Previously sendEmail() only logged on error/suppression paths; success
was silent. After prod deploy of c71c401 it was unclear whether the
first alert tick had actually delivered email or not - the answer was
yes but had to be derived from 'no error log + recipient query showed
matching device'. Add a log line on success so future observability
doesn't require detective work.
Replaces the unused EMAIL_WEBHOOK_URL stub with a real Microsoft Graph
Mail.Send pipeline via @azure/msal-node client-credentials flow. Prior
state on prod: every alert email was logged to journalctl and never
sent (21 fallback log lines per hour for the chronic-offline devices).
Four coordinated changes shipped as one commit since they're all part
of making email delivery actually work responsibly:
1. services/email.js (NEW): Graph send via plain HTTPS (no SDK), in-memory
MSAL token cache (refresh 60s pre-expiry), graceful stdout fallback
when GRAPH_* env vars absent. Drop-in replacement for the old webhook.
2. services/alerts.js refactored: sequential await around sendEmail (was
parallel fire-and-forget; first run hit Graph's MailboxConcurrency 429
ApplicationThrottled on a 30-device backlog). Sequential at ~250ms per
send takes 5-8s for the full backlog, well within the 60s tick. Also:
24h long-offline cutoff to stop nagging about chronic-offline devices
(the 20,000+ minute ones); 2-hour dedup window (was 1h) via a generic
shouldSendAlert(type, id, windowMs) helper that future alert types
(payment_failed, plan_limit_hit, etc.) can reuse.
3. Preferences UI: single checkbox in settings.js Account section bound
to users.email_alerts. Saved via the existing Save Profile button. PUT
/api/auth/me extended to accept email_alerts. requireAuth middleware
SELECT now includes email_alerts so it propagates via req.user.
4. Dev safety net: GRAPH_DEV_RESTRICT_TO env var as an allow-list. When
set, only listed recipients reach Graph; everyone else is suppressed
with a log line. Prevents local dev (which often runs against fresh
prod DB copies) from accidentally emailing real prod users. UNSET on
prod systemd unit so production fans out normally.
Also: package.json scripts use --env-file-if-exists=.env so local dev
picks up .env automatically (Node 20.6+ built-in, no dotenv dep). Prod
runs via systemd ExecStart and is unaffected. server/.gitignore added
to keep .env out of git.
Smoke verified end-to-end:
- Sequential send pattern verified (a prior parallel-send tick had hit
Graph's MailboxConcurrency 429 on 30 simultaneous sends; sequential
at ~250ms each completes the same backlog without throttling)
- 24h cutoff silenced 20/21 prod devices on the next tick
- Dev restrict suppressed the 1 within-24h send
- User-preference toggle flipped via UI -> DB -> alert path silently
continued before reaching even the suppression log
/me's accessible_workspaces query gains a device_count field via a
correlated subquery on workspaces.id - WHERE workspace_id = w.id
strictly excludes the unclaimed pair-pool (workspace_id IS NULL fails
equality). Added to both query branches (platform_admin LEFT JOIN and
regular INNER JOIN); microseconds per row at current scale (~37 rows
worst case), not optimizing.
Frontend appends the count to the muted org-name line with a middle-dot
separator: 'Acme Studios . 2 devices'. Singular/plural respected via the
existing tn() helper convention; 'No devices' for empty workspaces. New
formatResourceCount(n, keyBase, zeroKey) helper is generic so the same
shape can wire users/playlists/schedules counts later without refactor.
New i18n keys: switcher.devices_count_one, switcher.devices_count_other,
switcher.no_devices. Added to en.js only; other locales fall back to en
via the existing lookup chain (verified in i18n.js:19).
API smoke verified: switcher-test sees Studio A=2, Field Crew=2;
dw5304 (platform_admin) sees all 37 workspaces with their device counts
varying 0-4; single-workspace zero-device user (geoff.case) sees 0.
