At the ~1000px window minimum, the toolbar's right-aligned sort group
ran out of row width and overflowed leftward past the panel edge; egui
then advances the vertical cursor from the overflowed rect, so every
row below (view toggle, the whole list) shifted left and was clipped
under the sidebar.
The sort chips now right-align only when a font-metrics measurement
says they fit, and otherwise drop to their own wrapped row. The top
bar likewise wraps instead of pushing nav buttons off-screen, with the
right-aligned status label width-guarded the same way (truncating
fallback).
Verified via live GUI screenshots at 1000x700 and 1400x800; the wide
layout is unchanged.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Render List/Card/Grid through ScrollArea::show_rows so only the visible
row range is laid out per frame. Fixed-size cells (allocate_ui_with_layout
+ set_min_size) enforce a uniform lattice; titles truncate to one line with
the full title on hover. List separators are painted hlines inside the
cell; Grid is a manual horizontal run of fixed cells per lattice row with
cols computed from available width before show_rows.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Executes docs/superpowers/specs/2026-07-07-runtime-performance-pass-design.md:
- database.rs: every pooled connection now runs WAL + synchronous=NORMAL
+ busy_timeout=5000 + foreign_keys=ON via with_init (in-memory pools
get the latter two). Fixes the SQLITE_BUSY failure mode under the
parallel scanner and makes FK cascades work on every connection, not
just whichever one last ran the old one-shot pragma in delete_folder.
- database.rs: prepare_cached for the per-video info_cache queries; new
info_cache_put_many bulk upsert (single transaction) replaces the
per-row autocommitted info_cache_put, with a round-trip unit test.
- library.rs: enrich_with_cache collects cache misses and flushes them
once per channel; title sort uses sort_by_cached_key.
- app.rs: Title/ChannelAsc sorts use sort_by_cached_key instead of
allocating two lowercased strings per comparison.
- Cargo.toml: panic=abort in release (crash-hook still fires; Cargo
forces unwind for test harnesses). PACKAGING.md documents the local
RUSTFLAGS target-cpu=native opt-in; repo stays portable.
- .gitignore: catacomb.db-wal / catacomb.db-shm sidecars.
Verified: 146 tests pass (integration suite exercises the pragmas
end-to-end), x86_64-pc-windows-gnu check green, live smoke shows
journal_mode=wal + batched cache writes landing correct rows.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Mobile fixes (all in the embedded SPA):
- Move the mobile media queries to the bottom of the stylesheet — the
cinematheque design layer was appended after them, so its base rules
(header padding, video heights) silently overrode the phone layout by
source order. A comment now pins the ordering invariant.
- The design layer's uiDrop entry animation (fill-mode:both) permanently
overrode the sidebar's translateX(-100%), leaving the drawer stuck open
over the content on phones; the mobile block now disables it.
- Lower the single-column breakpoint 380px -> 350px: it was catching
360/375px phones and blowing every card up to full width.
- viewport-fit=cover so the env(safe-area-inset-*) paddings actually
resolve on iPhones; notch-safe padding on header/content/modals.
- 16px inputs on small screens (kills iOS Safari's focus auto-zoom),
same for the login page's password field.
- Mobile rules for the design-layer components that had none: full-height
command-palette sheet, clamped stats numerals, coarse-pointer touch
targets for the custom player (taller scrub track, always-visible
thumb, no volume slider).
- CSP font-src now allows data: — the embedded base64 woff2 fonts were
being blocked and silently fell back to system faces.
PWA:
- manifest.webmanifest, minimal service worker, and a new Catacomb arch
icon set (SVG source + rendered PNGs incl. maskable + apple-touch),
all include_str!/include_bytes!-embedded like the HTML.
- sw.js only intercepts GET navigations to "/" (network-first, cached
offline fallback) and the static assets; /api, /ws, /files,
/music-files, /feed are never touched, and it's served no-store so
binary upgrades keep propagating.
- Routes + auth_middleware allowlist so the browser can fetch the
statics pre-login; theme-color meta tracks the active theme's panel.
