374 pass, Pint clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Lists rendered loop rows without a wire:key. On re-render Livewire's morph
matched rows by position and cloned them, dropping the inline Alpine x-data
(modalTrigger) scope on the buttons — the console threw "Alpine Expression
Error: pending is not defined" (and a follow-on "undefined is not valid JSON"
as the snapshot merge broke). Reported in the file manager; the same pattern
existed across the app.
Added a stable wire:key to every loop whose rows carry inline Alpine / x-modal-
trigger state: files (entries + breadcrumbs), services, the release-channel
selector, settings (users, webauthn keys, login-protection bans) and the server
detail page (hardening, firewall rules, SSH keys, fail2ban exceptions). Keys use
a stable identifier (id / fingerprint / md5 of the natural key), not the loop
index.
372 tests pass (the Livewire tests render these blades, so a broken key would
fail them); /system, /settings, /files load 200 with no console errors.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The /audit page already rendered human-readable action labels (v0.9.48), but the
Dashboard's "recent events" panel still showed the raw codes (wg.set-endpoint,
deploy.update_request). It now uses the same AuditEvent::action_label + is_error
styling, so failures show red with an alert icon — consistent with the audit page.
Test: dashboard recent events render the readable label, not the raw code. 372
pass, Pint clean, dashboard loads 200 with no console errors and no raw codes.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Two fixes the operator asked for:
- The "what's new" preview was empty. availableChangelog fetched CHANGELOG.md at
ref="0.9.50", but release tags are "v0.9.50" — the raw fetch 404'd and the
panel silently stayed empty. Use the v-prefixed tag as the ref. The existing
test masked it with a wildcard fake; tightened with Http::assertSent.
- The sidebar now shows a "1" badge on the Version item when an update is
available, so an update is visible on every page, not only the Versions page.
Driven by a new ReleaseChecker service: updateAvailable() is a pure cache read
(no network in the sidebar); a scheduled clusev:check-update (every 30 min,
anonymous/read-only) keeps the cache warm so the badge is accurate even before
the operator opens the Versions page.
Refactor: the remote tag lookup (fetch + newest-version) moved out of the
Versions component into ReleaseChecker; the component delegates. nav-item gains
an optional badge prop.
Tests: ReleaseChecker (cache-read update-available, refresh caches, command warms
cache, sidebar badge renders only when available) + the v-prefixed-ref assertion.
371 pass, Pint clean, /versions loads 200 with no console errors; badge + the
"What's new in v0.9.50" panel verified in the browser. Lang parity kept.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The audit log loaded the latest 50 entries and filtered them in memory, so it
couldn't grow and search only spanned those 50. Now it is paginated (25/page)
with the page deep-linked in the URL via Livewire WithPagination (?page=), and
the search runs in SQL across the whole history — matching actor / action code /
target / server name AND the readable labels (a term is translated to the action
codes whose localized label contains it, so "Endpoint geändert" still finds
wg.set-endpoint). Pagination control mirrors the changelog's windowed pager.
Tests: 25/page with page 2 overflow, search resets to page 1, search filters in
SQL across all pages (a match on page 2 surfaces when filtered). 365 pass, Pint
clean, /audit loads 200 with no console errors, audit lang parity 62/62.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Versions page only rendered the INSTALLED CHANGELOG.md, so the operator
couldn't see what an available update contained until after applying it. Now,
when behind a published release, a "What's new in vX.Y.Z" panel shows the
changelog of the available version(s) — fetched from the remote CHANGELOG.md at
the latest tag (Gitea raw API, anonymous/read-only, briefly cached), parsed with
the shared changelog parser and filtered to versions newer than installed.
Degrades gracefully: a failed fetch yields an empty preview and never blocks the
update button. Refactored releases() to a reusable parseChangelog().
Tests: preview lists only newer versions + renders their entries; no preview
when current. 362 pass, Pint clean, /versions loads 200 with no console errors,
versions lang parity 51/51.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The audit log showed raw machine codes ("Administrator · wg.set-endpoint").
Now each entry renders a localized human-readable label ("WireGuard endpoint
changed"), via an AuditEvent::action_label accessor backed by an audit.actions
lang map (DE/EN); unmapped/dynamic codes (e.g. harden.*.on|off) fall back to a
tidied form, never a bare code. Search also matches the readable label.
Failures are now visible: AuditEvent::is_error flags failed/security events
(failed sign-in, IP ban, failed 2FA, wg.action-failed) which render in red with
an alert icon. And failed WireGuard dashboard actions are now written to the
audit log with the host error message (not just a transient toast), so the
operator can read later what went wrong.
Tests: label mapping + fallback, is_error, page renders label not code, search
by label, WG failure auditing. 360 pass, Pint clean, /audit loads 200 with no
console errors, audit lang parity 59/59.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The server config showed a "Port" and an "Endpoint" that repeated the same
:51820, so it looked like the port had to be entered twice. It never did (since
0.9.43 a bare host gets the listen port appended) — but the UI didn't say so.