Teams in its pre-Workspaces form is being paused while the feature is
redesigned as a user-grouping primitive within the new Workspaces
architecture. The original Teams data model had no workspace-awareness
and was effectively non-functional after Phase 2.2 (every route migrated
away from team_id), but the UI remained reachable and allowed users to
accumulate orphan data while believing they were configuring access
control.
Hide the Teams sidebar nav entry to prevent new entries to the UI.
/api/teams now returns 503 Service Unavailable with a 'feature
redesign in progress' message. Existing teams/team_members/team_invites
table data is preserved indefinitely for forward migration to the
future teams design.
Bonus: requireAuth middleware fires before the catch-all so unauthenticated
callers see the standard 401 instead of the 503 redesign message - avoids
exposing the 'feature being redesigned' signal to unauthenticated probes
or fingerprint scanners.
The previous comment claimed defParamCharset:'utf8' fixed multipart
filename header decoding. It doesn't - that option only fires for the
RFC 5987 encoded filename*=utf-8''... form, which clients rarely send.
The actual UTF-8 recovery happens in the storage.filename callback
(added in d679ca8) via Buffer.from(name,'latin1').toString('utf8').
The option is kept set for the rare RFC 5987 case but the comment no
longer overclaims what it does.
busboy reads the Content-Disposition filename="..." header value as
latin1 by default - even with defParamCharset:'utf8' set, that option
only applies to RFC 5987 encoded filename*=... params, which most
clients (browsers, curl, programmatic HTTP) don't send. Modern clients
send raw UTF-8 bytes for non-ASCII filenames; busboy interprets those
bytes one-byte-per-char as latin1, producing a JS string like 'A-tilde
+ quarter-mark' for 'u-umlaut'. JS then re-encodes that string as UTF-8
on the way to SQLite, yielding 4 bytes (c3 83 c2 bc) for what should be
2 bytes (c3 bc). Classic double-encoding mojibake - shows up in the UI
as 'BegrA-tilde...' instead of 'Begru-umlaut...'.
Fix: in the multer filename callback, re-decode file.originalname from
latin1 to utf8 to recover the original byte sequence. Mutating
originalname here propagates to every route handler reading
req.file.originalname (POST /, PUT /:id/replace, and any future upload
route using the same middleware).
This is the actual visible-mojibake bug semetra22 reported. The prior
commit b677752 (NFC normalize in safeFilename) handles a separate but
related case (macOS NFD clients sending decomposed forms); both fixes
compose correctly - latin1->utf8 first restores the byte sequence,
then NFC normalize collapses NFD into composed form.
Smoke verified by sending raw UTF-8 multipart from a Node https client
(no shell escaping). NFC input 'Begru-umlaut-essungsscreens.jpg' with
bytes c3bc c39f arrives clean (was c383c2bc c383c29f before). NFD input
'u + combining diaeresis' arrives as composed NFC c3bc after both fixes.
Single line change to safeFilename() in routes/content.js: add
.normalize('NFC') before sanitizeString. Covers all 4 user-facing
filename storage sites (POST /, POST /remote, POST /embed, PUT /:id
rename) since they all flow through safeFilename.
Fixes macOS NFD vs Linux NFC mismatch on filename storage that mangled
umlauts (ae/oe/ue/ss) in displayed filenames. macOS clients send
NFD-decomposed names (e.g. 'u' + combining diaeresis U+0308 instead of
the precomposed U+00FC); Linux + most renderers expect NFC. Without
this, names like 'Begruessungsscreens.jpg' arrive with the combining
char floating and display as mojibake.
Reported by semetra22 in Discord with extraordinarily good debugging
narrowing (rename works, upload doesn't = bug is in upload path).
Single-point fix at the convergence of all user-facing filename flows.
Existing NFD-mangled rows in DB not backfilled; users can re-upload or
rename to repair. Optional one-time UPDATE backfill captured as follow-up
in handoff doc.
Smoke verified by invoking safeFilename directly on NFD + NFC inputs of
'Begruessungsscreens.jpg' - both produce identical NFC-normalized bytes
(42656772c3bcc39f756e677373637265656e732e6a7067).