- tests/api.rs covers the new endpoints incl. the ungated-when-password
invariant.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Replace the old search modal with a keyboard-driven command palette: blurred
backdrop, centered input, results grouped by channel with highlighted FTS
snippets (char(2)/char(3) → <mark>), ↑↓ navigation, Enter to open, Esc to
close, recent searches + quick actions (Downloads/Stats/Health/Shortcuts) in
localStorage. Debounced queries to /api/search with a sequence guard so stale
responses don't clobber newer ones. Upload date is looked up client-side from
the loaded library (SearchHit has no date field). Opened by the 🔍 header
button and the `f` hotkey.
Verified: node --check passes, release builds clean, and a headless harness
screenshot confirms grouped results, highlighted matches, and selection.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Prose + commands updated to the new name: brand reads "Catacomb", while
binary/path/package references are the lowercase `catacomb` (commands, deb/rpm
names, ~/.local/share/catacomb, catacomb.db, catacomb.desktop). Codeberg repo
URLs left pointing at the existing repo. Also folds in the pending doc edits
from earlier in the session.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The login page was the last surface still on the old flat style. Give it the
same identity as the rest of the UI: serif wordmark with the glowing crimson
"recording" dot, an italic "your private archive" tagline, accent aurora +
film-grain atmosphere on a gradient card, a themed focus ring and accent
button, and a soft entrance. Login JS is unchanged; honours reduced-motion.
Standalone page, so the display font is a system serif stack (no CDN, no
duplicating the SPA's embedded woff2).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replace the browser's native <video controls> (used for direct /files/
playback) with one custom player for both direct and transcode streams, so
the chrome matches the reskin and gains features it never had. All seeking
still routes through playerSeek/effTime, so the transcode reload-at-offset
path is unchanged.
- Custom scrubber: accent played fill + buffered fill + chapter tick markers
+ a hover time tooltip; click/drag to seek (pointer events, works in both
modes — commits the seek on release so transcode only reloads once).
- Playback speed popover (0.5–2×) and persistent volume + speed + captions
across videos (localStorage).
- Captions (CC) toggle for overlaid <track> subtitles (off by default; the
searchable 📄 transcript pane is unchanged), PiP, fullscreen.
- Auto-hiding controls + cursor after idle while playing; a center play/pause
flash; gradient scrim.
- Expanded keyboard: space/k, j/l ±10, ←/→ ±5, ↑/↓ volume, m, c, p, f,
< > speed, 0–9 seek-to-percent — now active for direct videos too (they
previously relied on native controls).
Verified: node --check passes, release builds clean, live 8081 serves it, and
a headless screenshot of the control chrome (extracted CSS) renders correctly.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Sessions were an in-memory HashMap seeded empty at startup, so every restart
or upgrade invalidated every browser's cookie (the cause of the recent
"settings save → error" report: a restart out from under a logged-in tab).
Mirror sessions to a new `sessions(token, issued_at)` table:
- store issued-at as a UNIX timestamp (the map switches from Instant→u64 so it
can round-trip the DB; is_authed prunes by wall-clock age vs SESSION_TTL),
- insert on login, delete on logout, clear on password change,
- rehydrate the map at startup via load_sessions(), which also prunes rows
past the TTL. DB writes are best-effort — a failure just means that token
won't survive a restart, not that login breaks.
Verified end-to-end: set a password, log in, restart the server against the
same DB, and the original cookie still authenticates (200) while a cookieless
request is still 401. 128 unit + 11 integration tests green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Sessions are in-memory, so a server restart (or a session TTL expiry) leaves
a still-open SPA tab holding a stale cookie: cached views keep rendering, but
every mutating call 401s and surfaced only as an "authentication required"
toast (e.g. saving Settings) — a confusing dead end.
Make api() treat 401 as "session gone": flash a notice and location.reload(),
which the server answers with the login page. No reload loop (the login page
isn't the SPA), and authed sessions are unaffected.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reskin the Library-health modal into a systems-diagnostics readout, sibling
to the stats observatory and sharing its sx-* section language. Presentation
only — every id/class/handler the logic depends on (dup-chk, sim-chk, at-chk,
dedup-area, autotag-area, the async polling) is untouched.
- A pulsing health verdict ("Healthy" / "N items to review") whose count and
status colour update live as the async subsystems resolve.
- Four status tiles (Duplicate IDs, Missing assets, Similar content, Unfiled
groups) with health-coloured dots + top rules; sim/at start pending and are
filled by their render functions.
