Issue #319: Infinite Play toggle had no backend wiring after the V2
migration — toggle only wrote to localStorage and the queue-end handler
ignored it, so playback stopped at the last track.
Reimplemented in V2 (no legacy wrappers):
- New `v2_create_infinite_radio` command that takes the last few played
track IDs, uses the most recent track's artist as the radio seed via
`RadioPoolBuilder::create_artist_radio` (Qobuz `/radio/artist`
endpoint), and returns 50 V2QueueTrack ready to append. Does NOT
mutate the queue or playback context — caller decides how to use them.
- `playbackPreferencesStore`: expose `'infinite'` in `AutoplayMode`
type and add `isInfinitePlayEnabled()` helper.
- `+page.svelte`:
- `handleToggleInfinitePlay` now syncs to backend via
`setAutoplayMode('infinite' | 'continue')` so the choice persists
in `playback_preferences.db` across restarts (was localStorage only).
- On startup load infinite play from cached preferences (was
localStorage).
- When `nextTrack()` returns null AND infinite play is on: collect
the current track + last 5 history tracks, call
`v2_create_infinite_radio`, append the result via
`v2_bulk_add_to_queue`, then advance — playback continues without
interruption.
i18n: add `toast.failedToSavePreference` to all 5 locales.
cc
Four accessibility-focused themes with WCAG AAA contrast ratios:
- WCAG Light/Dark: 7:1+ contrast for all text levels
- High Contrast: pure white on pure black, yellow accent (21:1)
- Colorblind: deuteranopia-safe palette (Wong colors), no red/green
Addresses discussion #313 (user with severe vision impairment).
cc
Add pointer-based drag scrolling to HorizontalScrollRow so users can
click-and-drag to scroll album rows, matching iPad/browser behavior.
Distinguishes drag from click using a 5px threshold to avoid
intercepting album selection. Cursor changes to grab/grabbing.
Addresses user report from discussion #313 where horizontal album
rows were not explorable without arrow buttons.
cc
Tighten the viewBox so the vinyl fills ~97% of the canvas (matching
Steam/Chrome/Papirus sizing). Add a #3968be stroke rim around the
vinyl edge so the icon remains visible on pure-black backgrounds.
Include a subtle radial inner-light from top-left for depth.
Also adds 512x512.png for Flathub compatibility.
Regenerate all PNG exports: Tauri standard sizes, Windows Store
squares, ICO, Android mipmaps, and static logos.
cc
pkg_postinst and pkg_postrm in the binary ebuild call
xdg_icon_cache_update and xdg_desktop_database_update, which
come from the xdg-utils eclass. Without `inherit xdg-utils`
those helpers are undefined and the hicolor icon cache never
refreshes on upgrade.
The source ebuild template (qbz.ebuild.template) already
inherits xdg, so this only affects users installing via
qbz-bin (the prebuilt binary path). The qbz-overlay has a
direct patch for v1.2.5; this ensures v1.2.6+ release-gentoo
runs emit correct ebuilds.
cc
Triaged the 8 open CodeQL alerts raised on the 1.2.5 merge commit.
Critical (1) — real, fixed:
* .github/workflows/release-updater-manifest.yml:29 — the resolve
step interpolated `${{ github.event.workflow_run.head_branch }}`
directly into a shell heredoc. `head_branch` on `workflow_run` is
attacker-controllable (any outside contributor can push a branch
whose name embeds a payload). Route the value through an env var
(HEAD_BRANCH) and validate the shape with a strict regex before
letting it flow into shell or $GITHUB_OUTPUT.
Warnings (2) — real, fixed:
* .github/workflows/build-qbzd.yml had no `permissions:` block, so
every job inherited the repo's default GITHUB_TOKEN scope. Add a
top-level `permissions: contents: read`; the existing `release`
job keeps its own `contents: write` override for publishing
artifacts.
