The player dock used to sit fixed over the scrolling page, so it needed an
opaque background. The shell is now a fixed-height frame where only the
content column scrolls, letting the dock stay transparent like the sidebar
and queue. Scroll reset/restore on navigation moves to restoreScroll(), and
the stable scrollbar gutter moves from <html> to the content scroller.
Following lists are public; followers are visible to the user and to mutual followers.
Adds GET /users/{id}/followers and /following, exposes isFollowing and canViewFollowers
on the profile, and returns isFollowedByViewer from user search so the web player no
longer calls /is-following per user.
Discover used a hardcoded list of genre tiles that searched track titles,
and the upload form kept its own copy of the genre list and compatibility
rules. Both now come from the backend.
Backend:
- GET /genres returns genres with at least one track and their counts.
- GET /genres/available returns every assignable genre with the genres
it cannot be combined with, so the rules live only in
GenreCompatibility.
- GET /tracks accepts ?genre= (case-insensitive, 400 on unknown values).
web-player:
- Discover tiles are built from /genres; each opens /explore?genre=X, a
song list filtered by that tag. Unknown genres fall back to a label
derived from their name.
- The upload form loads /genres/available, disables incompatible genres
as you pick, and the server validates before reserving upload URLs.
- lib/data/trackGenres.ts is removed; its helpers merge into
lib/data/genres.ts.
- Regenerate the typed API client.
"Listening time this week" summed the full duration of every track whose
stream ticket was issued, so pressing play and skipping counted the whole
song. Playback is now measured on the client and reported to the API.
Backend:
- ListeningHistory gains PlayedSeconds, LastProgressAt and IsCounted
(migration TrackPlayedSeconds marks existing rows as counted).
- /tracks/{id}/stream returns a listenId; new
PUT /tracks/listens/{listenId}/progress stores the accumulated seconds,
capped by track duration and elapsed wall-clock time, monotonic and
owner-only.
- A listen counts once it reaches min(30s, 50% of the track). Weekly
stats, streak, recent history/albums, mixes and listensCount only use
counted listens; seconds sum PlayedSeconds.
- ListeningHistoryCleanupJob (daily) deletes uncounted listens older
than 8 days.
web-player:
- ListenTracker accumulates played time (seeks are ignored) and flushes
on pause, track change, ended, every 30s, and on pagehide.
- Proxy route /api/listens/[id]/progress; repeat starts a new listen per
loop.
Per feedback, the solid color-mix bar was wrong — sidebar, queue and
the mobile header should stay on the plain --mf-bg (the same subtly
hue-shifting black used everywhere else), exactly like light mode
already does, not a separately-boosted colored panel. Drop --mf-chrome-bg/
--color-chrome and switch those three back to bg-bg.
The general dark-mode richness bump (tc(), --mf-bg/--mf-elevated/etc.)
and the glass-bar background-color fix from the previous commit stay;
only the separate chrome treatment is reverted.
Mirror the earlier light-mode pastel/colorful pass for dark mode:
boost tc()'s lightness and chroma, and richen --mf-bg/--mf-elevated/
--mf-panel-bg/--mf-bar-bg/--mf-hairline the same way, so dark mode
reads as colorful instead of near-grayscale.
Sidebar, Queue, and MobileHeader used bg-bg directly, so in dark mode
they stayed almost black regardless of hue. Add --mf-chrome-bg (an
alias to --mf-bg in light mode, a solid color-mix of --mf-hero-tint
into --mf-bg in dark mode) and switch those three to bg-chrome, so
they read as a full, solid colored bar in dark mode instead of a
fading gradient (per feedback: a whole bar, not a gradient).
Also fixes glass-bar (PlayerBar/TabsBar): it set
`background-image: var(--mf-bar-bg)`, but --mf-bar-bg is a plain
color, not a valid <image>, so the declaration was silently dropped
and these bars never actually showed their tint in either mode. Changed
to background-color.
Button used Tailwind's bare `transition` utility, whose default
property list includes background-color. Since the button's bg is a
CSS custom property recomputed every frame by the same 600ms hue
animation the main background uses, that extra transition smoothed it
on a different (~150ms) timing, so its color change didn't match the
page background's.
Add a btn-transition utility that mirrors Tailwind's default
transition list minus background-color, so the button's background
tracks the hue animation frame-for-frame like the rest of the app,
while hover/focus feedback (text color, ring, filter) still transitions
normally.