- Sections become instrument panels: duplicate/similar groups as cards with
KEEP/REMOVE pills, missing-asset rows, a gradient dedup progress bar, and
auto-tag groups with confidence dots. Modal widened to 920px.
- Status colour flows through --mx-c off a data-status attribute, scoped to
.mx-body so it can't leak onto other nodes; pulse honours reduced-motion.
Verified: JS passes node --check, release builds clean, live 8081 serves it,
and a headless screenshot of a harness running the extracted CSS+JS against a
mock report confirms the verdict math, tiles, pills and panels all render.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Rebuild the Library statistics modal from plain tiles + flat bars + tables
into a real data dashboard, driven by the same /api/stats payload and the
theme variables (so all 10 themes inherit it).
- Metric cards with giant serif numbers that count up from zero on open
(size/runtime spin up through their units), an accent top-rule, and a
glowing hero tile.
- A self-drawing SVG area chart for downloads-per-week: Catmull-Rom-smoothed
line that animates in via stroke-dashoffset, a gradient area fill,
gridlines, a date axis, and hover dots with tooltips.
- A growing-column histogram for videos-by-upload-year with value labels.
- A ranked channel leaderboard with animated meter fills, gold/silver/bronze
ranks, and a Size/Count segmented toggle that re-animates on switch.
- Reveal animations gate behind .sx-go (added post-layout) so they play once
per open, not on re-render; honours prefers-reduced-motion; modal widened
to 1040px.
Verified: JS passes node --check, release builds clean, the live 8081 server
serves it, and a headless screenshot of a harness running the *extracted*
dashboard CSS+JS against a mock payload confirms the chart math, count-up,
histogram and leaderboard all render correctly.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A presentational redesign of the web SPA, appended as a design layer so it
overrides by source order while every colour still flows through the existing
7 per-theme CSS variables — so all 10 themes inherit it and no element
id/class the JS relies on changes.
- Typography: embed Instrument Serif (display) + Hanken Grotesk (body) as
base64 woff2 (SIL OFL) — offline-safe, no CDN, no privacy leak. Serif
wordmark/headers/empty-state; clean grotesque for UI.
- Identity: a glowing crimson "recording" dot before the wordmark; a soft
accent aurora + fine SVG film-grain fixed behind content (negative z, so
it never sits under text).
- Depth & motion: translucent blurred masthead with a hairline accent
underglow; cinematic card hover (lift + accent ring + thumbnail zoom +
shadow); one-time staggered shell reveal; themed focus rings; thin themed
scrollbars; restrained button hover (no full crimson flood). Honours
prefers-reduced-motion.
Verified: JS still passes node --check, release builds clean, and a headless
chromium screenshot of the live 8081 server confirms it renders (fonts +
atmosphere + masthead) without breaking the existing layout.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The tree already links a working x86_64-pc-windows-gnu exe; turn that into
a shipped artifact:
- scripts/package.sh gains a `win` target (and `all` picks it up when the
mingw toolchain is present) that cross-compiles and zips the exe +
LICENSE + a README listing the yt-dlp/ffmpeg/mpv PATH deps.
- Forgejo release workflow installs mingw-w64 + zip + the rust target, so
a tag push produces yt-offline-<ver>-x86_64-windows.zip from the same
Linux container — no Windows runner needed.
- main.rs: attach_windows_console() (cfg(windows), windows-sys
AttachConsole) reattaches a release build to the launching terminal so
--web/CLI output is visible, while a double-click stays windowless
(release sets windows_subsystem = "windows").
- ROADMAP 3.1 updated: Windows ships; macOS deferred (needs a Mac runner).
Native Linux build + 128 unit / 11 integration tests still green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add "download date" sorting (file mtime, distinct from upload date) plus a
broader set of options. Web select is now grouped via <optgroup>; desktop
gains SortMode::DownloadDesc/DownloadAsc/ChannelAsc with matching arms and
toolbar entries.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The startup channel scan used available_parallelism() (every core), which
could briefly peg the whole box on a cold/large library. Cap it at cores-1
so the machine stays responsive; the scan is disk-I/O-bound and runs as a
background task, so one fewer thread costs little. Mirrors the headroom the
fingerprint pass and thumbnail pool already leave.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Patch progress/speed in place on each tick instead of rebuilding the whole
modal body, so open per-job logs no longer collapse and the bar doesn't
flicker while downloading. Full rebuild only on a structural change
(job added/removed, state or retryability changed, queue length changed).