Warnings (3) — real, low severity, mitigated:
* crates/qbzd/src/{daemon.rs,session.rs} logged `user_id` at info
level (CodeQL rust/cleartext-logging). Demoted the three lines to
`log::debug!` so default-level logs stay free of the numeric user
id, with a comment noting RUST_LOG=debug re-enables them for
diagnostics.
Warnings (2) — false positives, dismissed in GitHub:
* crates/qbzd/src/login.rs:103 — `println!` to stdout inside the
interactive `qbzd login` subcommand; login confirmation UX, not a
persistent log. (alert #63 dismissed, reason: false positive)
* crates/qbzd/src/qconnect.rs:540 — `session_uuid` is an ephemeral
per-connection identifier QConnect exchanges in cleartext itself;
not a credential or token. (alert #65 dismissed, reason: false
positive)
cc
The v1.2.5 tag push flagged three macOS-only compile errors on the
release-macos workflow (Linux, aarch64 Linux, flatpak still building
successfully). All three are Linux-only symbols pulled unconditionally
into files that are compiled on every target:
* src-tauri/src/tray.rs: imports PathBuf inside the
`cfg(not(target_os = "linux"))` branch without `use std::path::PathBuf`.
Add the import under the matching cfg so macOS / Windows see it.
* src-tauri/src/commands_v2/playback.rs: `cached_audio_incompatible_with_hw`
is defined only on Linux (ALSA). Callsites are already
`cfg(target_os = "linux")`, but the shared `use super::{...}` block
was not. Move that one symbol out into a Linux-only import.
* src-tauri/src/commands_v2/legacy_compat.rs: same story with
`v2_prepare_notification_icon_bytes` — only the Linux notification
path calls it, but the import was unconditional.
No runtime behaviour change; this only unblocks the release-macos
(and any future Apple-target) workflow so it can build.
cc
Rewrite the opening paragraph into a personal, first-person note from
the maintainer (including the Spanish sign-off), drop the sprint-
level bullets that only described iterations kept in the narrative
but don't matter to end users, and tidy a couple of typos.
cc
Missed in the earlier bump pass. Confirmed every other version pin
reads 1.2.5: src-tauri/Cargo.toml, src-tauri/Cargo.lock,
src-tauri/tauri.conf.json, package.json, package-lock.json,
snapcraft.yaml, packaging/flatpak/com.blitzfc.qbz.metainfo.xml
(top <release> entry); the only remaining "1.2.4" string in the tree
is the historic release entry in metainfo.xml, which is correct.
cc
Follows the same pattern as docs/release-1.2.4/:
* CHANGELOG_FULL.md — short-hash + subject for every non-merge commit
since v1.2.4 (216 commits)
* TAG_DETAILS.md — narrative changelog grouped by Accolades, Release
Watch, Labels, Visual identity, Desktop integration, Streaming /
audio / player, Updater, UI polish, Internal architecture, Packaging
qbzd work is noted but flagged as scaffolding (hidden from the release
narrative per the plan to ship it separately).
cc
Bump the SHA in every <screenshot> URL to 5f759514d3
(the screenshot refresh commit) so Flathub picks up the new captures.
Add three new <screenshot> entries on top of the existing three —
immersive line-bed, enriched artist page, Hi-Fi DAC setup wizard — and
re-promote the Audio settings capture (default-settings.png) that was
sitting unused in the screenshots folder. New ordering for the listing:
1. home-discover (default — first impression)
2. immersive-cover-flow
3. immersive-line-bed
4. enriched-artist-page
5. local-library
6. hifi-wizard
7. default-settings
cc
Replace the four existing Flathub screenshots with captures from the
1.2.5 build (new logo, new sidebar, refreshed login chrome) and add
three new ones that showcase features worth highlighting on the
Flathub listing:
* enriched-artist-page.png — artist view with editorial data
* hifi-wizard.png — DAC setup wizard
* immersive-line-bed.png — immersive line-bed playback mode
Metainfo manifest is updated in the next commit because Flathub freezes
screenshot URLs by commit SHA, so we need this commit's SHA to land
first.
