Expose the maintenance job's rolling re-validation batch (videos re-checked per
run) as an admin control on the Scheduler dashboard, stored in app_state
(migration 0018; NULL = the env/config default). The job reads the effective
value each run via a state helper, clamped to a sane range. Trilingual.
Add per-job "Run now" buttons and a "Start all now" button to the admin Scheduler
dashboard (admin-gated endpoints). Triggers run the job in a background thread
independent of its interval, refusing a concurrent run (409). While running, the
long jobs (maintenance, enrich, backfill, shorts) report live progress through a
decoupled contextvar sink, shown as a progress bar on the job row via the existing
4s poll. A manually-triggered run posts a completion notification to the triggering
admin's inbox (scheduled runs stay silent to avoid spam); the inbox renders the
"scheduler" type trilingually from type+data. While here, give the maintenance job
its missing dashboard label/description in all three languages.
Add a daily maintenance/validation job that detects videos which can't be played
anywhere and retires them safely. Two phases: re-check already-flagged videos
(recover if available again, else hard-delete once the grace period elapses,
cascading to states/playlist items), and a rolling re-validation of the
least-recently-checked currently-available videos that flags newly-unplayable
ones (hidden from the feed immediately via unavailable_since).
Detection is ~free: a video missing from the videos.list response is
deleted-or-private; an `upcoming` premiere >2 days past its scheduled start that
never went live is abandoned. A still-live broadcast is kept (legit 24/7 stream).
Enrichment now also fetches part=status to populate the status columns. Grace is
7 days for removed videos, none for abandoned. Before deleting, affected users
get one batched notification (never per-video). Interval is admin-tunable via the
Scheduler dashboard; batch size and grace are config. Quota-attributed to the
system and bounded by the same backfill reserve as the other jobs.
Add a server-backed notification center that coexists with the client-side
transient bell: a per-user `notifications` table (type/title/body/data JSON/
read/dismissed), a `/api/me/notifications` CRUD API (list, unread_count, read,
read_all, dismiss, clear), and a left-nav inbox module with a live unread badge
polled via useLiveQuery. Known types render trilingual text from type+data
(English stored text is the fallback); read rows are trimmed past a soft cap.
Also adds the schema the maintenance job builds on: videos.list?part=status
columns (embeddable/privacy_status/upload_status) and the validation lifecycle
columns (last_checked_at, unavailable_since, unavailable_reason).
Page-id validation is centralized in one PAGES source of truth (isPage) so the
new page survives reload without a second allowlist to keep in sync.
- Channel-row tags show only what's attached; a '+' opens a per-channel picker. Clicking a tag
filters the feed by it; a 'Your tags' sidebar widget makes that filter visible/clearable.
- Tag manager dialog (add/rename/delete; delete only confirms when the tag is in use), reachable
from the Channel manager and the feed sidebar; hovering a tag's count lists its channels, each a
link that focuses it in the Channel manager (DataTable gains an external filter input).
- Video cards get a reset action that clears all watch state (incl. an in-progress position).
- api.req() now raises the error dialog on 5xx and 400/409/422 with the server's reason.
- Duplicate-tag create/rename hit the uq_tags_user_name constraint and 500'd; now caught and
returned as a clear 409 ('You already have a tag named …').
- New global error dialog (errorDialog store + ErrorDialog modal) for definitive server refusals;
the api layer wiring + mount land with the rest of the round.
- Modal now keeps a stack so ESC dismisses only the top modal — an error over the tag editor
closes the error and leaves the editor open.
- Add Last upload, total Length and a Normal/Shorts/Live breakdown, computed on read in the
existing grouped query (~35ms over the full catalog — measured, so no denormalization).
- Subscribers shown compact (919K), since the YouTube API already rounds them.
- Tag cell now shows only the tags actually attached to a channel; a '+' opens a per-channel
picker to attach/detach, instead of listing every user tag on every row.
POST /api/channels/{id}/reset-backfill (admin-only): clear the channel's backfill
markers, re-opt into deep backfill and re-run its recent pull now — a re-fetch-from-
scratch trigger regardless of current sync state. Idempotent (videos upsert by id).
The in-app player marks a video watched when it reaches the end, firing
POST /videos/{id}/state at the same instant as a progress checkpoint. The
progress endpoint deletes the 'new' video_states row near the end, so the
concurrent state UPDATE matched 0 rows and raised StaleDataError -> 500. The
watched status was never persisted (and Feed swallowed the error), so the
video stayed in the unwatched feed even after a refresh.
