Every page's floating header (search bar + sync chip) and the toolbar band now centre on the SAME
max-width the content column uses, instead of spanning the whole column while a max-w page centres
its grid. So the search bar lines up with the left edge of the cards/table on every page, and sits
centred with the content — not against the full page width.
- PageShell takes a `maxWidth`, published as --content-mw; the header adds max-w-[--content-mw] mx-auto
inside its left-4 / right-(1rem+scrollbar) span, so it centres on that width (`none` = full width).
- moduleRegistry PAGE_CONTENT carries each page's content max width (plex 1600px, channels 96rem,
playlists 64rem, downloads 56rem, notifications/users 48rem; feed stays full width) — one source.
- Playlists: its content (header, filter row, table) was left-aligned max-w-5xl → now mx-auto, so the
table is centred and the header lines up with it.
Verified at 1920px: header left edge == content left edge — feed 340, channels 213, playlists 487.
Plex browse now uses the same fixed-width ItemCount pill as the feed (its title count → "N items"),
placed as the leading element, and the toolbar gap tightened 2 → 1.5 to match. One consistent
toolbar shape across the source modules.
- New ItemCount: a generic, FIXED-WIDTH count pill (glass, non-interactive) — "192,050 items"
(neutral noun, dynamic plural), min-w-[9rem] so a changing number never shifts the controls
beside it; a million or more collapses to "1M+" with the exact figure in the title tooltip.
- Feed toolbar: ItemCount is the fixed leading element now; dropped the mx-1 separators so every
gap is a uniform 1.5 (Content-types↔Source used to be wider than Watch-state↔Content-types).
- Chrome-to-card alignment (app-wide, via the shared Header + PageShell): the floating header and
the toolbar band now span exactly the card grid's width — left-4, and a right edge of
1rem + the scrollbar gutter (measured once into --sbw), so the fixed chrome lines up with the
scrolled cards instead of running ~15px past them. Wraps cleanly to multiple rows on mobile.
Renames the count testid feed-result-count → item-count (E2Es updated).
The user no longer wants the header's "FEED" pill (module label + accent dot + ◀/▶ module-step
arrows). Drop it everywhere — the header now starts with the module's SearchBar; module switching
stays on the nav rail.
Deletes ModuleName.tsx, the now-unused stepModule() in lib/modules, and the Header's
module-order/label/setPage plumbing that only fed it.
The earlier bump measured the view-switcher's BUTTON (16px, inherited) instead of its visible
label span (text-sm = 14px), and 14px is the size the user actually finds comfortable. Walk the
over-sized triggers back to text-sm: FilterSelect, SortControl, ViewSwitcher, and the Combobox
input (which was 14px before the bump anyway). The secondary text already at text-sm/text-xs
(count, chips, widget headers, section labels) was correct and stays.
The user prefers FilterSelect's rounded-full glass pill over the ViewSwitcher's squarer
rounded-lg trigger. Give the ViewSwitcher trigger the same look — glass-card/glass-hover,
rounded-full, px-3 py-1, open→border-accent, and a 16px label (was text-sm) — so the two read
as one control family. Applies to every ViewSwitcher consumer (feed, channel manager, Plex
playlists).
Native <select>s, the Combobox input box, and playlist/list cards are a different control class
(form fields / variable-height inputs / cards) and keep their own shape.
The view-switcher's 16px reads comfortably; the toolbar and filter controls sat at 12px. Bring
the marked controls up to match, and lift the general scale on the shared filter surfaces.
- Toolbar (feed + channel + Plex, via the shared components): FilterSelect and SortControl trigger
pills text-xs → text-base (16px, the view-switcher size); the result count → text-sm; the
active-filter chips → text-sm with the "FILTERED BY" label text-2xs → text-xs.
- Filterbar: the Combobox picker input → text-base (matching the toolbar pills) + its chips → text-sm;
the panel/widget headers (SidePanel collapsed-rail + panel title, PanelGroup widget titles) up one step.
Channel info and the manager tables were already at 14–16px (readable), so they were left as-is.
The floating-header → toolbar → filter-chips → content stack carried ~2× the whitespace it
needed. Halve it, app-wide where the chrome is shared:
- PageShell: the --hdr-h gap below the floating header 8px → 4px (every page's header).
- Feed: toolbar wrapper pt-4 → pt-2, the toolbar's own pb-3 → pb-1.5, and the grid's top pt-2
→ pt-1. Measured on the feed: header→toolbar 24→12px, toolbar→chips 12→6px, chips→cards 12→8px.
- Plex / Playlists / ChannelDiscovery / DownloadCenter / Notifications / AdminUsers: toolbar
wrapper pt-3 → pt-2, so every page's top band matches the tighter feed.
PageShell.dom.test updated for the 4px gap.
Rebuilds SortControl's internals (API unchanged, so all five consumers — feed, Plex,
playlists, and the two table/card ColumnSortControl sites — get it for free) as a labeled
"Label: key ↓ ▾" pill on the AnchoredPopover platform, matching FilterSelect. The listbox
lists the sort keys; an asc/desc segment sits in the panel footer (with the trailing slot —
feed reshuffle, playlists dirty-first — beside it).
Two-dimensional key+direction, so it stays its own component rather than reusing FilterSelect:
the trigger is a single plain button (no nested button), picking a key keeps the panel open so
key AND direction can be set in one visit, and picking "no sort" (tables' defaultLabel) closes
it. Directionless keys (shuffle) hide the segment and drop the trigger arrow.
Drops the now-redundant external "Sort" label spans in Feed and PlexBrowse (the trigger carries
the label now). Rewrites SortControl.dom.test for the popover UI.
Replaces the native <select> "Source" control with a labeled FilterSelect pill, matching
the Watch state / Content types pills — one consistent toolbar idiom. Single-select over
the three librarySource values (organic / all / search-only).
Sort and View are intentionally left as-is: Sort is a two-dimensional key+direction control
(plus reshuffle and a tables-only "no sort") shared across five surfaces, and View is already
on the same AnchoredPopover platform with an icon-only trigger tuned for this packed toolbar —
FilterSelect's text trigger would regress it. Neither fits FilterSelect's one-dimensional model.
Adds a reusable FilterSelect: a "Label: value ▾" pill that opens a listbox (a checkmark
row for single-select, a checkbox row for multi), on the AnchoredPopover platform. It
replaces the two feed-toolbar chip walls:
- Watch state (single-select, default All) — picking a row replaces + closes.
- Content types (multi-select, default Normal; requireOne keeps at least one on) — rows
toggle and the panel stays open.
The trigger summarises multi picks as "First +N". The component is source-agnostic, meant
for reuse anywhere a small known option set currently renders as a chip group.