- "↻ Retry all failed" header button that re-issues every retryable failed
job (re-snapshots between retries since indices shift).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Download management UI (web + desktop):
- Cancel a running job (SIGKILL, marked "cancelled" so it isn't auto-retried
and shows a distinct state rather than a misleading error class)
- Cancel a still-queued job before it starts
- Manually retry a failed/cancelled job (fresh auto-retry budget)
- Live speed · ETA · % on running jobs, parsed from the yt-dlp progress line
- Expandable full per-job log fetched on demand
New endpoints: POST /api/jobs/:idx/{cancel,retry}, GET /api/jobs/:idx/log,
DELETE /api/queued/:idx.
Perceptual-dedup off-switch:
- backup.dedup_enabled (default true) hard-disables the "similar content"
scan in both UIs; the web scan endpoint returns {disabled:true} and the
desktop scan no-ops with a note. Wired through all five settings touchpoints.
Docs: README seeking note + ROADMAP recently-shipped updates for the prior
seekable-playback / federation / auto-tagging work.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Transcoded videos stream through a live ffmpeg pipe with no byte ranges, so
the browser couldn't seek/rewind them at all. Fix it end to end:
- get_transcode accepts ?start=<secs> and passes -ss before -i (fast keyframe
seek), rebasing timestamps to 0 from there.
- The player gets a custom control bar for transcode sources: play/pause, a
scrubber driven by the known duration, current/total time, ±10s, mute +
volume, and fullscreen, plus space / arrows / j / l / f keyboard shortcuts.
Seeking re-requests the stream at the new offset; effective time is
offset + element.currentTime. Every seek path (scrubber, chapters,
transcript click, resume, save-position, chapter/transcript highlight) now
routes through one offset-aware playerSeek()/effTime().
- Direct /files/ videos keep native <video controls> (already seekable).
Verified server-side: start=0 vs start=120 yield different streams.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Configure peers as [[remote]] entries (name, url, optional password) and
browse their libraries from this instance, read-only.
- remote.rs: RemoteClient (blocking reqwest + rustls + cookie jar). Fetches
the peer's /api/library, logging in first if the peer has a password and
retrying on a 401. Media is not proxied: the library JSON's media URLs
(/files/, /music-files/, /api/transcode/, /api/sub-vtt/) are rewritten to
absolute peer URLs with the peer's read-only feed token appended, so video
streams straight from the peer to the browser/mpv while only the small JSON
passes through us.
- web.rs: auth_middleware now also accepts the feed token for GET
/api/transcode/ and /api/sub-vtt/ (read-only media, same class as /files/),
so a transcoded peer streams remotely. New GET /api/remotes and
/api/remotes/:id/library (the RemoteClient runs under spawn_blocking).
- web UI: a 🌐 Remotes sidebar switcher; remote mode is read-only (hides the
download bar + per-card mutating actions, keeps Play).
- desktop: a 🌐 Remotes screen that fetches a peer's library off-thread and
plays via mpv on the tokenized URL.
- config.rs: [[remote]] section list; reqwest dep added (rustls/blocking/
cookies, no system OpenSSL).
Verified end-to-end against a second local instance, including a transcoded
stream returning HTTP 200 via the token. ROADMAP 3.5 + CLAUDE.md updated.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
`cargo check --target x86_64-pc-windows-gnu` is now green (only the
upstream egui f32:From<f64> warnings). The Linux-only desktop deps are
target-gated so the rest of the stack — which already cross-compiles —
can build off Linux:
- Cargo.toml: ksni and rfd's xdg-portal backend move to
[target.'cfg(target_os = "linux")'.dependencies]; non-Linux gets rfd
with its native Win32/AppKit backend (default features).
- tray.rs: cfg-split. The ksni/SNI implementation stays Linux-only; other
OSes get a no-op `start() -> None`, i.e. windowed-only (identical to the
Linux no-SNI-host path). TrayEvent/TrayHandle stay cross-platform.
- plex.rs: add a Windows `symlink_file` arm to make_symlink (also fixes an
unused-variable warning on non-unix).