Fix: at connect, enumerate the user's accessible workspace_ids (direct workspace_members + org_owner/admin paths + platform_admin 'all') via new accessibleWorkspaceIds() helper in lib/tenancy.js; socket.join one room per workspace. All 12 dashboardNs.emit sites across deviceSocket / heartbeat / server.js / devices route / video-walls route now route via dashboardNs.to(workspaceRoom(...)).emit() with the workspace looked up from the relevant device or wall. New lib/socket-rooms.js holds the helpers and breaks a circular dependency (dashboardSocket already requires heartbeat, so heartbeat can't require dashboardSocket).
Inbound 6 commands rewired to canActOnDevice(socket, deviceId, tier): request-screenshot is read tier (workspace_viewer+); remote-touch/key/start/stop and device-command are write tier (workspace_editor+). Platform_admin and org_owner/admin always pass via actingAs. Legacy admin/superadmin branch dropped.
Lifecycle note: workspace-switch already calls window.location.reload (Phase 3 switcher), which forces a fresh socket with updated memberships - no per-emit re-evaluation needed.
Smoke tested with 3 simultaneous socket.io-client connections (switcher-test, swninja, dw5304 platform_admin) + direct canActOnDevice invocation for 6 user/device/tier combinations. All 9 outbound isolation cells and all 6 permission gates pass. Fixture mutation: switcher-test's Field Crew membership flipped from workspace_editor to workspace_viewer to exercise the read/write tier split in one login.
KNOWN REGRESSION (Phase 3 fix): platform_admin / superadmin no longer has cross-workspace 'see everything' view. Every route migrated tonight (2.2a-2.2m) deliberately removed the role-based bypass per design doc - cross-workspace visibility will come via dedicated admin endpoints in Phase 3, not magic role bypasses. Until Phase 3 ships, platform admins must switch-workspace to see other workspaces' data.
Express's req.ip was resolving to a Cloudflare edge address (e.g.
172.70.x.x) for any request fronted by Cloudflare, because trust proxy
was set to '1' — that trusts the immediate hop, which IS Cloudflare.
All activity_log rows from API paths captured the proxy, not the
client. The WebSocket path was unaffected and recorded the real IP.
Two layers of defense:
1. trust proxy now lists Cloudflare's published v4 + v6 ranges plus
loopback / linklocal / uniquelocal (config/cloudflareIps.js). With
this list req.ip resolves to the original client when fronted by
CF, and X-Forwarded-For from any non-trusted source is ignored —
so the value can't be spoofed.
2. New getClientIp(req) helper in services/activity.js prefers the
CF-Connecting-IP header but only honors it when the immediate TCP
peer is itself a trusted address. Same gate as trust proxy, so a
visitor who hits the origin directly with a forged header is
logged at their real address.
Routed all five activity-log call sites (auth login success/failure,
admin password reset, generic activityLogger middleware, and the
in-memory rate-limiter key) through the helper.
Logging-only change. No schema changes. Existing rows are not
modified — fix applies to new entries going forward.
Verified locally:
- Bare loopback hit logs 127.0.0.1 (not a proxy address).
- Helper unit cases including an untrusted peer (203.0.113.7) sending
a forged CF-Connecting-IP correctly fall back to the real peer.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Wall editor: replaces the small grid with a Figma-style pan/zoom canvas. Each
display is a rectangle that can be dragged/resized to match its physical
arrangement; a separate semi-transparent player rect overlays the screens and
defines what content plays where. Drag empty space to pan, wheel to zoom,
"Center" button auto-fits content. Per-rect numeric x/y/w/h panel; arrow keys
nudge by 1px (10px with shift). Negative coordinates supported for screens
offset above/left of the origin. Coords rounded to integers on save.
Wall rendering: each device receives screen_rect + player_rect, maps the
player into its viewport with vw/vh and object-fit:fill so vertical position
of every source pixel is identical across devices that share viewport height.
Leader emits wall:sync at 4Hz with sent_at timestamp; followers apply
latency-adjusted target and use playbackRate ±3% for sub-300ms drift,
hard-seek for >300ms. Followers stay muted; leader unmutes via gesture with
AudioContext priming and pause+play retry to bypass Firefox autoplay.