FilterSelect takes a `testId` → the trigger gets it, each option gets `${testId}-${value}`,
so the feed/channel E2Es keep stable selectors. They now open the dropdown and assert the
option's aria-selected instead of the old inline aria-pressed chips (new expectContentType/
toggleContentType helpers).
Replaces the hand-rolled channel search list and the flat tag chip walls with the
searchable, ARIA-correct Combobox built for the Plex filter bar (E5), so the YT feed
filters get the same input+listbox+chips UX with facet counts.
- ChannelPicker now wraps Combobox (selectedLabels keeps picks named); drops its bespoke
input + scroll-fade list.
- Sidebar language/topic/tags: each Combobox owns only its slice of the shared
filters.tags list (setTagSelection replaces that list's ids, leaves the rest). Topic
keeps its any/all match toggle, Your tags keeps Manage. TagChip removed.
PageShell is reused (never remounted) across feed↔channel navigation — same
component position in App.tsx — so a one-shot useLayoutEffect kept its
ResizeObserver bound to the FIRST header node. On channel-open that node detached
and fired the observer with a 0 box, rewriting --hdr-h to 0+0+8=8px on the
still-mounted content div. Back restored the pt-(--hdr-h) pad class with that
stale value, sliding the content toolbar up under the header — stuck until reload.
Hold both the header and content nodes in state (callback refs) and key the
measurement effect to the live nodes, so it re-runs every time the header
mounts/unmounts; drop the stale pad when the header is absent (channel page).
Adds a regression test: header removed then restored must re-measure, not 8px.
Full-epic /deep-review (11 finders + 12 adversarial verifiers) surfaced no HIGH; every
mechanism finding calibrated to LOW on verification. Fix all 13:
Correctness / reliability:
- Combobox: Enter now commits the VISUALLY highlighted row (highlightIndex ?? activeIndex),
not the bare keyboard activeIndex — so arrow-then-hover-then-Enter no longer toggles a
different option than the one painted, and hover-with-no-arrow is no longer a silent no-op.
- parse_person: dedup the known_for pool by (id, media_type), not the bare id — TMDB movie
and tv id spaces are independent, so a same-id movie+show pair is two distinct credits
(matches the collision-safe key the rest of the module already uses).
- person-enrich: a per-person in-flight + 60s cooldown guard (people._should_schedule_enrich)
dedups concurrent opens and rate-limits re-triggering while TMDB is degraded (a transient
TmdbError leaves the watermark NULL), so bursts can't pile 8s tasks onto the shared threadpool.
- PlexInfo compact()/money(): format with i18n.language (keeping currency USD) instead of a
hardcoded en/en-US, matching the app's Intl convention.
- delete the dead `plex.filter.add` i18n key (both locales) — the +Add button that used it is gone.
Reuse:
- _continue_watching reuses rollup.agg_subquery(include_resumed=True) + query.apply_show_state_filter
instead of a third hand-inlined copy of the per-show rollup + in_progress predicate, so the
Discover rail can't drift from the grid's definition.
Test coverage (the gaps a regression would ship through):
- resolve_library_guids scope-exclusion (a credit in a non-scoped/disabled library is dropped)
+ the []-vs-None distinction; parse_person cross-media-id retention; enrich_person transient
(NULL watermark) and 404 (stamp) branches; _enrich_person_bg body (fresh session close/rollback,
None-guard) + the cooldown dedup; continue_watching fully-watched-show exclusion; Combobox
Enter-commits-highlighted + async/controlled-query mode; PopoverItem aria-checked contract;
RatingsRow missing-IMDb + imdb==audience branches.
Backend lane 413→420, FE vitest 351→357; all gates green. No migration.
The last of the deep-review deferred reuse items:
- extract the scroll-affordance machinery (watch the scroller AND its swappable/
lazily-growing children, diff two edge booleans) that useScrollFade (vertical) and
Carousel's useHScroll (horizontal) each hand-duplicated into one shared
lib/useScrollEdges(axis); both hooks now just map start/end to their own shape.
- migrate PlexInfo's two remaining hand-rolled `flex overflow-x-auto` strips — the
cast row and the collection strips — to the S1 <Carousel> primitive, so they get the
edge fade + hover/focus scroll buttons (only on the clipped side) for free, matching
the Discover rails / CastStrip / RelatedStrip.
Pure refactor, no behaviour change. Gate green (tsc, vitest 351, eslint, prettier,
knip). Browser-verified on a movie info page: cast row + collection strip render as
Carousels with the edge fade and the right-scroll button on the clipped cast row.
Address the deep-review's deferred backend items so E5 is review-complete:
- person-enrich OFF the request path (ABUSE finding): person_detail no longer makes
the TMDB call inline (even bounded to 8s it sat on the serve thread, un-throttled,
demo-reachable — it violated the module's "no live TMDB call at serve time"
invariant). It now schedules a FastAPI BackgroundTask (people._enrich_person_bg,
fresh session, best-effort) after the response; the first open serves the row as-is
and the bio lands by the next open. The inline try/rollback recovery is gone with it.
- reuse: a shared tmdb.resolve_library_guids owns the collision-safe guid→library
bucketing (movie vs tv ids are independent sequences) that resolve_recs and
people._resolve_known_for each hand-duplicated; a new rollup.item_state_map dedups
the "PlexState-by-item_id" fetch across the person + discover paths.
- test gaps: enrich_batch transient TmdbError leaves the watermark NULL to retry;
guid_to_tmdb_id resolves a tvdb source + tv media_type via /find's tv_results;
_continue_watching includes a NULL-status started movie and excludes a hidden one;
/api/search/youtube returns 422 (not 5xx) when ingest fails (route-level, B-core-1).
Backend lane 403 → 413, ruff clean. No migration; serve contracts unchanged.
The E5 filter pickers were a hand-rolled floating-ui combobox (MultiSelectChips)
with no combobox ARIA, no arrow-nav, and options in a body-portal — keyboard/SR
users could not operate it, and the bare role=listbox/option wiring actively
misled screen readers (deep-review A11Y-1, CONFIRMED 3/3 medium). Rather than
one-off-hack the a11y, add combobox as a first-class mode of the AnchoredPopover
platform and rebuild the pickers on a reusable component.
Platform (AnchoredPopover):
- new mode "combobox": reuses the whole positioning core (offset/flip/shift/size/
hide, isPositioned + transform-clobber flash guards, FloatingPortal + fullscreen
re-parent, topmost-only Escape layer). Differs only in the focus model — no
useClick (open is field-driven), useRole role="combobox" so Floating UI wires the
full ARIA tree (input role=combobox + aria-expanded/controls/autocomplete/
activedescendant, listbox surface, option ids), and useListNavigation virtual:true
so ↑/↓ move an aria-activedescendant pointer while DOM focus STAYS on the input.
- PopoverSurface: combobox surfaces host the item registry but are NOT focus-trapped.