--mf-hero-bg/--mf-spotlight-bg's overlay tint was at 16% alpha, which
after the pastel lightness/chroma boost read as a fairly solid block
of color rather than a soft wash. Halve it to 8% so more of the plain
page background shows through underneath.
Keep the secondary button's color-mix ratio (in layout.css) in sync at
8% too, since it's built to replicate exactly what the surface behind
it looks like.
Sidebar, Queue, PlayerBar, TabsBar, and the hero/spotlight surfaces all
had an extra CSS `transition: background 0.5s ease-out` (theme-transition/
glass-bar) layered on top of backgrounds that are already animated
frame-by-frame over 600ms by the hue-interpolation loop in +layout.svelte.
That extra transition made them visibly lag behind and take a different
amount of time than the main app-backdrop, which has no such transition
and therefore tracks the JS animation exactly. Drop the redundant
transition so every hue-driven surface updates in lockstep.
Also fixes a dangling --color-btn-hover: var(--mf-btn-bg-hover) that
referenced an undefined custom property (theme.css never defined
--mf-btn-bg-hover after an earlier partial commit), which silently
broke the secondary button's hover background.
backdrop-blur samples whatever is actually behind the element, so two
secondary buttons with identical classes could render differently
depending on nearby content (e.g. dark album art bleeding into the
blur near the spotlight card). Replace it with a fixed color-mix that
reproduces the same result mathematically instead of sampling pixels:
45% --mf-elevated over the hero/spotlight surface's own tint color
(16% --mf-hero-tint over --mf-bg, the same ratio hero-surface itself
uses), computed via CSS color-mix() rather than a backdrop-filter.
Verified against the previous blur version: both now compute to the
identical color(srgb 0.9329 0.9611 0.9332), matching what the blurred
button rendered against the same hero background, but fixed so every
secondary button is now guaranteed pixel-identical regardless of what
happens to be nearby.
Sampling whatever's behind the button via backdrop-filter meant two
secondary buttons could render differently depending on what happened
to be nearby (e.g. dark album art bleeding into the blur), so
"Explorar musica" and "Aleatorio" ended up visibly different colors
despite being the same variant. Replace it with a fixed, hue-tied
translucent tint (bg-accent-tint, the same token family accent already
uses) so every secondary button computes to the identical color, ties
into the same dynamic accent hue as the rest of the app, and reads as
a softer sibling of the accent button instead of an unrelated gray.
Hover now steps up to bg-accent-btn (same hue family, stronger) instead
of a flat, hue-agnostic gray. Removed the now-unused --mf-btn-bg-hover/
--color-btn-hover tokens.
secondary was fully bg-transparent (invisible at rest) or a flat neutral
overlay (clashed next to accent-tinted buttons and looked different
depending on what surface it sat on). Use bg-elevated instead, which is
already derived from the same hue as --mf-bg, so the button is always a
faint, minimal step above whatever page background it's on, and shifts
with it automatically.
Drop the now-unused --color-btn/--mf-btn-bg tokens.
secondary used a flat, static tint that only worked over a plain
surface, so it looked out of place on the hero's hue-shifting gradient.
Add a backdrop blur (no border, per feedback) so it always reads
against whatever's actually behind it. Gave the button its own
background tokens instead of aliasing --mf-surface-2, and dropped the
now-unused --color-btn-glass tokens left over from the removed glass
variant.
The secondary/glass Button variant used a flat, fairly opaque black or
white tint with no blur, so it didn't read as glass and clashed with
the hue-shifting hero background. Add a backdrop blur and a hairline
border (matching the app's existing glass-bar/glass-panel treatment)
and lower the tint opacity.
Wire up the already-computed but unused --mf-accent-muted token as
--color-accent-muted, and use it for the accent Button variant's text
instead of the higher-chroma --color-accent-soft, so it reads softer.
Moves the login screen from /login to /auth, logout from /logout to
/auth/logout (alongside the existing /auth/login and /auth/callback
endpoints), and the public profile route from /u/[id] to /user/[id].
AUTH_REDIRECT_URI and AUTH_POST_LOGOUT_URI already point at
/auth/callback and / respectively, so no Zitadel config change is
needed for the auth move.
Strip //, /* */ and XML doc comments from all backend .cs files, and
JS/TS/CSS/HTML comments from web-player source (.ts, .svelte, .css),
per the project's no-comments convention.