The rest was already portable (disk_space/statvfs, mpv IPC UnixStream,
chmod/PermissionsExt guards all cfg(unix)). macOS is Unix so those apply
there unchanged. Still not a *shipped* binary: needs a real per-OS tray
backend and a linking CI matrix. ROADMAP 3.1 + CLAUDE.md updated.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
New autotag.rs classifies each unfiled channel from already-scanned
metadata — source platform + median video duration + upload cadence —
into a suggested folder group (Music / Shorts / Long-form & Podcasts /
Streams & VODs), with a confidence and a human-readable reason. Mid-length
YouTube is left unsuggested rather than guessed at; channels already in a
folder are skipped. Pure arithmetic over the in-memory library, computed
on demand (no background job).
Surfaced in both Maintenance views:
- web: GET /api/autotag/suggest + POST /api/autotag/apply (create/reuse the
named folder and assign), rendered with per-channel checkboxes and an
"Apply -> group" button that re-analyzes after applying.
- desktop: a section listing each group with a "Move all -> group" button.
Apply mirrors post_assign_folder: DB write + in-memory library update +
ETag bump. Reuses an existing folder of the same name (case-insensitive).
Unit tests cover the classifier; verified the apply/revert round-trip live.
ROADMAP 3.4 marked DONE.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Several related UI and startup fixes from one session:
- Web grid: channel and music views rendered in a single column because
their content was wrapped in one nested div that became a single cell of
the outer CSS grid. Span the wrappers (and .empty) across all columns
with grid-column:1/-1.
- Subtitles: route every track through /api/sub-vtt and strip per-cue
position/alignment settings (align:start position:0% etc.) from both SRT
and VTT, so auto-generated captions render centered instead of left-
aligned. New strip_cue_settings/normalize_vtt + sub_vtt_url helpers.
- Comment downloads: add a global backup.fetch_comments toggle (the
per-channel option already existed; it's now a tri-state override that
defers to the global). Full five-touchpoint wiring across config,
download_options, the downloader's apply_comments resolver, both UIs'
global + per-channel controls, and downloader seeding.
- Desktop startup: run the initial library scan + search-index sync on a
background thread (mirrors the web server's deferred bind) so the window
appears immediately instead of blocking on a cold-cache scan; update()
swaps the library in when it lands.
- Perceptual dedup: cap worker threads via fingerprint::default_workers()
(~half the cores, always leaving at least one free) instead of using
every core, so the hashing pass doesn't starve the UI.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
serve() ran a full library scan + search-index rebuild before binding the
HTTP listener. On a large library with a cold cache (e.g. on a slow mount)
that scan is disk-bound for minutes, during which nothing listens on the
port and the server looks dead on the network (notably over Tailscale).
Build WebState with an empty library, bind + print the URL immediately,
then run the scan in a spawn_blocking task that swaps the result in and
bumps the library version (same pattern as post_rescan). The server is now
reachable in ~1s; content appears once the first scan lands and clients
pick it up via the bumped version/ETag.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The desktop forces the wgpu/Vulkan renderer because glow/OpenGL crashes
on NVIDIA + Wayland maximize. But on some systems wgpu creates the window
and never presents a frame, so nothing appears. Keep wgpu the default but
compile glow back in and let it be selected via `--renderer glow` or
`YT_OFFLINE_RENDERER=glow` for those machines.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Subscribe to the whole library or a single channel in any podcast/media
app.
- src/feed.rs: pure RSS 2.0 + iTunes-namespace renderer with absolute
enclosure URLs, MIME-by-extension, HH:MM:SS durations, and a
chrono-free RFC-2822 date (civil-date algorithm). Unit-tested
(mime / escape / duration / date / render).
- web.rs: GET /feed.xml (whole library, newest first, capped at 300) and
GET /feed/:platform/:handle (one channel); /api/feed-info returns the
token for the UI. Enclosure + thumbnail URLs are absolute (derived from
Host + X-Forwarded-Proto) and carry the feed token.
- Auth: a stable read-only `feed_token` (persisted setting) lets a
tokenized `?token=` GET reach /feed*, /files, and /music-files even when
a password is set — podcast clients can't do the cookie login. Scoped to
reads of feeds + media; never /api mutations.