- PopoverItem: virtual mode carries a stable id (the activedescendant target), never
takes roving DOM focus, keeps the input focused on click (mousedown preventDefault),
and its checked/selected truth wins over getItemProps' single-select guess.
Component (Combobox.tsx, new): the universal react-select/Overseerr combi-input —
chips-in-field + typed search + checklist. Supports multi/single select, client OR
server (async) filtering, custom rows, the big-vocabulary render cap + refine footer,
and a selected-label cache so a server-narrowed single-select still names its chip.
Rollout: MultiSelectChips (genre/age-rating/studio) → Combobox; the CollectionPicker
"+Add" popover → a single-select async Combobox (also fixes its arrow-keys-eaten
story), unifying the two filter idioms. PlexSearchBox is a documented future consumer.
Gate green (tsc, vitest 351, eslint, prettier, knip). Browser E2E on localdev:
role=combobox confirmed on both pickers; ↑/↓ moves aria-activedescendant with focus
staying on the input; Enter + click select; multi-select chips accumulate; Backspace
removes the last chip; async collection server-search picks + closes. No backend change.
Three verified deep-review findings (the 4th, the filter combobox a11y gap, is
deferred to a dedicated session on the combobox architecture):
- SEC (medium): TmdbClient sent the api_key as a URL query param, so an HTTP
error's httpx exception string embedded the full URL incl. the key, which
enrich_batch logged. _get now raises status+path only (never str(e)/the URL)
and severs the secret-bearing cause with `from None`. Regression test asserts
neither the status nor the network error path contains the key.
- STATE (low): person_detail's lazy-enrich except now db.rollback()s, so a
failed enrich_person commit can't poison the session and 500 the subsequent
SELECTs (SQLAlchemy 2.x deactivates the tx on a commit-time flush failure and
does not auto-reset it). Test locks that the page renders despite a poisoned
enrich.
- ABUSE (low): the request-path lazy person enrich now caps its TMDB client at
an 8s timeout (was the 30s default) so a slow TMDB can't tie up a worker;
TmdbClient gained a `timeout` param (batch keeps 30s). The fuller fix (move
person enrichment off the request path) is logged for later.
The combi-input popover flashed in the top-left corner for a frame, then
jumped into place. Two causes, both confirmed with a per-frame capture in
the browser:
1. The panel painted before floating-ui computed a position — gated its
visibility on `isPositioned` (hidden until positioned), same as
PopoverSurface.
2. The bigger one: the `popIn` entrance animation animates `transform`
(scale/translateY), and it sat on the SAME element floating-ui positions
via `transform` — so during the 0.16s animation the keyframes OVERRODE
the translate(x,y) and the panel rendered at the origin. Moved the glass
styling + popIn onto an inner child, leaving the positioned element's
transform untouched.
Frame capture after the fix: frame 0 hidden, first visible frame already at
translate(82,474) — never visible at the origin. Also gave the size
middleware a min width for a narrow rail.
The AnchoredPopover-based version had three UAT-caught bugs: the field
showed an ugly "Add" placeholder; the first click couldn't type (the
platform's FloatingFocusManager + focusItemOnOpen pulled focus onto the
list, blurring the field until a second click); and an outside click didn't
close the list. A visual pass also caught the dropdown rendering BEHIND the
sidebar panel (z-menu 20 < z-panel 35).
Rebuilt on floating-ui primitives directly — no focus manager, so focus
stays in the field; open is driven by field focus; useDismiss closes on
outside-press/Escape; the list is a z-popover (60) FloatingPortal sized to
the field width. Placeholder is now a plain "Search" (empty once chips
exist). Verified in the browser: type-on-first-click, filter, chip add,
outside-close, and the list painting above the sidebar.
User-reported issues after the E5 build:
- useCrossfadeBackdrop: the static branch (one image / crossfade off /
reduced motion) painted #root but returned NO cleanup, so the backdrop
persisted after leaving the page (e.g. switching modules). Every painted
branch now returns clear() so unmount always resets the default backdrop.
- scheduler: give the tmdb_enrich job a proper label + description (was the
raw "tmdb_enrich" id) in EN/HU.
- CastStrip: cap crew names to ONE line (truncate) so the row of circular
photos sits on one baseline instead of going wavy on 2-line names.
- PlexInfo: lay Details / Keywords / Crew out as side-by-side boxes
(items-start grid) instead of full-width stacked panels with dead space;
narrower Crew grid for the boxed width; each panel independently hideable
from the customize menu (new Details/Keywords/Crew toggles). Tighter
vertical spacing.
- MultiSelectChips: reshape the trigger from a "+Add" button into an
Overseerr-style combi input — removable chips inside a typeable field that
filters the checklist, opening on focus (backspace removes the last chip).
- Discover / person rails: reduce the vertical gap between rails.
Follow-ups from the epic-end /code-review of the E5 branch:
- tmdb.resolve_recs + people._resolve_known_for: match recommendations/
credits to the library bucketed by media_type. TMDB movie ids and tv ids
are independent sequences, so a bare `tmdb:<id>` match could deep-link a
movie rec to an unrelated show (or vice-versa) on a numeric collision.
Regression test added (test_resolve_recs_matches_by_media_type_not_bare_id).
- routes/search: wrap the _ingest_candidates call so a DB/enrich failure
degrades to a 422 with a message instead of a bare 500 hidden behind the
FE's generic "search failed" (extends the B-core-1 rationale); rollback on
failure. Stamp enriched_at with its own timestamp taken after the enrich,
not the stub-insert placeholder.
- routes/plex/discover: key serialized cards back by rating_key (globally
unique in Plex) instead of id(obj); fix the stale rows 3-tuple annotation.
The scrape source exposes no publish date, so its search stub carried
published_at=None. Migration 0068 (PERF-15, E2/0.68.0) made
videos.published_at NOT NULL, so insert_stubs raised IntegrityError —
outside the route's page-fetch try/except — and the whole
/api/search/youtube request 500'd ("YouTube search failed."). Every
scrape search (the default source) had been broken since 0.68.0.
- routes/search._ingest_candidates: stub published_at falls back to
now() (mirrors stub_from_playlist_item), satisfying the constraint.
- sync.videos.apply_video_details: prefer the enrich snippet's real
date whenever present, so the now() placeholder is corrected instead
of becoming the video's permanent date (the guard only filled a NULL,
which a non-null placeholder no longer is). Never clobbers a real
date with now().
Regression tests (DB lane): dateless scrape stub ingests and takes the
real enrich date; an enrich miss keeps a non-null placeholder, unkept.
The faint HTPC art backdrop now slowly cross-dissolves through a pool of images
instead of showing one static art — the last E5 sprint.