Add scripts/check-tokens.mjs (npm run lint:tokens, wired into lint) as
a guardrail: fails if src/**/*.svelte reintroduces arbitrary values
outside the B.4 allowlist, factory rounded-*, h-N w-N pairs (should
be size-N), quarter-step fractional spacing, or literal hex/rgba()
colors.
Cleans up the repo until the script reports 0 violations: ~40 h-N w-N
pairs -> size-N (mostly icons, via a one-off script), ~26 quarter-step
fractional spacings rounded to the nearest integer (delta <=1px, e.g.
py-2.75->py-3, gap-1.75->gap-2), stray factory radii (rounded-sm/lg,
bare rounded) migrated to the matching tokens, and two
tracking-[-0.02em] that matched the existing tracking-display token
exactly but had never been migrated. Removes the
stroke-thin/regular/bold recipes (added earlier, zero consumers) —
decision made, not blocking: strokeWidth={n} (13 values, across
almost every component) isn't migrated to CSS because there's no way
to verify without Docker that lucide-svelte honors stroke-width by
class instead of the prop, and a regression there would be silent
across the whole app; the script documents this gap explicitly
instead of pretending full coverage.
Orphan detector: 0 components with no importers. Removes
docs/design-system-refactor-plan.md (earlier plan, superseded by
frontend-refactor-plan.md). web-player/CLAUDE.md gains a "Styles"
section with the B.4 allowlist and the base components to reuse.
Recalculates the baseline table (A.4) and logs it: 61->47 components,
44->11 arbitrary values (the remaining 11 on the allowlist), 13->8
distinct radii, 0 lint:tokens violations.
Verified with npm run check (0 errors), npm run lint:tokens (0
violations) and npm run build. Smoke-tested /login with no console
errors attributable to this change.
This pass also tackles the task deferred earlier: removing
tracks.ts's TrackTarget wrapper and its 8 accessors
(targetId/targetTitle/targetArtist/targetOwnerUserId/targetExplicit/
targetListensCount/isTargetCurrent/queueItemForTarget/
targetForQueueItem). Explore, Home, Mix, menus.svelte.ts and mixes.ts
now use track objects directly, with
toQueueItem/trackFromQueueItem (new, in player.svelte.ts) as the only
converters. Along the way, TypeScript revealed that isExplicit
doesn't exist in the real /tracks schema: Explore's "explicit" badge
was already a no-op before this refactor (always undefined ??
false); the binding is removed instead of pretending support that
never existed.
Explore: GENRE_INFO/genreTiles are extracted to src/lib/genres.ts;
the search logic
(showGroup/capped/matchPlaylists/searchCounts/searchChips/findTopResult)
is extracted to src/lib/search.ts as pure functions parameterized by
filter/query instead of closing over component state. "Top result"
unifies the duplicated button/link with <svelte:element>. Decision
made without blocking: the 4 group blocks (Tracks/Albums/Users/
Playlists) are not collapsed into a single {#each groups} — each
group has genuinely different snippets, context menu and ListRow
props, and a generic config would have added more indirection than it
removed.
Home: uses Page instead of a hand-rolled page-x; the 3 stats become
an {#each}; the hero/spotlight backgrounds (inline style= with CSS
variables) become the hero-surface/spotlight-surface utilities, and
gap-11.5 is replaced by the --spacing-section token (2.875rem) via
gap-section.
Verified with npm run check (0 errors) and npm run build. /explore
without a session redirects to /login because +layout.server.ts
requires allowAnonymousListening from the backend (unreachable
without Docker this session) — pre-existing behavior, not a
regression. No authenticated visual QA; smoke-tested /login with no
console errors.
PlayerExtras is merged into PlayerBar, which now renders volume and
the queue button inline and uses the new Slider (a real range input
with thumb thickness controlled via CSS variables) for both the
progress bar and volume, replacing SeekBar and VolumeControl (grep
confirmed SeekBar's `compact` mode was dead code).
Sidebar and TabsBar migrate to appNavLinks(user, counts?), removing
the duplicated nav links between the two. Decision made without
blocking: Sidebar gains the gradient Logo mark it didn't have before,
for visual consistency with MobileHeader and the login screen.
PlayerDock and MobileHeader's z-index is fixed to z-(--z-sticky)
(they're fixed bars, not menus), and SegmentedControl's to
z-(--z-raised), per the plan's note B.1.5. glass-bar and
theme-transition are applied to PlayerBar (both variants), TabsBar,
Queue and Home (hero/spotlight), replacing duplicated
blur/transition styles.