cc
The collision check in positionPopup was reading the popup's measured
rect right after mount, but the popup grows after that runs (the
genre tree fills in asynchronously). The first measurement reported a
height much smaller than the final ~500px / ~700px max-heights from
CSS, so the "would I overflow the bottom?" branch evaluated false even
when the final popup didn't fit. Result: bottom of the popup got clipped
by the window edge (visible in the Discover -> Filter by genre flow on
short windows).
Use the CSS max-heights (530/630 already used for width) as the
collision target instead. If the down-placement still overflows, try
flipping above the trigger; if neither fits (very small window), clamp
to the bottom padding and let the popup's own internal scroll handle
the overflow.
cc
* src-tauri/Cargo.toml and Cargo.lock: 1.2.4 -> 1.2.5
* src-tauri/tauri.conf.json: 1.2.4 -> 1.2.5
* package.json (+ package-lock.json regenerated): 1.2.4 -> 1.2.5
* snapcraft.yaml: 1.2.4 -> 1.2.5
* packaging/flatpak/com.blitzfc.qbz.metainfo.xml: new <release> entry
dated 2026-04-15 covering the visual refresh (new logo + login
branding), Plasma/Klassy auto-detect for window controls,
opt-in match-system-chrome rounded corners, Editor's Picks
ribbon cleanup, the two sidebar menu positioning fixes shipped
alongside this bump, and the DAC wizard contrast fix.
Tag not applied per request — release notes are staged, ready for
the screenshots pass and manual tag afterward.
cc
Two bugs surfaced during 1.2.5 QA:
* The per-playlist right-click context menu opened at the cursor and
could fall below the bottom of the window when invoked on items near
the last of the sidebar. It now measures its own rect and flips above
the cursor if there's no room below, or clamps to the bottom padding
as a last resort. Same pattern already used for the main playlists
dropdown menu.
* The Sort submenu inside the sidebar's general playlists menu was
being clipped by its parent .dropdown-menu because the submenu lived
inside that menu's DOM subtree. Teleport it via a local `use:portal`
action so absolute positioning escapes any ancestor `overflow: hidden`.
Also fix a duplicated "License" label in the About dialog's Build Info
block: the second row meant to show the platform was using the license
key. Swap to about.buildInfo.platform (the key already existed in both
locales).
cc
The inset dark frame + inset highlight + outer drop shadow combo
didn't produce a convincing edge on WebKitGTK — rounded corners are
the only piece of "match system chrome" that actually works today, so
drop the rest and leave the CSS minimal.
Real native shadow/border around a CSD Tauri window requires hints
like `_GTK_FRAME_EXTENTS` (X11) or `zxdg-toplevel-decoration-v1`
(Wayland) that wry doesn't expose yet. That path stays as a Phase 3
TODO — either upstream wry/tao or an FFI to gtk::Window via
raw_window_handle.
cc
Per user request, put explicit window edges above the ambient drop
shadow. Restore clip-path for clean corner anti-aliasing (it doesn't
clip inset shadows, only outer paints), drop the body padding
workaround, and paint two inset bands on .app:
inset 0 0 0 1px rgba(0,0,0,0.75) — crisp dark frame, 1px
inset 0 0 0 2px rgba(255,255,255,0.06) — subtle inner highlight
Keep a low-weight outer drop shadow (0 8px 20px) so the window still
has a hint of lift, but the edges — not the shadow — define the
window boundary now. Works the same on any compositor because it's
all painted by the webview, not by KWin/Mutter.
cc
Phase 2 produced clean rounded corners but no visible shadow because
.app filled 100vw/100vh, leaving no pixels for box-shadow to paint
into — and the clip-path added for anti-alias was clipping the shadow
as well.