- useCrossfadeBackdrop (replaces useArtBackdrop): a double buffer — the SETTLED
image lives on #root's background (the proven full-viewport layer); the next
image cross-fades in on a z-index:-1 overlay layer, then is committed and the
overlay reset. ⭐ #root is given `isolation: isolate` so it's a stacking context
— without it the z:-1 overlay paints BEHIND #root's own background and is
invisible (found + fixed live: a red test layer + a real TMDB backdrop both
only showed once #root was isolated).
- Art pool: the item's own art + its TMDB backdrops (E5 S2, rendered straight
from the TMDB CDN — the CSP allows it) + its collection siblings' art; the show
page adds related art; the PERSON page cross-dissolves through the person's
films' art (the roadmap's signature touch).
- HARD-OFF under prefers-reduced-motion (static, no overlay) + a new usePref
"Crossfade backdrop" toggle in the detail customize menu (fade 1.6s / 7s cycle).
Wired into PlexInfo, PlexShowView, PlexSeasonView, PlexPersonView. Tests:
plexBackdrop.dom.test.tsx (5: paint+isolate+overlay, reduced-motion off, single
image, animate=false, disable/unmount cleanup). Browser-E2E'd (isolation +
overlay render + a real backdrop crossfade + the toggle). Gate: vitest 347,
backend lane 403 (unchanged — FE only).
▶ E5 · Plex Experience epic COMPLETE (S1–S8).
Reworks the always-expanded filter chip walls into compact, Seerr-style
"+ Add" → searchable-popover pickers (the user's space-saving ask), and adds
the missing studio filter.
- New MultiSelectChips: selected values as removable chips + a "+ Add" trigger
opening a searchable checklist (AnchoredPopover platform — menuitemcheckbox,
arrow/Escape, anchor flip). Options carry facet counts; a large vocabulary
renders a 50-row cap + "N more — refine with search" footer (studios run to
~1500). Replaces the flat Genre + Age-rating chip lists.
- New Studio filter via the same picker. Backend: /facets now aggregates the
studio vocabulary (distinct studio + count, self-excluded like genres). The
studio filter itself already existed (only addable from the detail page before).
- Collection converted to the same "+ Add" pattern (CollectionPicker): a
single-select popover with server-side search + the 50-cap; the selected
collection still shows as a removable chip. Retires the always-inline search+list.
New i18n plex.filter.studio/add/moreRefine + common.search/remove/noResults
(en+hu). Tests: MultiSelectChips.dom.test.tsx (5, incl. the cap) +
test_plex_facets studio-vocabulary. Browser-E2E'd: Genre→Horror(402),
Studio→A24(14, capped/searched), Collection→James Bond(26)/harry-search.
Gate: vitest 342, backend lane 403.
The person page now leads with a TMDB bio, lazily fetched the first time it's
opened (off the S2-linked `tmdb_person_id`) and then served from the row:
- Migration 0072: ADD-only nullable person columns on plex_people — biography,
birthday, deathday, place_of_birth, known_for_department, also_known_as,
known_for ([{tmdb_id, media_type, title}]), tmdb_enriched_at.
- tmdb.py: TmdbClient.fetch_person (/person/{id}+combined_credits), parse_person
(bio + a popularity-ranked, deduped known_for), apply_person, enrich_person
(lazy one-shot, key+id guarded, commits, 404 stamps the watermark).
- people.py person_detail: loads the full person row, lazily enriches (best-effort
— never fails the page), serves the bio fields + resolves known_for to library
rows by guid (in-library only, popularity order).
- FE PlexPersonView: bio/birth-facts/aka hero + a "Known for" Carousel rail above
the existing Movies/Shows filmography; PlexPersonDetail type + i18n (en+hu).
Tests: test_plex_person_bio.py (6: parse / lazy-enrich guards / serve) + extended
PlexPersonView.dom.test.tsx + the unknown-name contract in test_plex_person.py.
Browser-E2E'd on Jim Carrey (live TMDB enrich → bio 1937 chars, born 1962-01-17
Newmarket, Known-for resolved). Gate: vitest 337, backend lane 402.
Consumes the TMDB block the S2 engine already persists + serves (movie/show
detail's `tmdb`), turning the movie info page into a Seerr-style enriched detail
(page variant only — the player overlay stays lean):
- Multi-source ratings row: IMDb (links out) · TMDb (with vote count) · Plex
audience — whichever are present. Only the primary score (IMDb, else audience)
carries the ratingMin filter, mirroring the backend rating column.
- Facts panel: release · runtime · cert · budget · revenue · studios · countries.
- Keyword chips → tap to filter the library. Backend: a new `keywords` filter on
LibraryFilters + meta_filter_conds (`keywords @> [kw]`, both models), wired
through PlexFilters (appendPlexFilters / plexFilterCount / active-chip / clear).
- Full crew (TMDB key-crew) → tap a name to open their person page.
Tests: test_plex_keyword_filter.py (3) + PlexInfo.dom.test.tsx (5, incl. the
overlay-stays-lean guard). New i18n plex.info.facts.* / keywords / crew / audience
(en+hu). Gate: vitest 337, backend lane 396.
A dedicated Plex "Discover" landing — a stack of horizontal Carousel rails
served from ONE consolidated read (serve-from-DB, no live Plex/TMDB call):
- GET /api/plex/discover → continue_watching (in-progress movies + shows, most
recently progressed) · recently_added · genres (top-N tiles) ·
because_you_watched (recs of the most-recent watched, TMDB-enriched movie,
resolved to library rows by guid via serve_block's resolver).
- GET /api/plex/surprise → a fresh random, scope-bounded sample for the ↻ rail.
Frontend: PlexDiscover.tsx (rails reuse the S1 Carousel + S3 PlexPosterCard;
colored genre tiles → tap to filter the grid; ↻ reshuffles the surprise rail
alone). Discover is now the root sub-view; the flat grid becomes the bookmarkable
"Browse all" secondary view (?v=browse) reached by a header search/filter (a
browse-intent flip) or the new Discover⟷Browse toggle.
resolve_recs is promoted from _resolve_recs (now cross-module). Tests:
test_plex_discover.py (rail shape/order, continue-watching, because-you-watched
+ null, surprise bounds, scope) and PlexDiscover.dom.test.tsx; plexSubUrl
round-trip updated for discover-root/grid-browse. Backend lane 393, vitest 332.
The poster card (grid + person page) gets the Seerr-style hover: on hover a
bottom scrim slides up with the title, a "year · runtime · ★rating" meta line,
and a 3-line synopsis — the info you'd otherwise click through for — plus a
less-flat play ring (radial-gradient + ring shadow) and a top-right quick-action
cluster (Info + watched-toggle, revealed on hover / always on touch). Rest-state
badges (season / type tag) yield to the scrim on hover; the playable-cost tint
stays. (§1 A + C from the enrichment study.)
Backend (the data the scrim needs, mirroring S1's `art`): the grid
(browse.unified_library) now emits a lang-coalesced, truncated `summary` (new
query.summary_col) + the Plex `rating`; movie_card/show_card do the same (EN) for
the detail/related/person cards. PlexCard.rating type added (summary already
existed for episodes).