- set_video_state: atomic INSERT ... ON CONFLICT (uq_user_video) DO UPDATE for
watched/hidden, immune to a concurrent delete; tolerate StaleDataError on the
'new' branch and in the progress endpoint.
- PlayerModal: skip the redundant near-end progress checkpoint when we just
auto-marked watched, removing the self-inflicted race.
- Feed: drop the optimistic override if the server rejects the change, so a
failed request can't leave a card phantom-hidden.
Add scheduler_settings (per-job interval override, migration 0015) and a
PATCH /api/admin/scheduler/jobs/{id} that persists the override and live-
reschedules the running APScheduler job. Intervals load from the DB (env
defaults as fallback); the snapshot reports the live trigger interval.
Record each scheduler job's run activity in-process (running / last result /
error / timestamps) and expose a snapshot. New admin-only GET
/api/admin/scheduler returns the per-job activity (with APScheduler next-run
times), the DB-derived work still queued (channels/videos pending, deep ETA,
live-refresh backlog) and the shared quota picture.
enrich_pending only touches enriched_at IS NULL, so a video first seen
while live was stamped live + duration-null and never revisited — staying
'live' with no duration forever after the broadcast ended. Add refresh_live
(run after each enrich pass) that re-fetches anything still live/upcoming,
plus just-ended was_live videos that haven't got their duration yet, until
they settle. Cheap: videos.list is 1 unit per 50 ids.
The shared demo account no longer hits YouTube affordances: the onboarding
wizard never opens (or renders) for it, the empty-feed prompt nudges into
the shared library instead of the connect wizard, and the browser-facing
/auth/upgrade redirects the demo home instead of returning a raw 403 JSON.
Admin endpoints to manage the demo email whitelist (DB-backed, no env)
and a manual reset that wipes the demo account's per-user state and
re-seeds a few sample playlists from the shared catalog.
Whitelisted emails enter the shared demo user via /auth/demo (lazily
created, no OAuth token/scope), rate-limited per IP and answering
uniformly so it can't be hammered as an enumeration oracle. Add a
require_human dependency that blocks the demo account from quota-spending
sync endpoints and the OAuth upgrade flow, and surface is_demo on /api/me.
Add the shared demo-account plumbing: users.is_demo marks the single
shared demo user, demo_whitelist holds the admin-curated emails that may
enter it without Google sign-in, and a small in-process RateLimiter
(generic groundwork) for throttling the demo-login endpoint per IP.
1) Move the Show view filter (Unwatched/In progress/All/Watched/Hidden) up into the
toolbar chip row as its own single-select group, divided from the content-type chips;
removed the 'show' sidebar widget (sidebar now = Upload date / Language / Topic).
2) Rework ordering like the Playlists page: a sort-key dropdown (Date, Popular, Duration,
Name, Channel subscribers, Channel priority, Surprise me) + a single asc/desc arrow
toggle, instead of separate directional entries. 'Most viewed' is now 'Popular'.
Backend gains the missing directions (views_asc, title_desc, subscribers_asc,
priority_asc); the frontend maps (key, dir) -> the backend sort string, so FeedFilters
is unchanged. Trilingual (feed.sortKey.*, feed.dirAsc/dirDesc). Feed.tsx normalized to LF.
Sessions created before multi-session existed had no account_ids, so adding a second
account left the first one (e.g. the long-lived session) absent from the switcher.
current_user now ensures the active user_id is always present in account_ids, so the
switch list is never missing the account you're using.
Track every account that completes OAuth in a browser session (session 'account_ids',
~14-day cookie), so the user can switch between them without another Google round-trip.
New GET /api/me/accounts (switchable accounts, active flagged) and POST /api/me/switch
(guarded: target must be in the session list — proof it signed in here — and re-checked
against is_allowed in case the invite was revoked). Logout now signs out the active
account and falls back to the most recent remaining one, else clears the session. UI: the
account popover lists the other signed-in accounts (click to switch, reloads) plus an
'Add another account' action (-> /auth/login, which uses prompt=select_account).
Trilingual. Third step of Epic N.
The session has autoflush off, so recompute_dirty's query read the last-committed
state, not the current request's pending item/order edits — so the first reorder
after a sync compared against itself and missed dirty, and only the next edit showed
it. Flush pending changes before computing the current fingerprint. Verified: a swap
now flips dirty immediately, and reverting to the original clears it.
Store a SHA-256 of each playlist's name + ordered video ids as last synced from /
pushed to YouTube (migration 0013, backfilled for already-clean linked playlists).