Reserve a 12px padding on body while match-chrome-transparent is active
and shrink .app by the matching 24px in width/height, so the outer
drop shadow has somewhere to render. Drop clip-path (it was consuming
the shadow); keep transform: translateZ(0) for GPU AA on the corners.
Two-layer shadow: a broad soft ambient (18/40 rgba(0 0 0 / 0.65)) plus
a tight 4/12 crisp shadow for edge definition without painting a hard
1px stroke that jaggies at the rounded joint.
cc
Remove the hard 1px outline that the "match system window chrome"
style was painting inside the radius: on WebKitGTK the outline
amplified anti-aliasing jaggies at the rounded corners because the
stroke has no blur. Replace with a softer two-layer drop shadow,
add clip-path: inset(... round ...) on top of border-radius (clip-path
gives noticeably cleaner AA than border-radius + overflow:hidden alone
in WebKitGTK), and promote the container to its own compositing layer
with translateZ(0) + backface-visibility:hidden so the GPU handles the
edge.
cc
Rounded-corner window chrome now actually works. The toggle in
Settings > Appearance persists to the window_settings SQLite store and
gates the Tauri window transparency decision at startup, so on next
launch the main window is rebuilt with `transparent: true` and the
border-radius plus edge outline apply without painting white corners.
Flow:
* Toggle in Settings > Appearance -> windowChromeStore saves to
localStorage and calls v2_set_match_system_window_chrome which
writes match_system_window_chrome to window_settings.db.
* Toast reminds the user to restart.
* On next launch, should_use_main_window_transparency() reads the
setting and returns true, Tauri builds the window transparent.
* Setup stores the decision in MAIN_WINDOW_TRANSPARENT. A new command
v2_main_window_is_transparent reports it back to the frontend.
* +page.svelte only applies .app.match-chrome when the window is
actually transparent AND the custom titlebar is visible, so
stale-state toggling never produces white wedges.
* CSS sets html/body transparent via a global class on <html>, then
.app.match-chrome.floating paints the rounded corners plus a
12/28 rgba drop shadow and a 1px darkening outline.
* Maximized state clears the radius automatically.
Env vars QBZ_FORCE_TRANSPARENT_WINDOWS / QBZ_FORCE_OPAQUE_WINDOWS still
take precedence over the DB-persisted toggle.
cc
Adds an opt-in "Match system window chrome" toggle in Settings >
Appearance. When enabled (and the custom title bar is active, i.e.
neither "Use system title bar" nor "Hide title bar"), QBZ rounds its
own corners to match the detected desktop's decoration radius and
paints a thin edge outline so the window reads as its own surface
against the desktop instead of a bare rectangle.
Phase 1 scope:
* New windowChromeStore (localStorage) holding the toggle and the
detected corner radius.
* desktop_theme module gains a best-effort window_corner_radius_px:
10 for Klassy, 6 for Breeze Plasma, 12 for GNOME/Adwaita, 8 fallback.
* +page.svelte subscribes to the detector; .app.match-chrome.floating
applies the radius via a CSS variable plus a 1px darkening outline
and a deeper shadow. Maximized state removes the radius.
* SettingsView: new toggle, disabled when hideTitleBar or
useSystemTitleBar is on; i18n for en/es.
External shadow that matches each DE's compositor remains a phase 2
item — it requires `transparent: true` on the Tauri window, which has
compositor stability trade-offs and must be restart-gated.
Also cleans up redundant AlbumCard ribbons in the Editor's Picks
TAB (distinct from the home tab that was fixed earlier): removes
ribbon={album.ribbon} from the `editorPicks` and `qobuzissimes`
sections inside the editor-picks tab block. Press Accolades and
Popular Albums keep their ribbons (they vary per card).
cc
qbzd is feature-complete on disk but not ready for public release yet.