Gate: tsc x3 / eslint / prettier / knip / vitest 326 green; ruff green; backend
lane full suite green. Self-E2E on the live dev server (rebuilt on this branch,
S2-enriched): the grid serves rating+summary (Shawshank 9.3, Godfather 9.2, real
synopses) and the card renders the play ring + scrim (The Expanse: ring, synopsis,
★8.1) — verified via the DOM (the pane wouldn't composite for a screenshot).
A live smoke against the real TMDB API showed two crew problems the mocked
tests missed: (1) a TV show's creator/showrunner lives in the payload's
top-level `created_by`, not the crew list, so it was dropped entirely; and
(2) TMDB's crew order is arbitrary — a TV show's per-episode Directors (dozens)
flooded the key-crew list and buried the show-level roles.
Fix: split the accepted crew jobs by media type — movies key on
Director/Writer/…; TV keys on Creator (from created_by) + the show-level
Executive Producer/Writer/Producer/Composer, deliberately excluding the
per-episode Director/Editor/DP noise. Stored crew is sorted by display
priority and capped, so the lead roles come first.
Verified live: Breaking Bad -> [Gilligan (Creator), Johnson (EP), … , Writer];
The Godfather -> [Coppola (Director), Puzo (Screenplay), …]. test_plex_tmdb
green (9).
Matches each mirrored Plex movie/show to TMDB by its stored canonical guid,
fetches once (append_to_response), and PERSISTS the enrichment onto the row —
served from the DB, never a live TMDB call at serve time. Additive: the
Plex-sourced imdb_rating/rating are untouched; tmdb_rating is the distinct
TMDb community score (ratings scope = TMDb + Plex's IMDb/audience, no OMDb/RT).
- config.py: tmdb_api_key + api_base/language/interval/batch defaults.
- sysconfig SPECS: tmdb_api_key (secret, Fernet) + tmdb_language, auto-rendered
in the admin Config page; read like the Plex token.
- migration 0071 (ADD-only nullable, reversible): plex_items + plex_shows gain
tmdb_id / tmdb_enriched_at (watermark) / tmdb_rating / tmdb_vote_count /
budget / revenue / keywords / crew / studios / production_countries /
recommendations / backdrops; plex_people gains tmdb_person_id (the S2
foundation for S6 person bios). No watermark reset needed.
- app/plex/tmdb.py: TmdbClient (mirrors PlexClient; throttled; 404->None,
transient->TmdbError), parse_tmdb / apply_tmdb_detail (one-parser rule),
guid_to_tmdb_id (tmdb: direct, imdb:/tvdb: via /find), enrich_batch (movies
then shows, transient error leaves the watermark NULL to retry, definitive
no-match still stamps so it isn't re-fetched), _link_people (upserts
tmdb_person_id by name), serve_block (recommendations resolved to library
rows by guid), run_enrich (no-op without a key).
- scheduler: tmdb_enrich nightly job (default 120 min, no YouTube quota),
runtime-tunable + "run now" from the Scheduler dashboard.
- detail endpoints emit `tmdb` (movies + shows) so it's curl-verifiable now;
the FE render lands in S5.
Note: _link_people inserts plex_people rows for TMDB cast/crew; unreferenced
crew (composers etc., not in cast_names/directors) get pruned by the orphan
sweep — acceptable, the person page is cast-driven and cast rows survive.
Gate: ruff + py_compile green; backend test lane FULL suite 387 passed
(incl. migration drift downgrade->re-upgrade + test_plex_tmdb 9 tests).
Frontend unchanged since S1. Not merged (epic branch, stacks on S1).
Wave 5a building blocks under the Seerr-inspired Plex enrichment (E5).
Carousel (frontend/src/components/Carousel.tsx): the one horizontal-rail
component, replacing the hand-rolled `flex overflow-x-auto` strips and the
home of the future Plex-Discover rows. Edge-fade mask, hover/focus scroll
buttons that appear only on the clipped side (useHScroll — the X-axis twin
of useScrollFade, same watch-the-children discipline), native scrollbar
hidden behind .no-scrollbar, reduced-motion-aware paging. First consumers:
CastStrip + RelatedStrip now scroll inside it (dedup + free affordances).
New i18n: common.seeAll / scrollLeft / scrollRight (en+hu).
Card-art serializer (backend/app/plex/cards.py): movie_card + show_card now
emit `art` (the wide backdrop, /api/plex/image/{rk}?variant=art) so rails and
the crossfading hero get a landscape image with no per-item fetch — gated on
art_key so an art-less item stays null and falls back to the poster. FE type
PlexCard.art?.
Tests: Carousel.dom.test (children/header/see-all, clipped-edge buttons +
paging), test_plex_cards (art present/absent by art_key). Added a
ResizeObserver stub to PlexPersonView.dom.test (it renders CastStrip, now a
Carousel consumer).
Gate: tsc x3 / eslint / prettier / knip / vitest 324 green; ruff green;
backend test lane (docker-compose.test.yml) green on the plex card tests.
Closes every ▶ DEFERRED test-coverage item from the E4 deep-review (Modal
focus-trap, useHistorySubview popstate URL-fallback, PLEX_SUB_URL round-trip,
loadInitialPage precedence + channelViewFrom, readSelectedId ?pl guard,
useUrlMirror dep-keying, usePref flush-on-unmount) + the a11y sweep, and the
findings from two /code-review high rounds. Test-only + behaviour-preserving
seams (parseId/navigationInit/playlistsInit/plexSubUrl extractions); no
user-facing change, no version bump. vitest 257→321. Not deployed.
Closes the re-review's test-coverage finding: a plexCards.dom.test asserting the
icon-only card controls (watched-toggle, Info, add-season/episode) each expose
their aria-label as an accessible name via getByRole({name}). A future edit that
drops a label now fails the gate instead of silently reopening the unnamed-button
gap the a11y sweep closed.
4 tests; full FE gate green (tsc/eslint/prettier/knip, 321 vitest).
Resolves the 5 /code-review high findings on this branch:
- a11y sweep: extend aria-label to the remaining icon-only Plex card buttons the
first pass missed — PlexPosterCard Info, SeasonCard watched-toggle + add-season
(EpisodeCard already had them). "Fix a principle everywhere, in one sweep."
- new shared lib/parseId.ts `parsePositiveIntId`, used by BOTH readSelectedId
(now guarding the stored ?pl fallback the same as the URL — was unguarded) and
plexSubUrl's ?v=playlist decode; removes the hand-duplicated guard.
- PAGE_KEY exported from navigationInit and imported by NavigationProvider (was
defined in both).
- hoist a `watchedLabel` const per card so title + aria-label can't drift.