'dirty' is now recomputed on every edit by comparing the current fingerprint to that
baseline instead of being stickily set true. So manually restoring the original order
(or undo back to it) clears dirty on its own — the Reset to YouTube button disappears
and no YouTube read is needed. Baseline is set on read-sync, repull (reset), and a
clean push; a missing baseline counts as dirty (unknown YT state). Verified the
migration's fingerprint matches the app's runtime fingerprint exactly.
The in-session undo/redo is lost when switching playlists, leaving no way to undo
local edits to a YouTube-linked list afterwards. Add a per-playlist 'Reset to YouTube'
action (shown when a linked playlist is dirty) that force-repulls that single playlist
from YouTube, replacing its items/order/name and clearing dirty — even though the bulk
read-sync skips dirty playlists. Backend: repull_playlist() + POST /api/playlists/
{id}/revert-youtube (read-scope gated). Confirm dialog (destructive). Trilingual.
1) Fix: the selected playlist is now persisted (localStorage) and scrolled into view,
so F5 keeps it instead of jumping to the first one.
2) Consolidate the in-detail sort: one key select (Title/Duration/Channel) + an
asc/desc direction toggle (was separate A-Z / Z-A / shortest / longest options);
add Channel as a sort key. Direction also orders the channel groups when grouping.
3) Left rail sorting: by name / item count / total length, asc/desc, plus an
'unsynced first' toggle that floats playlists with unpushed edits to the top.
Backend: list/summary now return total_duration_seconds (grouped sum). The rail
also shows each playlist's total length. Sort prefs persist to localStorage.
Mirrors (source='youtube') were read-only; per the original two-way design they can
now be edited locally (add/remove/reorder/rename) and pushed back. The read-sync
skips a dirty mirror so it won't clobber unpushed edits; a successful push clears
dirty and lets the mirror refresh. _mark_dirty now covers any YouTube-linked list
(exported local or mirror), and rename marks dirty too. push/delete no longer reject
mirrors. Push reconciles the title as well (cheap playlists.list compare, then
playlists.update only if changed) so a local rename doesn't revert on the next pull.
Frontend: editable = not-built-in (was also excluding mirrors); rows/reorder enabled
for all owned lists; AddToPlaylist popover lists mirrors too (with a YT marker); the
origin chip now reads 'edits sync back'. Trilingual.
Local 'user' playlists (not Watch later, not YouTube mirrors) can now be pushed
to YouTube: create a new playlist on first push, else reconcile membership + order
so YouTube matches local (local wins). plan_push computes a dry-run estimate
(inserts/deletes/reorders + quota units + divergence) read from the live YouTube
state; push refuses if the estimate exceeds the remaining daily quota so a big
playlist can't strand half-pushed. Reorder uses an insertion-sort move count that
matches what executes. Edits to a linked playlist set dirty; a successful push
clears it. DELETE accepts on_youtube to also delete the playlist on YouTube. The
read-sync now skips YouTube playlists already owned by a local export, avoiding
duplicate read-only mirrors.
Add OAuth-only write helpers to YouTubeClient (each 50 quota units):
create_playlist, add_playlist_item, move_playlist_item, delete_playlist_item,
delete_playlist, plus iter_playlist_items_with_ids for diffing. A shared _write
helper records quota and raises YouTubeError on non-2xx.
Mirror each user's own YouTube playlists into local source='youtube' playlists,
one-way (YT -> local). New client methods iter_my_playlists / iter_my_playlist_video_ids
(OAuth, so private playlists work); sync/playlists.py reconciles the mirror (matching YT
order) and ingests any playlist videos not in the shared catalog yet (with stub channels).
POST /api/playlists/sync-youtube for manual sync (read-scope gated, per-user quota) plus a
scheduler job (playlist_sync_minutes, default 6h) that syncs all read-scope users. YouTube's
Watch Later / History are not API-accessible and are never synced.
The old per-video status='saved' becomes membership in a built-in, undeletable
'watch_later' playlist. Migration 0012 moves existing saved videos into each user's
Watch later list and demotes the states to 'new'. The feed/playlist serializers now
expose a 'saved' boolean (EXISTS in watch_later); the card bookmark toggles watch_later
via new /api/playlists/watch-later add/remove endpoints (create-on-demand) instead of
setting a status. Removed the 'saved' show-filter and status. Watch later shows a
localized name and hides rename/delete in the Playlists page and add-to-playlist popover.
Add per-user local playlists: Playlist + PlaylistItem models (with forward-looking
kind/source/yt_playlist_id/dirty columns for the later YouTube-sync phases) and
migration 0011. New /api/playlists routes: list (with optional contains=<video_id>
membership flags for the add-to-playlist popover), create, get detail (items
serialized via the feed serializer, with per-user watch state), rename, delete,
add/remove item, and reorder. Watch later (built-in) is protected from deletion.