"Tap to enable audio" overlay as a final fallback.
Reconnect handling: server re-evaluates leader on device:register so the
top-left tile reclaims leadership when it returns. Followers emit
wall:sync-request on entering wall mode (incl. reconnect) so they snap to
position immediately instead of drifting until the next periodic tick.
Group dissolve: removing a device from its last group clears its playlist
to mirror wall-leave semantics. Leaving a group with playlists on remaining
groups inherits the next group's playlist.
Dashboard: walls render as their own card section (hidden the device cards
they contain). Multi-select checkboxes on cards + "Create Video Wall" toolbar
action that creates the wall, removes devices from groups, and opens the
editor. dashboard:wall-changed broadcast triggers live re-render. Per-card
playback progress bar driven by play_start events forwarded from devices.
Security: PUT /walls/:id/devices verifies caller owns each device (or has
team-owner access via the widgets pattern), preventing cross-tenant device
takeover. wall:sync and wall:sync-request validate that the sending device
is a member of the named wall; relay re-stamps device_id with currentDeviceId
so clients can't spoof or shadow-exclude peers.
Schema: video_walls += player_x/y/width/height, playlist_id;
video_wall_devices += canvas_x/y/width/height. All idempotent migrations.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Password reset for other users:
- New PUT /api/auth/users/:id/password endpoint
- Superadmin can reset any local user; admin can reset role=user
members of teams they own only (cannot reset other admins or
superadmins, cannot self-reset — that goes through PUT /me with
current_password)
- OAuth users are excluded (no password to reset)
- Rate-limited 20 req/min/IP to cap blast radius if an admin session
is compromised
- Explicit audit log entry "password_reset_for_user / target: <email>"
on every reset; activity logger's summarizeAction never reads the
password field, so the password value is not stored anywhere
Frontend: Reset Password button in the Admin user table and Settings
> User Management table. Shown only for local-auth users that aren't
the current user; prompts for an 8+ char password.
Widgets visibility fix:
- routes/widgets.js had `const isAdmin = req.user.role === 'superadmin'`
which mislabeled superadmin as admin and silently restricted real
admins (role=admin) to seeing only their own widgets. Now matches
/auth/users behavior: superadmin sees all, admin sees own + public
+ widgets owned by members of teams they own, user sees own + public.
7 new i18n keys (admin.reset_password, admin.prompt_reset_password,
admin.toast.password_min_8, admin.toast.password_reset, and the
matching settings.user.* / settings.toast.* trio). 1024 keys total,
parity 100% across en/es/fr/de/pt.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Add-Display modal in index.html: marked translatable elements with
data-i18n / data-i18n-placeholder / data-i18n-html attributes
- app.js: translateStaticDom() walks data-i18n* on init and on every
language-changed event so static HTML stays in sync
- server/player/index.html: standalone player gets its own inline
PLAYER_I18N table (en/es/fr/de/pt) with a tiny _t() helper. Reads
rd_lang from localStorage (set by dashboard) so the player picks up
the same language. Translates info overlay, setup screen, and
status messages.
- 1018 keys total in dashboard locales, parity 100%.
This completes the wiring; Android resources are next.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Landing page (frontend/landing.html):
- Title now includes "Self-Hosted" for that keyword
- Description appended "MIT licensed."
- Keywords aligned to spec (digital signage raspberry pi, digital
signage android tv, video wall software, kiosk software, etc.)