+8 tests (parseId unit + guarded-stored ?pl); full FE gate green (tsc/eslint/
prettier/knip, 317 vitest).
The hover watched-toggle on Plex movie/episode cards is icon-only (WatchedIcon,
no text), so it carried a `title` but no accessible name for screen readers —
and the empty-Circle unwatched glyph could read like a radio. Mirror the title
into `aria-label` (the Modal close-button idiom) so the control announces its
action. The PlexInfo watched buttons already have visible text labels, so they
need nothing.
FE gate green (tsc/eslint/prettier/knip, 309 vitest).
The remaining E4 deep-review deferred gaps, in one sprint:
- readSelectedId split into playlistsInit.ts (react-refresh-clean, testable) —
?pl accepted only as a positive integer; empty/0/negative/fractional/non-numeric
fall back to the stored selection; null when neither is present.
- useUrlMirror: mounts write the param, a same-serialized rerender does NOT
re-fire (stable dep), a null value clears the key, other params preserved.
- usePref: a pending debounced write is FLUSHED on unmount (not dropped); an
unmount with nothing pending stays silent.
20 new test cases; full FE gate green (tsc/eslint/prettier/knip, 309 vitest).
Split the two URL-derived nav seeds out of NavigationProvider into
navigationInit.ts so they're unit-testable and the provider stays
react-refresh-clean (component-only exports). channelViewFrom also dedups the
identical _chan/?chan resolve the channelView seed and the onPop handler each
hand-rolled — locking the S1 fix (a popstate on a URL-only bookmark entry must
restore the channel from ?chan, not close it). Pure code motion.
Tests: the full loadInitialPage precedence matrix (?page > ?v > ?pl >
filters/?chan > stored landing > feed, with the beats-a-lower-param cases) and
channelViewFrom (_chan+name, ?chan fallback, state-over-url, closed → null).
15 tests; full FE gate green (tsc/eslint/prettier/knip).
Move `type Sub` + `PLEX_SUB_URL` out of the 538-line PlexBrowse into a small
plexSubUrl.ts so the encode/decode round-trip and its guards can be unit-tested
without dragging the component (and its lazy hls.js player) into the test graph.
Pure code motion — PlexBrowse imports the two symbols; the `SubviewUrl` import
moves with them.
Tests lock: encode of each bookmarkable kind, grid/player → null, encode→decode
identity, a person name containing ':' round-tripping, and the decode guards
(positive-integer playlist id; season needs both ids; non-empty show/info/person
arg; unknown kind/garbage → null).
12 tests; full FE gate green (tsc/eslint/prettier/knip).
Backfills the E4 deep-review's deferred coverage for the subview history
primitive. The keeper case is the S2 bookmark fix: a popstate that lands on a
URL-only entry (no history.state._sub) must restore the sub-view from the ?v
param via the SAME fallback the mount seed uses, not collapse to root.
Covers: seed from URL vs state._sub vs root, the no-SubviewUrl state-only
path, open() push+mirror + encode→null key-clear, and popstate restore from
state / from URL (the fix) / collapse-to-root.
9 tests, gate green.
Backfills the E4 deep-review's deferred Modal focus-management test. jsdom
returns null for every offsetParent (no layout), so focusablesIn() would see
zero controls; the test stubs HTMLElement.prototype.offsetParent to report
"visible" for connected nodes (what a real browser reports for the shown
dialog), and a second describe documents the raw-jsdom empty-guard path.
Covers: initial focus on open, Tab/Shift+Tab edge-wrap, pull-back on escaped
focus, the active===panel leak guard, no-trap in the interior, Escape closes
the topmost only, and restore-to-trigger on unmount. useBackToClose is mocked
(its history stack is covered by history.dom.test.ts).
10 tests; full FE gate green (tsc/eslint/prettier/knip, 257 vitest).
The reveal-class swap in S6 shortened two className lines (VideoCard action row,
SavedViewsWidget default-star), which prettier wants collapsed to one line. The
local `npm run lint` doesn't run prettier --check, so it slipped past the sprint
gate and only the release `siftlode check` caught it. Formatting only, no behavior
change. Fixes the pre-publish gate for the already-merged 0.70.0 on main.
Bump VERSION to 0.70.0 and add the user-facing release-notes entry for the E4
Frontend Platform epic: full bookmarkable/shareable address-bar URLs (retiring the
"Share view" button), touch-visible card action buttons with a shape (not
colour-only) watched marker, and the bookmark-page-closes-on-modal-close fix — over
a large internal front-end refactor (god-component splits, usePref, useUrlMirror,
FilterRail, Modal focus). No DB migration.
Deep-review (6 dimensions, 3-verifier adversarial pass) surfaced 7 findings, all
verified; fixed here.
- NavigationProvider + history.ts (medium, correctness): the popstate handlers
restored the open channel / Plex sub-view from history.state ONLY, but a fresh
?chan / ?v bookmark lives in the URL with no marker in state — so any overlay's
programmatic history.back() (closing a modal/player) silently closed the
bookmarked channel or collapsed the sub-view to the grid. Give onPop the same
URL fallback the initializers already use (history.ts extracts a shared readView
so seed and popstate can't diverge again).
- plexCards SeasonCard (medium, a11y): the "In progress" season badge was a
reveal-idle swap-badge → display:none on touch, and a season has no progress bar,
so an in-progress season was indistinguishable from unwatched on touch. Moved it
to the top-left (the toggle owns top-right) and made it always visible.
- plexCards watched toggle (low, a11y / WCAG 1.4.1): watched state was color-only
on touch (green vs white CheckCircle2). New WatchedIcon distinguishes by SHAPE —
CheckCircle2 (watched) vs an empty Circle (not) — used by all three card toggles.
- Modal.tsx (low, a11y): the focus trap leaked on Shift+Tab when focus rested on
the panel container itself (Node.contains(self) is true, so the pull-back branch
never ran). Guard now also checks `active === panel`.
- index.css (low, a11y): .reveal-idle faded only on .group:hover though .reveal is
revealed on hover AND focus-within → the badge overlapped the control on keyboard
focus. Added the focus-within fade.
- PlexBrowse.tsx (low, correctness): the ?v=playlist decoder accepted a non-integer
id (Number.isFinite); tightened to Number.isInteger(n) && n > 0, matching the ?pl
guard.
Gate green (tsc, knip, lint 0-err, 247 vitest). Self-E2E (localdev): ?v=show and
?chan bookmarks survive a modal open+close (were collapsing/closing before); Modal
Shift+Tab from the panel stays inside; in-progress season badge visible on touch;
watched toggle shows an empty-circle-vs-check shape difference; 0 console errors.