Keep all code paths in place so it can be unhidden with a one-line
flip, but remove it from the user-facing surface:
* build-qbzd.yml: comment out the `push.tags` trigger so neither a
regular `v*` version tag nor a `qbzd-v*` tag auto-builds artifacts.
workflow_dispatch is preserved for manual testing.
* CastPicker.svelte: guard the QBZ Daemon tab, the manual-connect
section, and the three `v2_qbzd_*` Tauri invocations behind a
`QBZD_ENABLED` flag (currently false). Default activeProtocol
switches to 'chromecast' so the picker doesn't start on a hidden
tab.
* SettingsView: DAC Setup Wizard icon filter swapped — button icon
drops the `invert(1)` so it reads on light accent colors (e.g.
Catppuccin Lavender), and the tooltip copy of the same icon now
has `invert(1)` so it reads on the dark surface bg. Unrelated to
qbzd but bundled since both are pre-release polish.
cc
Adds a runtime-detected "Klassy (auto-detect)" preset for the custom
window controls that reads the user's Plasma decoration theme and
applies matching colors and shape. When Klassy itself is not installed
but Plasma is, the option re-labels to "Plasma (auto-detect)" and
still applies the desktop color scheme. Hidden on non-KDE / non-Linux.
Backend:
* New desktop_theme module parses kwinrc, kdeglobals, and klassyrc
without external deps. Exposes detect_desktop_theme Tauri command
returning DesktopThemeInfo: desktop id, isKlassy flag, WM active/
inactive background + foreground, accent, DecorationHover, and the
user's Klassy ButtonIconStyle / ButtonShape when present.
Frontend:
* windowControlsStore gains applyKlassyPreset(), a cached detection
helper, and mapKlassyShapeToQbz() that maps Klassy's button shapes
(FullHeightRectangle, FullHeightRoundedRectangle, Tab, Circle,
Square, IntegratedRoundedRectangle, ...) onto QBZ's shape set.
* Adds the full-height-rounded shape to ButtonShape to cover the
most common Klassy style that QBZ was missing; CSS matches Klassy
FullHeightRoundedRectangle visual (full-height tab with ~7px
radius and margin).
* SettingsView fires detection on mount, derives the preset list
reactively so the auto-detect option only appears when relevant,
and adds the new shape to the Style picker.
* i18n keys added for the new shape and both preset labels in en/es.
cc
Apply the new matte-vinyl QBZ logo (Flathub-compliant: square canvas,
74% footprint, no baked shadows/gloss) across every icon asset: Tauri
standard sizes, Windows Store squares, Android icons and mipmaps,
static/, README, and the tray.
Ship two tray variants (black and white glyphs) and pick the right
one at runtime by reading GNOME color-scheme, GTK settings, and KDE
kdeglobals. Defaults to light-theme glyph when nothing matches.
Login screen: logo 80 -> 175 px with a subtle var(--text-muted)
drop-shadow using color-mix() for alpha, plus QBZ and "Qobuz(TM)
Player" under the icon. Update app.tagline i18n accordingly.
cc
The package's `postInstall` wraps the installed `qbz` binary with
`LD_LIBRARY_PATH` so it can `dlopen(libayatana-appindicator3.so.1)` at
runtime. Inside `nix develop`, users run `target/debug/qbz` directly
without that wrapper, so the tray init panics at startup.
Add a `shellHook` that mirrors `postInstall`'s library path.
Same root cause as the previous LIBCLANG_PATH fix: `inputsFrom` makes
dependencies available at build time but does not set runtime/env vars
the way the installed wrapper does.
Refs #312.
cc
rodio deprecated `DeviceTrait::name()` in favour of `description()`
(name) / `id()` (stable identifier). The legacy fallback in
`preferred_name()` was unreachable on our target platforms anyway —
the only path that matters is the `description()` one, since that is
what `get_audio_devices()` persists in settings.