Verified with npm run check (0 errors) and npm run build; smoke-tested
/login with no console errors. No authenticated visual QA (Docker
unavailable this session).
CoverForm.svelte is the shared skeleton (form + enhance + cover +
error message + buttons) for AlbumForm and PlaylistForm, which now
only keep their own fields via the `fields` snippet (next to the
cover) and `extra` (full width — not literally in the spec, but
needed: the album description and the playlist visibility picker
don't fit in the column next to the art).
ImageDropzone gains `size="hero"` (size-cover-hero + rounded-art-lg)
and an `onselect` callback; the upload-track page now uses this
component instead of reimplementing the cover picker by hand (wrapped
in {#key} so it can reset after publishing). Its empty-state icon
goes from 64px to icon-xl (36px), as the plan recommends.
SegmentedControl.svelte (generic, N options) replaces PlaylistForm's
Private/Public picker. Chip.svelte replaces Explore's filter chips
(selected/count) and the suggested-genre chips in "No results" (href,
outline style).
@utility field-control in layout.css unifies the repeated class
string shared by Input and Textarea.
Verified with npm run check/build; no authenticated visual QA (Docker
unavailable this session).
PageHeader absorbs CollectionHeader
(title/eyebrow/description/meta/cover?/actions?/align?): without a
cover it behaves like the old simple PageHeader (inline actions),
with a cover like the old CollectionHeader (actions below). It sets
its own bottom margin (mb-7/8); the mt-* each page used to add
afterward are removed. Migrates the 9 places that used either one
(albums, library, playlists, upload, Explore x2, liked, mix,
profile). CollectionHeader is removed.
AlbumHeader and PlaylistHeader are inlined into their pages and
removed. Decision (recommended, not blocking): PlaylistHeader gains
icons on Edit/Delete/Play, matching what AlbumHeader already had.
New ConfirmDialog replaces the two delete dialogs (album and
playlist), which had different typography. Modal now uses
.glass-panel (instead of duplicating the inline style=) and
z-(--z-modal); its eyebrow and title use the
text-eyebrow/text-display-3 recipes.
SectionHeading gains `count?` (a "N results" line, replacing 4 copies
in Explore) and `href?`/`linkLabel?` ("See all", replacing Home's).
Applies typography recipes (text-display-2/3, text-eyebrow,
text-body) and width tokens (max-w-prose-sm, max-w-hero) across
EmptyState, +error, "No results", and the Home/Queue/Sidebar hero and
stats. The "You haven't listened to anything yet" empty state
(duplicated twice on Home) becomes a local snippet.
Verified with npm run check/build; no authenticated visual QA (Docker
unavailable this session).
MediaCard now builds its own art (src/trackIds via Artwork
size="fill", or a custom art snippet) and its own play button
(onPlay, with a scrim and a glass PlayButton on hover) instead of
relying on each wrapper passing it a snippet. Replaces AlbumCard,
PlaylistCard and MixTile, which are removed along with
MediaGrid.svelte (their grids move to the
grid-cards/grid-cards-lg/grid-tiles/grid-wide @utility recipes).
format.ts gains albumMeta() and playlistMeta() (previously ad-hoc
calculations inside each card). Explore's genre tile moves from an
inline style= gradient to the genre-tile @utility recipe +
--tile-hue. Along the way, fixes the "follower/s" pluralization on
the user profile that had been missed.
Delta documented: the art's hover-zoom and the scrim+PlayButton
(previously exclusive to MixTile) become MediaCard behavior gated on
the onPlay prop — they'll apply to any future card that uses it, for
consistency.
Verified with npm run check/build; no authenticated visual QA (Docker
unavailable this session).
Add TrackList.svelte, replacing TrackTable+TrackRow+TrackIndexCell+
TrackMeta and the 6 track tables (library, liked, mix, album,
playlist, album's library picker): declarative columns
(added/uploaded/plays) with fixed widths as module constants, an
index with an EQ indicator for the active track (or an overlay on the
art when index={false}, for the library picker), and a rowAction that
builds the form+IconButton (removeTrack/addTrack/deleteTrack) or a
direct onclick (Like).
AlbumTrackTable, PlaylistTrackTable and LibraryPicker are inlined
into their pages and removed, along with the 4 components TrackList
replaces.
Decision (recommended by the plan itself, not blocking): remove the
"Album" column, which in all 3 tables that had it always showed "—"
(a placeholder with no real data).
Verified with npm run check/build; no authenticated visual QA (Docker
unavailable this session).
Admin