Applies the /code-review high findings on the E4 branch:
- usePref: flush the pending debounced server write on unmount instead of
clearing it, so the last change isn't lost (and can't later be overwritten by
the stale server value via the adopt effect). [correctness]
- PlaylistsProvider.readSelectedId: accept ?pl only when it's a positive integer
— "?pl=" gave Number("")===0 (finite) and "?pl=1.5" a fractional id, both of
which selected a non-existent playlist. [correctness]
- usePref: extract mergePrefsCache() so useSavePref and usePref.set share the
one me-cache merge shape instead of duplicating it. [reuse]
- New lib/useUrlMirror: the shared replaceState(urlWithParams(...)) mirror
effect, so FeedFiltersProvider and PlaylistsProvider stop hand-rolling the
identical block (keyed on the serialized delta so it only fires on change).
[reuse]
- plexCards: swapBadgeReveal() helper for the repeated
reveal-idle/group-hover:opacity-0 swap-badge ternary. [simplification]
Gate green (tsc, knip, lint 0-err, 247 vitest). Self-E2E (localdev): feed filter
mirrors to ?show and clears on leaving the feed / re-mirrors on return via
useUrlMirror; ?pl= (empty) and ?pl=abc no longer select a bogus playlist (fall
back to the stored selection); Plex reveal-idle badges still display:none on
touch after the helper refactor; 0 console errors.
The three filter rails — feed Sidebar, PlexSidebar, PlaylistsRail — each rendered
the same SidePanel chrome (title bar, customize `editing` state, edit-toggle,
reset-to-layout-default, edit hint) around a PanelGroups body, differing only in
identity + filter content. Extracted that shell into components/FilterRail.tsx:
each rail now supplies only what differs (title/icon, active-filter count +
Clear-all action, its `items`, and optional content above/below the groups —
`top` for the feed's scope toggle, `children` for its tag manager). FilterRail
owns the `editing` state, so all three drop their duplicated useState + SidePanel
wiring.
MediaSource facade (lib/mediaSource.ts): a thin `MediaSourceKind = "feed" | "plex"`
naming the two browsable, filterable sources (a PanelId minus "playlists", which
is a rail but not a source). Already unified in practice — one SortControl, the
source-agnostic ActiveFilter[] chips, PanelId layouts, now the shared shell — so
this is formalization, not new capability: a single named seam for a future third
source or a cross-source feature. The two media-source rails annotate their `kind`
with it; PlaylistsRail passes the plain PanelId. Per the recon, the playlist-dialog
merge was deliberately NOT forced (it proved forced + low value). UX-3 control-vocab
sweep deliberately deferred — the rails already use the canonical primitives.
Behaviour-preserving. Gate green (tsc, knip, lint 0-err, 247 vitest). Self-E2E
(localdev): all three rails render via FilterRail — feed (scope toggle + panel
groups + Customize→editing with reset + 56 reorder affordances, exits clean), Plex
(filter groups + customize), Playlists (sections + customize); 0 JS console errors.
UX-7 — hover-only card actions were `opacity-0 group-hover:opacity-100`: fine on
a mouse, but invisible AND still tappable on a touch device (no hover). New
`.reveal` primitive (index.css) gates the hide on `@media (hover: hover)`, so a
touch device keeps the control visible; its partner `.reveal-idle` is the idle
badge a control replaces (watched/in-progress pill under the watched-toggle),
faded on hover and dropped on touch so the two don't collide at the same corner.
Migrated the true controls — feed card action toolbar (VideoCard), Plex poster
card watched-toggle/info/add across all card variants + their swap badges
(plexCards), message pop-out (Messages), edit-interval pencil (Scheduler), and
saved-view star/rename/delete (SavedViewsWidget). Pure play-scrim overlays (the
whole card is the tap target) stay hover-only by design.
FE-9 — Modal.tsx had aria-modal but no focus management. Adds: a Tab/Shift+Tab
focus trap (topmost layer only, wraps at the edges and pulls escaped focus back);
initial focus into the dialog unless an autoFocus child already placed it; focus
restore to the trigger on close (captured in a useState initializer during render,
before commit runs autoFocus and moves activeElement inside); and aria-labelledby
wiring the dialog to its title.
Gate green (tsc, knip, lint 0-err, 247 vitest). Self-E2E (localdev): FE-9 trap
both directions + pull-back + Escape-restore-to-trigger + aria-labelledby verified;
UX-7 feed toolbar opacity 0 on desktop / 1 under mobile (hover:none) emulation,
Plex reveal buttons opacity 1 + reveal-idle badges display:none on touch.
Pure code-motion, DownloadCenter 1088->359 L. Group by kind, mirroring the S4
plexCards/plexViews split:
- components/downloadRow.tsx (309 L): the user-list presentation + shared row
primitives — StatusPill, Thumb, IconBtn, SourceRef, ChannelLine, DownloadRow,
UsageBar (+ jobTitle/jobChannel/displayUrl helpers, status/phase constants).
- components/downloadModals.tsx (178 L): the per-item user modals — MetaEditModal
(rename/metadata) and ErrorModal (failure detail + retry).
- components/downloadAdmin.tsx (267 L): the admin "System" tab — AdminSystem plus
QuotaModal and QuotaUserPicker.
Every extracted piece is prop-driven with no shell-state reach-back, so the shell
just imports them back; the tab logic, mutations and modal state stay in
DownloadCenter. Behaviour-preserving.
Gate green (tsc first try, knip, lint 0-err, 247 vitest). Self-E2E (localdev):
Queue/Library/Shared/System tabs render, UsageBar, admin storage/footprint/job
list, MetaEditModal open/close, 0 console errors.
The native-controls-yield effect registered addEventListener("blur", onBlur) but its
cleanup removed "blur-sm" (a stray Tailwind-class find-replace artifact), so the blur
listener was never removed — leaking one per player open/close. Match the names.
Mirror the Plex player split (plexPlayerHooks.ts / plexPlayerControls.tsx). The shell
keeps owning the shared refs and passes them into each hook; the once-bound keyboard
handler + layer-stack registration stay in the shell (they orchestrate every cluster).
Behaviour-preserving code-motion — PlayerModal 1332 -> 890 lines.
New lib/playerModalHooks.ts:
- useYtVolume (per-account wheel/arrow volume, on shared playerVolume.ts)
- useNativeControlsYield (iframe-focus overlay yield)
- useYtBigView (F-maximise + our true-fullscreen; YT-specific, not Plex's useFullscreen)
- useYtProgress (player create + resume + 5s checkpoint + auto-watch + auto-advance)
New components/playerModalControls.tsx: PlayerTransportControls (moved as-is) + the
AutoMode/LoopMode types.
Self-E2E (localdev): play a feed video -> plays; transport bar (Auto/Loop); wheel volume
persists ({"volume":100}); F maximises + docked bar; exit big view; Close destroys the
player. Gate green: typecheck, knip, eslint (0 err), prettier, 247 vitest.