Skip the device if `description()` fails instead of falling back to
the deprecated API. Silences the build warning.
cc
- src-tauri: rand 0.10.0 → 0.10.1 via cargo update (alert #36)
- qbz-player: rand 0.8 → 0.9, `gen_range` → `random_range` (alert #38)
The underlying RUSTSEC issue only triggers for code that calls
`rand::rng()` alongside a custom logger — we don't do that anywhere,
so actual impact is nil. Bumping closes the alerts.
cc
`inputsFrom` does not propagate the package's `env.*` attributes into
the dev shell, so running `npm run tauri dev` inside `nix develop`
failed at `mupdf-sys(build)` with "Unable to find libclang".
Re-export LIBCLANG_PATH on the shell itself.
Closes#312.
cc
Level A of issue #307 was firing 'device no longer available' spuriously
because settings.output_device stores the ALSA-normalized form
(front:CARD=X,DEV=0 per normalize_device_id_to_stable) while cpal's
output_devices() enumeration returns different strings entirely —
DeviceTrait::description() (friendly) and DeviceTrait::name() both
produce values that don't match what the UI persisted. Comparing them
marked every configured device as 'missing' on normal playback.
Removes the pre-flight emit_missing_if_needed() calls from
v2_play_track and v2_resume_playback so the false-positive toast
stops firing. The init_device fallback still routes audio to default
on a truly missing device — we just no longer surface that to the UI,
which is net neutral for the user vs. wrong notifications.
Keeps audio_device_watch module, DeviceMissingThrottle state and the
frontend listener in place so Level B (emit from inside the audio
thread when the fallback actually triggers) can reuse the plumbing
without another refactor. Hardens the presence check with
empty-enumeration = inconclusive and description()-based matching for
whenever Level B actually wires up a call site that has a comparable
device name at hand.
cc
Root cause traced from user diagnostic: only one playback:state event
fired (position=0), then nothing. Backend's current_position() returned
0 forever because thread_state.duration was 0 — the PlayStreaming
handler at player/mod.rs:1849 stores duration verbatim from the
command parameter and current_position() at player/mod.rs:789 ends
with .min(duration), clamping to 0.
Parameter names also didn't line up. cmdPlayTrack was invoking
v2_play_track with qualityStr and lowQualityFallback, neither of
which matches Tauri's camelCase mapping of the Rust signature
(quality, forceLowestQuality, durationSecs). Quality and force-fallback
were being silently ignored; durationSecs was never sent at all.
The normal queue playback path in playbackService uses the correct
camelCase names and passes track.duration, which is why the seekbar
worked everywhere except the session-restore first play — that path
goes through togglePlayback → cmdPlayTrack.
Fix:
- Correct the parameter names in cmdPlayTrack to match the backend
(quality, forceLowestQuality, durationSecs).
- Add durationSecs as a first-class arg and pass it from
togglePlayback when resuming a restored track.
- Drop the temporary [Diag:*] console logs added while investigating.
cc
evaluateQconnectSessionPersistence now unconditionally returns
shouldPersist: true. Previously it refused to save the local session
whenever Qobuz Connect was in remote mode, which meant closing QBZ
with QConnect active discarded the current track + position. On next
launch the app fell back to whatever QConnect held, which is
playlist/release-level only.
QConnect-vs-local priority on restore is a separate concern and
belongs at the restore path, not here.
Tests updated.
cc
Level A of issue #307. When the configured output device has
disappeared from the host device list at the time of play/resume
(KVM switched away, USB sink unplugged, PipeWire sink removed…)
the backend emits an 'audio:device-missing' event and the frontend
shows a warning toast:
'Output device "XYZ" is no longer available. Using system default.'
The existing init_device fallback already routes audio to the system
default, so this is purely informational — the user now understands
why audio came out of a different sink instead of staring at a silent
progress bar.
Does NOT touch the sacred audio init path (pipewire_backend.rs,
init_device, audio_settings.rs). New module audio_device_watch.rs
adds a pure presence check, a new-state throttle so we don't emit on
every resume once the user's been notified, and a v2 command
v2_check_audio_device_presence so the frontend can poll on demand.