In AND ("All") topic mode the facet endpoint still excluded the topic selections
when counting topic chips, so every topic kept its full count and none dropped
out as you narrowed — e.g. picking Comedy left Cooking visible even though no
channel has both. Count topics conjunctively in AND mode (keep the selected
topics applied) so each remaining chip reflects channels that ALSO have all
already-selected topics; non-co-occurring tags fall to zero and hide. OR mode
stays disjunctive. Verified: Comedy selected narrows topic chips 21 -> 6.
Add a facets endpoint that returns per-tag channel counts for the current filter
context (scope, channel, date, content type, search, watch state, and the other
category's tags). Each category is counted with its own selections ignored —
standard drill-down faceting — by a new exclude_tag_category param threaded into
_filtered_query, so selecting one topic doesn't zero out the other topics. Count
is distinct channels with a matching video, keeping the channel-count chip
semantics. Reuses the feed's filter query so both stay in lockstep.
The feed query was always scoped to the user's own non-hidden subscriptions
via an INNER JOIN on subscriptions. Add a scope param: scope=my (default)
keeps that behaviour; scope=all LEFT-joins the subscription instead, so every
video in the shared catalog shows while per-channel priority still resolves
for channels the user is subscribed to. Per-user watch state stays private via
the VideoState outer join in both modes. The priority sort is made null-safe
(coalesce to 0) since unsubscribed channels have no subscription row in all-mode.
Make APP_VERSION come from the VERSION file (Dockerfile reads it for the SPA build
and ships it for the backend to read at runtime) instead of depending solely on a
deploy-time build-arg. This keeps the version correct even for a plain
`docker compose build` and removes the first-deploy bootstrap gap. git_sha/build_date
stay best-effort build-args.
Set up react-i18next with locale files auto-loaded per area (Vite glob), a compact
LanguageSwitcher, and language as a server-persisted preference (preferences.language)
mirrored to localStorage. On first login the default UI language is guessed from the
Google-reported locale (hu/en/de, else English). vite-env.d.ts types the build-time env.
Add a VERSION file (0.1.0) and inject APP_VERSION/GIT_SHA/BUILD_DATE as Docker
build-args (both stages; Vite inlines them into the SPA). New public GET
/api/version returns app_version, git_sha, build_date and the Alembic head as the
database revision. deploy.sh and the localdev build pass the args.
A channel whose stored uploads already meet or exceed YouTube's advertised
video_count holds its whole history, but backfill_done could stay false forever
when the deep cursor never reached the end (e.g. a small channel that was never
deep-requested, so the demand-driven deep job never ran). Such channels nagged
as 'needs full history' despite having every video. Add reconcile_full_history()
(idempotent, no quota) and run it at the end of each recent/deep backfill cycle
so backfill_done self-heals.
Show 'N yours / M total' in the header (your subscriptions vs. the whole shared
catalog) with a tooltip, backed by a new total_videos field on /sync/my-status.
The admin pause button now only appears when there's sync work to pause; Resume
still shows whenever sync is paused.
Add position_seconds (+progress_updated_at) to video_states so watch progress
survives across devices and can drive a feed filter. New POST
/api/videos/{id}/progress checkpoints the player position (clearing trivially
-early and near-finished positions). Feed serialize exposes position_seconds and
a show=in_progress filter lists started-but-unfinished videos. Un-marking
'watched' now keeps a stored position instead of deleting the row.
- requirements: cryptography >=46.0.7 (was pinned <46, which excluded the fix for
the CVEs pip-audit flagged in our Fernet/crypto library). pip-audit now clean.
- Dockerfile: upgrade pip before installing deps (patches installer-level CVEs).
- auth: /auth/upgrade now defaults to the least-privileged read scope; only an
explicit access=write requests the write scope.
Base login now requests only openid/email/profile (non-sensitive), so a new user
gets a clean Google consent with no "unverified app" warning and no 7-day refresh
token expiry. YouTube read (youtube.readonly) and write (youtube) are granted later
by the onboarding wizard via a parameterized /auth/upgrade?access=read|write.
Security fixes folded in from the baseline audit:
- config: refuse to boot in production (https OAUTH_REDIRECT_URL) with the
placeholder/short SECRET_KEY or a missing TOKEN_ENCRYPTION_KEY, closing a
session-forgery / admin-impersonation hole.
- main: mark the session cookie Secure when served over HTTPS.