- SoftwareApplication JSON-LD: added applicationSubCategory
"DigitalSignage", license URL, refreshed description
- Image alt text + og:image:alt + twitter:image:alt now include
"open-source digital signage"
- New Resources section above the CTA with 6 cards linking to all
new guides and comparison pages
- Footer rewritten as a 5-column grid (Brand / Guides / Compare /
Project / Legal) with the new internal links
New SEO pages, all dark-themed, mobile-responsive, ASCII-only:
- frontend/css/seo-page.css (shared nav/footer/article/table styles)
- frontend/compare/yodeck-alternative.html
- frontend/compare/screencloud-alternative.html
- frontend/compare/optisigns-alternative.html
- frontend/guides/raspberry-pi-digital-signage.html
- frontend/guides/digital-signage-android-tv.html
- frontend/guides/self-hosted-digital-signage.html
Each new page has unique title/description/canonical, OG and Twitter
card tags, BreadcrumbList JSON-LD, single h1, proper h2/h3 nesting,
visible breadcrumb, comparison table or step-by-step ordered list,
"Related guides" cross-link block, and a CTA.
Sitemap (frontend/sitemap.xml): added all 6 new URLs with appropriate
priority (0.8 for compare pages, 0.9 for guides). Existing landing
(1.0) and legal pages preserved.
Robots (frontend/robots.txt): allow /compare/ and /guides/, disallow
/player (was previously allowed by mistake).
Server (server/server.js): added explicit GET /sitemap.xml and
GET /robots.txt routes ahead of the static middleware so the
Content-Type is guaranteed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
video.play().catch(() => {}) silently swallowed the rejection from the
browser's autoplay policy, so when a user click triggered the unmute
path the video paused (browser side-effect of unmuting a muted-autoplay
video) and never resumed.
Surface the play() rejection in the log, and fall back to muted playback
if the unmuted play() is blocked. Same for the YT side: explicitly set
volume on unmute. Bumped SW cache to v9.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The remote-control feature dispatches synthetic click events on the
player when the dashboard forwards touches. The global click handler
called requestFullscreen() on every click, but the browser only honors
that API for trusted user gestures — synthetic events rejected with
"Permissions check failed" / "API can only be initiated by a user
gesture", spamming the console for the duration of any remote session.
Gate the fullscreen request on event.isTrusted. Local user clicks still
trigger fullscreen; remote-control taps no longer try (and fail).
Bumped SW cache to v8.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
createYoutubeEmbed set container.style.position = 'relative' to anchor
the click-to-unmute overlay. That overrode #playerContainer's
position:fixed/inset:0 — the container fell into normal flow with
zero height (the YT iframe inside has no intrinsic size), so the new
absolute-positioned iframe rendered as 100% of 0 = black screen.
The container is already position:fixed, so absolute children anchor
to it correctly without the override. Removed the line. Bumped SW
cache to v7.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The previous CSS fix used 100% width/height but YT.Player can bake in
300x150 fallback pixel dimensions if the placeholder isn't laid out at
construction time. Inline pixel dimensions beat percentage CSS at
equal specificity, so the iframe stayed small.
Use absolute positioning with !important to force fullscreen over
whatever YT set inline. Bumped sw cache to v6 to invalidate the
previously-cached player HTML.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The .zone iframe sizing rule only applies to multi-zone layouts. In
fullscreen single-zone mode the YT IFrame API replaces our placeholder
div with an iframe directly inside #playerContainer, where no CSS rule
sized it — leaving it at the iframe default size (~300x150) and
producing a tiny square in the corner. Added explicit rules so any
iframe child of #playerContainer fills the viewport.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
userHasInteracted was initialized from localStorage('rd_audio_unlocked')
on every page load. Browser autoplay policy is per-document, so a flag
from a prior session does not actually grant autoplay rights — but the
player code used it to decide whether to start the YouTube embed muted
(autoplay-able) or unmuted (blocked). Result: kiosks with the flag set
loaded a YT embed with mute=0 that the browser refused to start.
- userHasInteracted now always starts as false. The cold-load tap
overlay flips it to true on real gesture; the 5s auto-dismiss leaves
it false and playback stays muted (still allowed).
- unlockAudio() now also calls activeYtPlayer.unMute() so the muted
embed unmutes immediately when the user finally taps the overlay.
- Removed the now-unused localStorage writes of rd_audio_unlocked.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Browser autoplay policy is per-document — a previous session's
localStorage flag does not grant the new page autoplay rights. The
'audio previously unlocked, skipping tap overlay' branch was racing
with YouTube's autoplay block, leaving the player stuck on a paused
embed.