Now the field is labelled "Public endpoint (host)", the placeholder shows only
the host (port stripped via Str::before), and the hint explains the port is
taken from the listen port automatically; a different port is only needed for
port-forwarding (host:port). Display/copy only — no behaviour change.
Pint clean, /wireguard server tab loads 200 with no console errors, lang
parity 89/89, 354 tests pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- The Traffic panel header ("Empfangen/Gesendet") drew from the sampled
throughput, which is 0 until the minutely sampler accrues deltas — so a peer
with live traffic (e.g. 244 B) showed "0 B" in the header. It now sums the
live per-peer rx/tx counters, so the header always matches the peer rows. The
chart below stays windowed throughput (fills from the sampler).
- import_hint reworded to state plainly that the WireGuard phone app ALWAYS
prompts for a tunnel name on a QR scan (app behaviour, not changeable); the
.conf download is the path to an automatic name (filename = tunnel name).
Tests: header sums live peer counters. 354 pass, Pint clean, /wireguard 200 with
no console errors, lang parity 89/89.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Three things the operator asked for on the WireGuard page:
- DNS is no longer hardcoded to 1.1.1.1. New host action set-dns (_set_dns +
cmd_set_dns + write-bridge branch + WgBridge validation, IPv4 list only),
stored as WG_DNS in wg.env (appended for pre-existing tunnels), surfaced via
the collector + WgStatus, baked into new peer configs (DNS = ${WG_DNS:-1.1.1.1}).
Editable in the Server tab; e.g. point clients at your own gateway.
- The page is split into two independent tabs: "Peers" (list / add / remove /
traffic) and "Server configuration" (endpoint / DNS / port / subnet / gate /
SSH lock + server identity). #[Url] $tab makes it deep-linkable (?tab=server).
- Offline detection: ONLINE_WITHIN 180s -> 150s (PersistentKeepalive=25 keeps
active peers re-handshaking well within it, so no flapping; a disconnect now
shows offline ~2.5 min sooner). Endpoint hint clarified: the port is optional
(auto-appended), no need to repeat the listen port.
Tests: set-dns (write+audit, reject), tab switch/clamp, dns in status; the two
configured-render tests now switch to the server tab for server content. 353
pass, shellcheck clean, Pint clean, both tabs load 200 with no console errors,
lang parity 89/89.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- QR: 0.9.42 over-enlarged it; bring the show-once peer QR back to a balanced
~240px (still crispEdges on a white quiet-zone frame) so it scans on a phone
without dominating the modal.
- Download filename now becomes a useful tunnel name on import: build it from
the endpoint host + peer name (e.g. "10.10.90.165-laptop.conf") instead of the
generic "clusev-wireguard.conf" — WireGuard apps name a tunnel after the file.
A QR scan can't carry a name (the app prompts for one); the import hint now
explains both paths.
Tests: download-filename cases (host+peer, fallback); 349 pass, Pint clean,
/wireguard loads 200, lang parity 83/83.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
If the operator entered an endpoint without a port (e.g. "10.10.90.165"
instead of "10.10.90.165:51820"), that value was written verbatim into every
peer config. WireGuard requires the endpoint as host:port, so the WireGuard app
rejected the QR/file import as "keine gültige WireGuard-Konfiguration" — the
real cause of the invalid-config errors (the QR-size fix in 0.9.42 was secondary).
Normalise host-side: append the listen port when the endpoint has no colon, in
both _write_server_conf (setup — covers the dashboard and interactive paths) and
_set_endpoint (later changes). Verified: bare host -> host:port, host:port left
as-is, DNS host -> host:port; write-bridge smoke test still passes; shellcheck
clean. setup_endpoint_hint (de/en) now documents the host:port format.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The show-once peer config QR rendered at 176px (~2.7px per module) with
anti-aliasing on. A WireGuard config QR is dense (~57 modules), so a phone
camera misread it and the WireGuard app rejected the result as "keine gültige
WireGuard-Konfiguration". Render it large (up to ~300px) on a white framed
quiet zone with shape-rendering:crispEdges, so modules stay sharp and ~4.6px
each. The Download button (.conf import on a PC) remains as the alternative.
Verified: old-vs-new render comparison (2.7 → 4.6 px/module, crisp modules);
Tailwind compiles the arbitrary shape-rendering property; /wireguard loads 200
with no console errors.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The host handler `clusev-wg.sh serve-request` aborted under `set -euo pipefail`
BEFORE writing run/wg-result-<id>.json, so the dashboard waited ~30s and showed
"Keine Antwort vom Host" even though the systemd watcher was running correctly.
The write-bridge had therefore never actually worked on a real deployment.
Two root causes, both fixed:
1. Field extraction: each `id/action/name/endpoint/port/subnet` value is pulled
with `grep -oE | head | sed`. A field that is empty or absent (an "auto"
endpoint at setup; gate/add-peer/remove-peer payloads that carry no
endpoint/port/subnet) makes grep fail → with pipefail the whole pipeline
fails → `set -e` aborts the handler before _write_result. Appended `|| true`
to every extraction so a non-match yields "" instead of killing the handler.