Ready for Level B (background watchdog) — the same helpers plug into
a timer without further refactor; Level B only needs to add a tokio
task that calls emit_missing_if_needed() every N seconds while
playback is active.
cc
Referenced by AwardView hero / Other Awards mini-card / AlbumDetailView
sidebar award avatar in commit d719fb67 but the asset itself was never
tracked. Ships the SVG so the placeholder renders after a clean clone.
cc
appendPageReleases already skipped releases whose id already lived in
artist.albums/epsSingles/liveAlbums/others, but convertPageArtist
(the first /artist/page response) pushed everything blindly. When
Qobuz returned the same release in multiple groups (e.g. both the
main category and the 'download' distribution bucket) we ended up
with side-by-side duplicates on the discography grid — reported by
user on Holly Humberstone's 'Cruel World' entry.
Adds a seenIds Set scoped to the whole response so a release is
assigned to exactly one category on first encounter and silently
dropped on subsequent references. totalAlbums is now counted per
unique accepted item instead of raw group size, so the header count
matches the rendered grid.
cc
The compact QualityBadge in Popular Tracks rendered a gold-gradient
chip that clashed with the rest of the track row. Homologates to the
same look AlbumDetailView uses in its TrackRow: plain 12px grey text
('24-Bit/96 kHz' / 'CD Quality'), centered in an 80px cell, no
coloured background.
Uses formatQuality() with bit-depth > 16 as the hires signal so the
output string is identical to what AlbumDetailView shows for the
same track.
cc
Uses /static/laurels.svg as the default award artwork everywhere a
placeholder was rendering the lucide Award (trophy) icon:
- AwardView hero circle (140px, 68%)
- AwardView 'Other awards' mini-card (120px, 56%)
- AlbumDetailView sidebar award avatar (44px, 58%)
Laurel is rendered as an <img> filtered white (brightness(0) invert(1))
over the existing gold gradient. Awards that ship with their own
image/magazine cover keep theirs — the swap only affects the default
fallback.
cc
- AwardView stayed on the same activeView when switching between
awards via the 'Other awards' carousel, so onMount didn't re-fire
and the old award's data persisted. Replaces the onMount trigger
with a keyed on awardId: resets loading flags and calls
loadHero/loadAlbums/loadOtherAwards whenever the prop changes.
- Essential Discography albums all carry the Ideal Discography award
by definition — the ribbon is redundant on every card. Drops the
ribbon prop from the two AlbumCard call-sites for that rail.
Ribbons still appear on New Releases, Press Accolades, Most
Streamed, Qobuzissimes, Album of the Week, Editor's Picks and
Forgotten Favorites.
cc
1. Press ribbon restyled to a solid gold gradient (#f5c042 → #d49511)
with dark (#1f1407) uppercase text and a subtle white text-shadow.
Drops the red border-left. Qobuzissime / Album of the Week keep
their dark-background treatment with coloured left accent.
2. AwardView now loads /award/explore on mount (first 30 entries,
current award filtered out) and renders them as a horizontal
'Other awards' carousel below the album grid — same layout as
LabelView's 'More labels'. Each card shows the award laurel
(or magazine cover when present) + name + magazine name; clicking
routes into that award's AwardView.
i18n key award.otherAwards added for all five locales.
cc
Root cause of 'Discoteca Ideal Qobuz not clickable': /album/get
returns awards as LegacyAwardDto (per a20/c.java + LegacyAlbumDto
in the Android decompilation) whose id field is named awardId, not
id. Our AlbumAward struct only looked for 'id', so every award from
/album/get deserialized with id=None and the click fell back to the
catalog resolver — which was itself incomplete.
Adds serde aliases 'awardId' / 'award_id' for the id field and
'awardedAt' for awarded_at. Applies to both qbz-models (V2) and the
src-tauri legacy mirror.