Removed the skip-overlay optimization. The existing 5s auto-dismiss
+ muted-connect fallback still handles unattended kiosks, and a real
user only needs to tap once per cold load to get audio.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The cached-playlist restore at the top of the script synchronously calls
playCurrentItem -> renderContent -> createYoutubeEmbed, which references
ytGeneration / activeYtPlayer / ytApiReady / ytApiCallbacks. Those were
declared with `let` further down in the script, so the references hit
the temporal dead zone and threw on every cold start with a YouTube
item in the cached playlist:
Uncaught ReferenceError: can't access lexical declaration
'ytGeneration' before initialization
Hoisted the four declarations to the top of the script alongside the
other player state.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two pre-existing bugs surfaced during deploy:
- /api/devices/:id/screenshot fell back to a query referencing
screenshots.created_at, but the schema column is captured_at. Threw
SqliteError 500 whenever the in-memory cache was cold (e.g. just
after a server restart).
- The SPA catch-all at /* served index.html for non-/api paths but did
nothing for unmatched /api/ paths — the response hung until the
upstream timeout (524 from Cloudflare at 15s). Now returns 404 JSON.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
LOW 1 (notes XSS): device.notes textarea content now goes through
esc(). Notes weren't in the sanitizeBody allow-list at write time, so
HTML in the field would render unescaped on the device-detail page.
LOW 2 (CSP): enabled Helmet contentSecurityPolicy with default-src
'self', script-src 'self', style-src 'self' 'unsafe-inline', plus the
data:/blob:/https: image and media sources the player needs. Strict
script-src blocks <script> injection; script-src-attr 'unsafe-inline'
keeps existing inline onclick handlers working until they can be
refactored to addEventListener (TODO comment in code).
CSP applies to /app and most other paths. Skipped on the public
widget and kiosk render endpoints, the landing page, and /player —
those legitimately need inline scripts/styles. upgrade-insecure-
requests is explicitly disabled so HTTP-only self-hosted LAN
deployments aren't broken.
Refactored two inline onclick handlers in index.html to data-close-
modal attributes wired by a delegated listener in app.js. Was the
only blocker for /app under strict script-src.
LOW 3 (CORS): Express CORS now only allows screentinker.com (and
subdomains) + localhost in production. SELF_HOSTED=true bypasses the
allowlist (operator owns their deployment). Development mode stays
open. Same policy applied to the Socket.IO CORS config which was
previously origin: '*'. Native clients (Android, server-to-server,
kiosk iframes) send no Origin and pass through unchanged.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
HIGH 1 (teams IDOR): POST/DELETE /api/teams/:id/devices now require the
caller to own the device before assigning or detaching it. Without this
check, any team member could pull any device into their team via UUID
guess and gain remote-control access.
HIGH 2 (schedules IDOR): PUT /api/schedules/:id now re-verifies
ownership of every changed target field — device_id, group_id,
content_id, widget_id, layout_id, playlist_id. Previously only the
schedule owner was checked, letting users fire arbitrary content on
victim devices via update.
HIGH 3 (filename XSS): file.originalname captured by multer bypassed
sanitizeBody. New safeFilename() wraps every INSERT path (multipart
upload, remote URL, YouTube). Frontend sinks now go through esc() in
content-library.js, device-detail.js, video-wall.js. Web player gets
an inline escHtml helper for its info overlay where filenames, device
name, and serverUrl land in innerHTML.
HIGH 4 (kiosk public XSS): config.idleTimeout is now coerced via the
existing safeNumber() helper at both interpolation sites. A crafted
value with a newline can no longer escape the JS line comment to
inject arbitrary code into the public render endpoint.
HIGH 5 (folder DoS): POST /api/folders enforces a per-user cap of 100
folders (429 on overflow). Superadmin exempt.
MED 1 (SSRF): ImageLoader.decodeUrl rejects any URL scheme other than
http(s) so a malicious remote_url can't read local files via file://.