2. Action dispatch: gate-up/gate-down/gate-ssh-on/off and set-endpoint/port/
subnet (and remove-peer) called helpers that run require_setup, which `die`s
(exit) when the tunnel isn't configured yet — aborting serve-request before
the reply. Each handler now runs in a subshell `( ... )` so an internal exit
becomes a caught non-zero and still yields a clean failure result.
(add-peer/setup were already isolated via their $(...) command substitution.)
Verified end-to-end: all ten actions now write a result for minimal/empty-field
payloads instead of timing out. New smoke test docker/wg/serve-request.test.sh
(root-bypassed, no WireGuard needed) guards both regressions. shellcheck clean,
348 PHP tests pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Rebuild the README around the product's "Tactical Terminal" design system
(dark graphite, signal-orange, cyan, mono type) so it reads as a professional,
self-hosted enterprise product instead of a plain text file:
- docs/assets/clusev-banner.svg — hero banner with the Clusev wordmark, tagline,
brand chips, and a live "fleet" node-graph (control-plane hub → server cards
with the online/warning/offline status triad + CPU bars and animated data
pulses). Self-contained: the real product fonts (Chakra Petch + JetBrains Mono)
are embedded base64 so Gitea's <img>-rendered SVG shows pixel-correct type.
- docs/assets/clusev-architecture.svg — Operator → Control plane → Fleet diagram
(HTTPS/WS · Reverb in, SSH+SFTP · phpseclib out), same brand language.
- README.md — centered hero, brand-matched shields badges, anchor nav, a
"why" three-up, a Features table, the architecture diagram + a tech-stack
table, and the existing install / domain-TLS / updating / security-recovery
content kept verbatim (now with a WireGuard feature row). All graphics are
external SVG files (Gitea strips inline <svg>); relative paths resolve on the
default branch.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Verified dead-code + duplication cleanup over the WireGuard feature surface
(found via an audit workflow, every item adversarially re-verified by grep):
Dead code removed:
- lang keys not_configured_title / not_configured_body (both locales) — orphaned
when the unconfigured empty state became the in-page setup form (setup_title /
setup_intro). 0 references anywhere. de/en parity kept (83/83).
- WgTraffic::series() 'buckets' return key — written, never read by blade/route/test.
- blade $fmtB byte-formatter — byte-identical duplicate of the top-level $fmtBytes
already in scope; collapsed onto $fmtBytes.
DRY / readability (behavior-preserving):
- Index: PEER_NAME_RE constant (was the same regex literal at 5 sites), portInRange()
+ clampWindow() helpers (window default now self::WINDOWS[0]), openConfirm() helper
(5 confirm-modal openers shared one dispatch), onFlag() helper (2 apply handlers).
- WgBridge: validPort() helper (port check was duplicated verbatim).
- clusev-wg.sh: _write_server_conf() shared by both setup paths (the byte-identical
server bring-up was copy-pasted); single CIDR_RE constant for the 3 subnet checks;
fixed a stale header comment that still claimed SSH is "NEVER matched" (the v0.9.40
ssh-lock can match 22); documented the defence-in-depth re-validation in _set_*.
- WgStatus: documented server.pubkey is part of the /wg-status.json contract (kept).
No DB/migration changes (peer_name write-only column left intact — needs a
coordinated migration). 348 tests pass, shellcheck clean, Pint clean, R12 verified
in both the configured + unconfigured states (200, no console errors, no leaked
keys, traffic formatter renders 1.5 MB / 200 KB correctly).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Two dashboard additions to the WireGuard page:
- Peer config download: the show-once new-client view gets a "Download"
button next to the QR code (streams clusev-wireguard.conf) plus an
inline hint explaining QR-on-phone vs download-on-PC import.
- SSH lock (port 22): an optional toggle in WireGuard settings that walls
off host SSH to the tunnel via a dedicated CLUSEV-WG-SSH chain on the
host INPUT chain. Fail-open by construction: ESTABLISHED,RELATED + lo +
wg0 + WG-subnet RETURN before the DROP (live sessions survive, SSH over
the tunnel stays open); _ssh_gate_on refuses while wg0 is down; the boot
unit's ConditionPathExists=wg0 and an in-function self-guard keep a
broken tunnel from ever applying the DROP; ssh_gate_apply re-validates
the subnet shape and fails open on a corrupt wg.env. Strong lock-out
warning shown inline AND in the confirm modal (recommend a backup peer).
New CLI escapes: clusev wg ssh-lock / ssh-unlock; clusev wg down clears
both gates. Panel gate and SSH gate are independent toggles.
Write-bridge: new gate-ssh-on/gate-ssh-off actions (no args, host
whitelist + ConfirmToken wgSshGate, single-use uid-bound). Status collector
now reports gate.ssh. Adversarial lock-out review: all 8 failure modes
refuted. 348 tests pass, shellcheck clean, Pint clean, R12 verified.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
boban