- Web UI: a "📡 Podcast feed" section in Settings (library URL + Copy) and
a "📡 Feed URL" button in each channel's options dialog.
Integration tests cover the served RSS (enclosure path, MIME, pubDate),
the channel feed + 404, and the full token gate (401 without token once a
password is set, 200 with it, /files reachable with it). 127 tests pass.
Desktop URL display follows.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The 🔍 library search now matches spoken words, not just titles /
channels / descriptions.
- database.rs: add a `transcript` column to the `video_search` FTS5 index.
Fresh DBs get it directly; existing ones are migrated (FTS5 has no ADD
COLUMN, so the table is recreated and search_meta cleared to force a
one-time reindex). `sync_search_index` reads each video's first subtitle
(mtime-gated like the description), flattens it via the shared `vtt`
parser (`transcript_text`, capped at 256 KB), and indexes it. Search
snippets now use column -1 so the excerpt comes from whichever column
matched (description or transcript).
- library.rs: `build_search_entries` passes the first subtitle path.
- Both UIs: search copy updated to mention transcripts.
Tests: a unit test (a word only in the .vtt is found), a migration test
(seed an old 5-column index + stale meta, open, confirm the recreated
index indexes transcript text), and the search integration test now seeds
a subtitle and asserts a spoken-only word hits. Also bumped the test
server's startup budget so the ffmpeg-heavy dedup test can't starve a
sibling server into a flaky timeout. 120 tests pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Extracts the dedup pipeline (mtime-gate -> parallel fingerprint -> prune
-> group) into `fingerprint::rebuild_and_group`, returning groups of file
paths. Both front-ends now call it and map paths to their own review
rows — web's run_dedup is refactored onto it (no behaviour change; the
integration test still passes).
Desktop: the 🩺 Maintenance screen gains a "Similar content (perceptual)"
section — a Scan button that runs the fingerprint job on a background
thread (live "N / total" progress via shared atomics + an mpsc result
channel), grouped results with a recommended-keep marker, and a per-group
"Remove non-recommended copies" that deletes via the existing
remove_files path and re-scans. Closes the desktop/web parity gap for
content-aware dedup.
118 tests pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Wires the fingerprint engine into the web Maintenance view.
- web.rs: a DedupState job (one at a time) on its own thread. POST
/api/maintenance/dedup/scan snapshots the library, mtime-gates against
stored fingerprints, fingerprints the new/changed videos in parallel
(progress counter), prunes vanished entries, groups by visual
similarity, and builds review rows (title/channel/size/files + a
recommended-keep = largest copy). GET /api/maintenance/dedup/status
polls progress + results. Deletion reuses /api/maintenance/remove.
- index.html: a "Similar content (perceptual)" section in the Maintenance
modal — Scan button, live progress bar, grouped results with
checkboxes (recommended-keep pre-unchecked), and a re-scan that drops
deleted copies. Poller self-cancels when the modal closes.
Integration test (ffmpeg-gated): generates orig + CRF-38 downscaled
re-encode + an unrelated clip, runs the real scan end-to-end, and asserts
the first two group while the third stays out. 118 tests pass.
Desktop Maintenance UI next.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The core of content-aware duplicate detection — finding the same video
under a different ID (reupload / mirror / re-encode), which the ID-based
maintenance scan can't.
- src/fingerprint.rs: samples 6 frames per video via ffmpeg keyframe seek
(fast — never full-decodes), downscales to a 9x8 grayscale grid, and
dHashes each to a 64-bit frame hash. `group_similar` clusters videos
whose frame hashes match within a Hamming threshold, comparing only
within a duration-tolerance window (sorted sliding window + union-find)
so it stays near-linear instead of O(n²). Parallel `compute_batch` over
a hand-rolled scoped thread pool (no rayon), with a progress counter.
- src/database.rs: `video_fingerprint` table keyed by (path, mtime) like
info_cache — hash once, reuse forever; new downloads hashed as they
arrive. Store / load / mtime-gate / prune helpers.
Unit tests cover dHash, Hamming, duration-bucketed + transitive grouping,
and DB round-trip/replace/prune. An opt-in `real_ffmpeg_groups_reencodes`
test (run with --ignored) generates a video + a downscaled CRF-38
re-encode + an unrelated clip and asserts the first two group and the
third doesn't — measured ~0.5s/video serial (≈65ms/video across 8 cores).