On the server, validateRemoteUrl() is extracted and now also runs on
PUT /api/content/:id remote_url updates — previously the SSRF check
only fired on POST.
MED 2 (fingerprint takeover): the WS device:register fingerprint
reclaim path now rejects takeover while the target device is online or
within 24h of its last heartbeat. A leaked fingerprint can no longer
hijack an active display.
MED 3 (npm audit): bumped uuid 9.x -> 14.0.0 (v3/v5/v6 buffer bounds
CVE; we only use v4 so not exploitable, but clears the audit). path-
to-regexp resolved to 0.1.13 via npm audit fix. 0 vulns remaining.
MED 4 (folder admin consistency): ownedFolder() and the content.js
folder_id move check now both treat only superadmin as privileged,
matching GET /api/folders. Previously a plain "admin" could rename
or delete folders they couldn't see, and could move content into
folders they couldn't list.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
multer/busboy decode multipart filename headers as latin1 by default,
which mangled umlauts and other non-ASCII characters end-to-end
(Größe.jpg arrived as Größe.jpg and was stored that way). Setting
defParamCharset: 'utf8' on the multer options makes the entire
upload pipeline consistent UTF-8.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
New content_folders table with hierarchical parent_id and per-user
scoping. content.folder_id added (ON DELETE SET NULL so deleting a
folder drops items back to root). New /api/folders route exposes
list/create/rename/move/delete with cycle detection on move.
Content library UI: breadcrumb navigation, subfolder grid, "+ New
Folder" creates inside the current folder, drag-and-drop content
items onto folder cards to move them, and the edit modal has a
folder dropdown showing each folder's full path.
Per-user scoping is enforced server-side: every folder query
filters by user_id, and folder ownership is checked on both folder
mutations and content.folder_id updates.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Server-side: when a device reconnects on a fresh socket while the old
TCP zombie is still around, the old socket's eventual disconnect handler
flipped the device offline and removed the new heartbeat entry. Now we
proactively evict any prior socket on register and ignore disconnects
from sockets that are no longer the registered one for that device_id.
Frontend: dedupe devices by id from the API response and only render
each device in the first group it belongs to (multi-group membership
is still tracked for the Manage modal).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
When DISABLE_REGISTRATION=true (or 1), POST /api/auth/register returns
403 with a clear error. OAuth endpoints (/google, /microsoft) also
refuse to auto-create new accounts — existing OAuth users can still
sign in. First-user setup (empty users table) is always allowed so a
fresh install can still be initialized.
GET /api/auth/config now returns registration_enabled so the login
view can hide the "Create Account" button and the trial banner when
registration is off. Absence of the flag is treated as enabled for
back-compat with older servers.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Extend the public /api/content/:id/file gate to unlock content referenced
by widgets (previously only playlists unlocked it), so device browsers
and kiosk iframes can fetch logos and background images that widgets
embed.
Security: scope the widget lookup to the content owner's widgets only
(w.user_id = content.user_id). Otherwise a user could unlock another
user's content file by creating their own widget whose config references
the victim's content UUID. The pre-existing playlist gate has the same
shape and is left for a separate fix.
Also adds a 30/min rate limit on POST /api/widgets/preview, which
inlines user content as base64 and is memory-intensive.
Perf note: the widgets.config LIKE scan is O(n). Fine at current scale;
revisit with a content_widget_refs join table if the widget table grows.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Lobby-style tenant/room directory with vertical marquee, seamless loop via
content cloning, pixel shift + bg pulse for anti-burn-in, rotating background
images with crossfade. Supports logo, title, footer, subtitles per entry,
and Available (green) state. All user strings rendered via textContent in
browser — no server-side HTML escaping of entries needed.
Also refactors render dispatch into renderWidgetHtml() and adds a POST
/preview endpoint that inlines user-owned image content as base64 data
URIs so the editor can preview unsaved widgets. Preview is gated by:
- image/* MIME only
- 10 MB size cap
- user_id ownership check
- path traversal guard via basename + resolve
Unknown widget_type on /preview returns 400.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
