Ranked feed — affinity was broken:
- Was looking up user_tag_affinity by v.category (e.g. "Science & Technology")
but affinity is stored using fine-grained video tags ("linux", "rust", etc.)
- Now uses SUM across all matching affinities: category OR any tag found in the
video's tags JSON via instr() — up to 5 matches to prevent runaway scores
Ranked feed — completion rate now influences channel scoring:
- Added avg_completion_pct to channel_stats CTE (AVG of completion_percent)
- Channels where you finish videos score higher; channels you bail on score lower
- Defaults to 50% (neutral) for channels with no tracked completions
Progress endpoint — backend auto-watched safety net:
- If completion_percent reaches ≥90% on a video >60s, mark watched automatically
- Catches cases where browser closes before the 10s debounce fires
- Guards against double-calling _update_affinity with not prev_watched check
VideoPlayer — seamless local file switch:
- Removed switchedToLocal state which caused a race condition: video loaded with
local_file_url already set but flag was still false, requiring a page refresh
- local_file_url from the backend is the single source of truth (backend gates
it with os.path.exists so it only appears when the file is actually on disk)
- Show spinner while video metadata loads, then immediately show local player
if file exists — no YouTube flash for already-downloaded videos
- After download completes, single refetchVideo() picks up the new URL and
React re-renders directly into local player
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Rewrite list_channels to run exactly 4 SQL queries regardless of channel
count: channel rows, aggregated video stats (GROUP BY), new-video counts,
and latest video (derived-table JOIN replaces per-row correlated subquery)
- Remove dead _CHANNEL_STATS_SELECT (orphaned after the rewrite)
- Fix upload_frequency_days: use pre-computed date_span_days from vstats
instead of a broken per-channel db.execute() call
- Restrict new_counts query to id_csv so it uses idx_videos_channel_indexed
- markChannelsSeen: optimistic setQueryData instead of invalidateQueries,
eliminating a full channel-list re-fetch on every Following page visit
- DownloadIndicator idle poll: 10s → 30s (no need to hit DB when idle)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The CTE approach returned 0 rows — likely a SQLite/SQLAlchemy interaction
with :user_id appearing in multiple CTEs. Reverted to the original
correlated-subquery form which is proven correct.
The 4 indexes added in the previous commit still apply and will make
the per-channel subqueries faster once the DB is indexed on startup.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The old _CHANNEL_STATS_SELECT ran 9 correlated subqueries for each
channel row. With 1266 channels that was ~11000 sub-executions per
GET /channels request, causing multi-second (or timeout) delays.
New approach: 2 CTEs (vinfo for counts/sums, nc for new_count) each do
a single aggregated pass over all followed-channel videos, joined back
to channels. Only 2 correlated LIMIT-1 subqueries remain for
latest_video_id/title (fast with the new index).
Also adds 4 indexes on startup (IF NOT EXISTS — safe to deploy):
- videos(channel_id, published_at DESC) — latest video lookups
- videos(channel_id, indexed_at) — new_count filter
- user_videos(video_id, user_id) — watch/download aggregation
- user_channels(user_id, status) — followed channel filter
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Sync throttling:
- sync-all now skips channels crawled within the last 6 hours (prevents
re-scraping 1266 channels on every button press)
- Channels are queued into a single _index_channels_batch task that runs
with 1.5s delay between each yt-dlp call instead of firing 1266
background tasks simultaneously
- Startup enrich task reduced from 10 to 3 videos (3 yt-dlp calls on
each container restart)
- Enrich task adds 2s sleep between metadata fetches
SQLite stability:
- busy_timeout=5000 prevents SQLITE_BUSY errors under concurrent load
- synchronous=NORMAL speeds up writes without data loss risk (safe with WAL)
Following page:
- staleTime: 60s on channels query so cached data is reused immediately
on revisit; gcTime keeps it in memory for 5 min
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
All inline SQL queries in the feed endpoint (chronological, random,
inbox, ranked scored CTE, and discovery injection) were missing
c.thumbnail_url AS channel_thumbnail_url — only _VIDEO_SELECT had it.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Reading from the channels query cache was unreliable (cache might not be
loaded, or channel not followed). Add c.thumbnail_url AS channel_thumbnail_url
to _VIDEO_SELECT so every video response carries its channel avatar directly.
VideoCard uses it with cache as fallback.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Comments: switch from CLI --write-comments to yt-dlp Python API with
getcomments=True — more reliable, proper extractor_args dict format
Dislikes: add dislike_count column, fetch from returnyoutubedislike.com
after each video metadata upsert (5s timeout, non-fatal)
UI: replace emoji like count with a like/dislike ratio bar — blue fill
showing like proportion, labels on each end; views stay in meta row
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Same pattern as view_count: model column, yt-dlp extraction, SQL select,
VideoDetail field, startup migration, and display in Watch meta row.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Video model: view_count column (Integer, nullable)
- ytdlp._normalize_video: extract view_count from yt-dlp info
- _VIDEO_SELECT: include v.view_count in all queries
- VideoDetail schema: view_count field
- Watch page: formatViews() helper, show "X.XM views" in meta row
alongside date and category
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- VideoComment model (video_id, author, text, likes, is_pinned, published_at)
- fetch_video_comments() in ytdlp.py: top 20 comments, no reply threads,
sorted pinned-first then by likes
- GET /videos/by-yt/{id}/comments — returns cached comments instantly
- POST /videos/by-yt/{id}/comments/refresh — fetches from YouTube, stores, returns
- Watch page: CommentsSection shows "Load comments" button when uncached,
renders comments with author/likes once loaded; Refresh link to re-fetch
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Backend: DELETE /stats/taste/{tag} removes the row from user_tag_affinity
- API: deleteTasteTag(tag) helper
- Stats UI: × button on each tag chip, faint by default, full opacity on hover;
invalidates stats query so the tag disappears immediately
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
nodeenv is a Python package that downloads a pre-built Node.js binary —
no apt repos, no compilation, guaranteed to work in python:3.12-slim.
The 'node' binary is linked into /usr/local/bin so yt-dlp can find it.
With Node.js available the web client works fully (37 formats) and can
solve YouTube's n-challenge that every other approach was failing on.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
web_embedded: supports cookies, no Node.js/JS runtime needed, 23 video
formats available. android_vr was skipped by yt-dlp when cookies are
present since that client doesn't support cookie auth.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
android_vr provides pre-signed format URLs that bypass YouTube's
n-challenge and signature JS requirements entirely. Tested: 23 video
formats available without any JavaScript runtime installed.
Reverts Node.js Dockerfile addition (which failed to build anyway).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Debian installs nodejs as /usr/bin/nodejs but yt-dlp looks for 'node'.
The symlink ensures yt-dlp can find the runtime.
Diagnostics now report node path/version and yt-dlp version so we can
verify the environment without shelling into the container.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- _cookie_args() no longer falls through to --cookies-from-browser when
cookies_file is configured but missing. Firefox isn't installed in the
Docker image, so that fallback caused yt-dlp to exit with empty stdout
and every metadata fetch to return "Video not found on YouTube".
- fetch_video_metadata() now retries without auth args if the first call
fails, so a broken cookie config can't block public video fetches.
- Add use_oauth2 setting + full device-auth flow (POST /settings/oauth2-init,
GET /settings/oauth2-status) with OAuth2Section UI in Settings page.
- Add GET /settings/ytdlp-test diagnostics endpoint.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Self-hosted personal YouTube management app.
FastAPI + SQLite backend, React + Vite + Tailwind frontend.
Dockerfiles and compose included for Portainer deployment.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