UI wiring (background build + review screen in both front-ends) follows.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Surface the downloaded .vtt subtitles as a searchable, click-to-seek
transcript — a thing Tartube doesn't do at all.
Web (pure frontend): a 📄 button in the player opens a transcript pane
beside the video. It fetches the already-served .vtt, parses cues
client-side, and offers full-text search (highlighted), click-to-seek,
and a live highlight + auto-scroll of the current line as the video
plays. Reuses the chapters-pane layout; hidden on narrow screens.
Desktop (egui): a 📄 Transcript button on the detail panel opens a
floating window that reads the .vtt off disk (new `vtt` parser module),
shows the searchable cue list, and seeks the running mpv via its
JSON-IPC socket when you click a line.
New `src/vtt.rs`: a small, tolerant WebVTT/SRT cue parser (start time +
tag-stripped text, consecutive-duplicate collapse) with unit tests.
Closes the desktop/web parity gap for transcripts.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A new FTS5 index (`video_search`) over every video's title, channel, and
description — searchable across the whole library, not just the loaded
grid. `search_meta` tracks each video's mtime so a routine rescan only
re-reads a description sidecar when the video actually changed; vanished
videos are evicted. The index is refreshed after every scan in both
front-ends via the shared `library::build_search_entries`.
- database.rs: schema + `sync_search_index` (mtime-gated, one txn) +
`search_videos` (ranked, with a highlighted snippet) + a safe prefix
MATCH builder. Unit-tested (index/search/prefix/AND/evict/garbage).
- web.rs: `GET /api/search?q=&limit=`; index synced after the initial
scan, rescan, and maintenance-remove.
- Web UI: a 🔍 header button + `f` shortcut open a debounced search modal
with ranked results and highlighted description snippets; clicking a
result jumps to it.
- Desktop (egui): a 🔎 Search button opens a floating results window
querying the same index; clicking a result plays the video. Closes the
desktop/web parity gap for search.
Integration test seeds a real video + description, rescans, and asserts
title / prefix / description-only matches hit and unrelated misses.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Comment viewer (web UI): the flat tree gains search (highlight + keep
ancestor context), sort (threaded / top / newest / oldest), per-thread
collapse/expand with collapse-all/expand-all, an OP badge for the
uploader, and a "new since last visit" highlight + count (localStorage
per-video timestamp). API now returns each comment's `timestamp` and
`author_is_uploader`.
SponsorBlock: was a hardcoded `--sponsorblock-mark all`. Now a real
setting (off / mark / remove) with a per-channel override, threaded
through the full five touchpoints — config `[backup].sponsorblock_mode`
(default "mark", preserving prior behavior), DownloadOptions override,
a downloader resolver (apply_sponsorblock), and both UIs (egui Settings
+ channel dialog; web Settings modal + channel dialog + SettingsPayload).
Integration tests extended: settings round-trip asserts the sponsorblock
default + persistence; channel-options round-trip asserts the override.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A yt-dlp process can stall with the OS process alive but no progress — a
wedged TLS handshake, an unresponsive server, a dead socket mid-fragment.
Before, that job hung forever, holding a concurrency slot until the user
noticed and cancelled.
Now each Job tracks last_activity (bumped on any stdout/stderr line or
progress message). poll() runs check_watchdog(): a Running job silent for
HANG_TIMEOUT (5 min) gets SIGKILLed by pid via libc::kill. The threshold
sits well above any legitimate quiet gap — yt-dlp prints a line per
download chunk, polls every 30s under --wait-for-video, and our longest
sleep is ~30s — so 5 min of total silence is a real stall.
To capture the pid, spawn_job + enqueue_transcode now spawn the child
*before* the reader thread (the thread still owns it for the blocking
wait()). A watchdog kill sets watchdog_killed, and drain() forces the
resulting failure to NetworkError (retryable) instead of the Other that
classify() would return for an output-less kill — so the existing
auto-retry path re-queues it (a hang is transient).
Non-Unix builds keep the flagging but can't force-kill (no portable
pid-kill; Windows isn't first-class). JobState gains Debug for tests.
2 tests: watchdog kill → retryable class; normal failures still
classify from the log. 98 pass.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