- me: expose can_read; sync/subscriptions returns a friendly 403 (not a 500)
until YouTube read access is granted.
- Narrow the title hover target to the actual text, not the whole row.
- Linkify descriptions: timestamps (mm:ss / hh:mm:ss) seek the player; emails
become mailto:; hashtags link to YouTube's hashtag feed; other URLs open in a
new tab. Blank lines are stripped so the popover isn't mostly whitespace.
- YouTube links play in the inline player: a link to the current video seeks
(honoring t=), a link to another video navigates the player to it, with a Back
button to the original. While on a linked video the title/author come from the
player and its views/date/duration + a clickable channel come from the detail
endpoint, which falls back to the YouTube API (videos.list, attributed to the
user) for videos not in our DB.
- Watched: an explicit toggle in the modal (Mark watched / Watched→unmark) plus
auto-mark when playback reaches the end (within 10s, or on the ended event).
- Compact layout: drop the header bar and the redundant 'Open on YouTube' button
(the embed's own YouTube logo already jumps out); Close moves to the title row,
channel + meta share one line — fits without a scrollbar at higher zoom.
- Card actions reflect status: watched shows a double-check, saved a filled
bookmark, with matching tooltips.
- Description: new GET /api/videos/{id} exposes the already-stored description,
shown in a popover when hovering the modal title (fetched lazily).
Track who burned how much YouTube API quota. A QuotaEvent audit log (migration
0009) records every spend with the triggering user (NULL = background/system) and
an action label, set via a request/job-scoped contextvar (quota.attribute) so no
call signatures change. User-initiated work (sync subscriptions, unsubscribe,
opt-in recent backfill, manual enrich) attributes to the user; scheduler work to
System, split by action.
- backend: QuotaEvent model + migration 0009; quota.attribute() contextvar;
record_usage logs events; entry points wrapped (routes/sync, routes/channels,
scheduler); GET /api/quota/my-usage + GET /api/quota/admin
- frontend: admin-only Stats page (header nav, page=stats) with daily bars +
per-user breakdown by action and range picker; 'Your API usage' in Settings ->
Sync for every user
Verified: attribution + endpoints compute correctly; events are per-user vs System.
deep_requested is per-subscription, but deep backfill is channel-wide and shared:
once any subscriber requests full history, the whole back-catalog arrives for
everyone. The chip only looked at the current user's flag, so a second subscriber
saw a misleading 'get full history' on a channel already queued by someone else.
Add a channel-level deep_in_queue to /api/channels and show an informational
'full history queued' badge (vs the owner's cancelable button) in that case.
Personalize the access-request and approval emails, add a real Date header and a
Reply-To (approval -> admin, new-request -> requester) so they read as a
conversation rather than a no-reply blast — small nudges for deliverability;
sender reputation still does most of the work.
Move the access whitelist from the ALLOWED_EMAILS env var into a DB Invite table
(env kept as bootstrap fallback), and add a self-service request + admin approval
flow with fail-soft email.
- models: Invite(email, status pending|approved|denied, requested_at, decided_*)
- migration 0008: invites table; seed env ALLOWED_EMAILS u ADMIN_EMAILS as approved
- auth: is_allowed() (DB-first, env fallback); a denied Google login records a pending
request and bounces to /?access=requested instead of a raw 403; public POST
/auth/request-access; upsert is idempotent so repeats don't re-spam admins
- routes/admin.py (admin-only): list/approve/deny invites + manual add
- email.py: smtplib + Gmail App Password, fail-soft (skips if SMTP unset)
- /api/me exposes pending_invites; config + .env.example gain SMTP_*
- UI: Login 'Request access' form + access=requested/denied handling; Settings ->
Access requests (approve/deny + add); admin nudge toast on pending requests
Verified locally: request-access creates a pending invite and emails the admin;
seed approved npeter83; guinea-pig yt.trash2023 denied until approved.
When recent backfill stopped partway through a page (age cutoff or count cap),
it stored the *next* page token as the deep-backfill cursor, so the older items
remaining on that same page were collected by neither pass — a channel could
report backfill_done while silently missing a band of videos right around the
365-day cutoff. Now resume deep from the page we stopped on (re-fetched; inserts
are idempotent). Verified: Pánczél went 200 -> 209 stored (= full uploads
playlist) after re-backfill.
When a user requests full history for a channel whose recent uploads aren't
fetched yet, run a one-channel recent backfill synchronously in the request so
the feed populates at once instead of waiting for the scheduler. Deep paging
still follows on the scheduler (recent-then-deep). The deep-toggle mutation now
also refreshes my-status and the feed.