Endpoint coverage:
- /discover/index → {id, name, awarded_at} (primary match)
- /album/get → {awardId, name, awardedAt, …} (alias match)
- /artist/page → {id, name, awarded_at} (primary match)
cc
Previous loop short-circuited when a page returned fewer than the
requested limit, but Qobuz may return a partial page with has_more=true.
Only stop when items is empty, has_more is explicitly false, or the
page repeats ids already seen (server ignoring offset).
Also bumps the sessionStorage key (v2) so prior truncated caches are
discarded, and logs every page (offset, returned count, has_more) plus
a summary (distinct ids, names, total items, pages) for debugging.
cc
- Console-dumps album and album.awards when AlbumView loads so we can
see exactly what Qobuz returns for a given album (user reported
Discoteca Ideal Qobuz still not resolving even after /award/explore
cached 15 entries).
- Adds rememberAward / rememberAwards / rememberAwardsFromAlbums to
the catalog store plus sessionStorage persistence. Wiring these
into the data paths comes next once we know what shape /album/get
actually returns for awards.
cc
User pushed back on the hardcoded list of Qobuz editorial award names
across locales. Removed the seed map entirely — the catalog is now
100% what Qobuz returns from /award/explore in the user's current
session locale, so id resolution always matches whatever labels the
API is currently emitting.
Kept:
- Cache-first lookup (hit the in-memory map before touching the API).
- Diacritic-insensitive + whitespace-agnostic normalization so accent
or case differences don't miss a real match.
- Single fetch per session, paginated.
Added a console.log summary after the initial fetch and a warn when a
name misses the catalog so future 'X award doesn't open' reports are
debuggable without guessing.
cc
'Discoteca Ideal Qobuz' was not resolving because /award/explore
either doesn't list editorial Qobuz awards or returns them under a
different locale's label. Adds a hardcoded seed of the three
stable-id editorial awards (88 Qobuzissime, 151 Album of the Week,
70 Ideal Discography) with their Spanish/English/German/French/
Portuguese aliases so they resolve instantly regardless of the user's
Qobuz locale.
Name normalization now strips diacritics (NFD + combining chars) and
collapses whitespace so 'Discoteca Ideal Qobuz',
'Qobuz Ideal Discography' and 'Discothèque idéale Qobuz' all map to
the same id.
Seed runs first on every lookup (sync). /award/explore only kicks in
for press awards the seed doesn't know about, once per session.
cc
AwardView now shows a single grid of the first 20 albums with a See
all link in the section header (visible only when there are more
albums behind it). Grid layout stays — no horizontal carousel.
Clicking See all navigates to a new 'award-albums' view
(AwardAlbumsView) with a full paginated list: header with kicker +
title, top-right search, virtualized grid with load-more. Same
structure as LabelReleasesView for visual consistency.
cc
/album/get's AwardDto has id: String? — Qobuz sometimes returns an
award entry with only a name and no id. Rather than disabling the
click or toasting, we look up the id in the Qobuz award catalog.
- Adds v2_get_award_explore (paginated /award/explore) all the way
through qbz-qobuz → core → bridge → command → invoke_handler.
- New frontend awardCatalogStore caches the full (id, name) list
lazily, one fetch per session, resolved on first click that needs
it. Further clicks are instant.
- handleAwardClick goes straight to navigation when id is present,
falls back to resolveAwardIdByName otherwise. Toast only fires if
the name isn't in the catalog either (genuinely unknown award).
cc
Previously disabled={!onAwardClick || !award.id} grayed out the award
card whenever /album/get returned an award entry without an id. User
reported the award was 'no longer clickable' on albums opened from the
artist page. Now the button always accepts clicks and the handler
shows a toast when no id is available instead of silently doing
nothing — keeps the affordance honest without sending 'null' to
/award/page.
cc