Noted a pre-existing bug (kept byte-identical here, flagged separately): the native-yield
effect removes a "blur-sm" listener but adds "blur" — the blur listener leaks.
The query string is now the source of truth for the current view, so every URL is
bookmarkable/shareable. Reverses the early minimalist-URL rule (user's call).
- urlState: urlWithParams (read-modify-write delta on the live query, each provider
owns its keys) + clearParams (strip only transient auth params) replace the whole-
query stripUrlParams; shareUrl deleted; filterParamUpdates helper.
- NavigationProvider: page + ?chan mirrored to the address bar (pushState URL arg;
history.state mechanics untouched); loadInitialPage resolves deep links
(?v->plex, ?pl->playlists, filter/?chan->feed, bare->stored landing).
- FeedFiltersProvider: feed filters mirrored to the URL, gated on page===feed.
- useHistorySubview: optional SubviewUrl mapping; PlexBrowse mirrors the sub-view to
?v=<kind>:<args> (show/season/info/person/playlist; grid+player stay ephemeral).
- PlaylistsProvider: selected playlist mirrored to ?pl on the Playlists page.
- App: the 5 stripUrlParams calls become targeted clearParams (link/oauth/admin/
access/verify/reset); the share-link hydrate effect removed.
- Retire the "Share view" feature: Sidebar + SavedViewsWidget buttons + i18n keys.
Browser tab title stays the short module/channel name (document.title untouched) —
the query never leaks into it, per the user's request.
Gate green: typecheck, knip, eslint (0 err), prettier, 247 vitest (urlState.dom.test
+9). No browser-E2E yet (most visual sprint — flagged for the epic-end review).
Fold the hand-rolled server-pref plumbing (savePrefs + local mirror + me-cache
merge + adopt-on-login) into a single usePref<T>(serverKey, default, opts) hook,
plus useSavePref for dynamic multi-key patches.
Migrated (6 savePrefs call sites + 2 duplicated adopt effects -> the hook):
- FeedViewProvider: view
- App: filterCollapsed, playlistsCollapsed (+ their adopt block removed)
- PlayerModal: playerAutoAdvance, playerLoop
- plexDetailUi: plexInfoArtBg/Cast/Related (savePref via useSavePref)
Left as-is by design: panel layouts (a dynamically-indexed Record, not one key),
language (i18n store, no React state), PrefsProvider (debounced multi-field
auto-save + baseline echo-suppression - a different model).
Behaviour-equivalent (net -84 LOC at the call sites). usePref.dom.test covers
the three seed sources, persist path, debounce, adopt, parse, and useSavePref.
Frontend gate green: typecheck, knip, eslint (0 err), prettier, 238 vitest.
An <img> can't send the X-Siftlode-Account header, so the poster.jpg request
resolved to the session-default account and _accessible_job 404'd it for any
download owned by a different (per-tab) account — every card then fell back to
the source's remote thumbnail, which for Facebook is an expiring fbcdn URL, so
the Library showed blank thumbnails in a multi-account session. The 0.69.1
local-poster-primary change made this the primary image, exposing it broadly.
Append ?account= to the poster URL via a new api.posterImageUrl(), the same
wallet-gated header-less selection downloadFileUrl / storyboardImageUrl already
use (resolved_user_id reads ?account= for exactly these contexts).
Three prod bug fixes: local-poster-primary Library thumbnails, viewport-scaled
feed read-ahead (fixes the fast-scroll "Load more" stall), and the channel About
popover fitting/blank-line cleanup.
Review follow-up. The docstring said "channel/video description" but the helper is
wired only to the channel store sites; clarify it is deliberately channel-only
(video descriptions are user-facing prose where paragraph breaks carry meaning,
unlike a channel's metadata block) so it doesn't read as an asymmetric fix. Also
document that str.splitlines() folds CR/CRLF and Unicode line separators to "\n",
and lock it with a test.
Review follow-up. The viewport-scaled lookahead was unbounded, so a tall window
prefetched several feed pages on every mount / filter change before any scroll —
wasted work on the app's heaviest query. Clamp it to MAX_LOOKAHEAD_ROWS, and move
the policy into a pure readAhead.ts (readAheadLookahead / shouldPrefetchNext) with
unit tests covering the floor, the scaling, the cap boundary, and the fetch trigger
— the core of the feed-scroll fix now has coverage instead of only manual E2E.
Keyset pagination is strictly serial (a page needs the previous page's cursor)
and the read-ahead only fired within a fixed 4 rows of the loaded end, pulling
one 60-item page per cycle. A fast free-spin flick (MX Master & co.) out-scrolls
that tiny lookahead, lands on an empty bottom, and waits out a chain of serial
round-trips — the multi-second "Load more" hang. Scale the lookahead to the
rendered row count (PREFETCH_VIEWPORTS screenfuls); since the effect re-runs when
a page settles it chains, keeping a couple of screens buffered ahead of the fold
so the content is usually already there. Bounded, and a no-op for finite lists.
YouTube channel descriptions pad their contact/link blocks with runs of blank
lines, which the About popover renders verbatim (whitespace-pre-wrap) — wasted
space and storage, and noise in the search_vector. Add normalize_description()
and apply it at both channel store sites (full details + initial import) so the
DB holds the clean text. New channels are clean immediately; existing rows heal
on their next metadata re-sync. Unit-tested in test_utils.
At elevated zoom the CSS viewport shrinks; a long channel About wanted its fixed
max-h-112 (448px) height, so when floating-ui flipped it above the trigger its
top rendered above the viewport edge, clipped and unreachable (the inner
overflow-y-auto can't scroll past its own top). Enable the popover's fitHeight so
the `size` middleware caps the panel to the available space. Because the surface
splits into an outer positioned element and an inner styled/scroll wrapper, the
middleware now also publishes the available height as a --fit-max-h custom
property the inner scroller consumes via maxHeight: min(28rem, var(--fit-max-h)),
keeping the 28rem design cap on a tall viewport and shrinking to fit on a short one.
The Library card used the source's remote thumbnail_url as the primary <img>
src and only fell back to our self-hosted poster on error. For Facebook
downloads (the bulk of a typical library) that URL is an fbcdn signed link that
expires within days, and is also blocked by ad-blockers / Brave Shields on the
facebook domain — so the thumbnail broke even though the local poster was
present on disk. Flip the priority: serve /api/downloads/{id}/poster.jpg first
(always present, never expires, same-origin) and keep the remote as the fallback
for the rare poster-less asset.
Round-3 review of the S3 config-fold, all in routes/scheduler.py:
- stale docstring: "Persisted to app_state" -> "Persisted via sysconfig" (the
batch moved stores in B-core-11; the comment didn't follow).
- drop the manual min/max check in the PATCH — sysconfig.set_value already
validates against the same spec bounds; catch its ValueError -> 400, matching
the config PATCH pattern (bounds enforced in one place).
- fetch sysconfig.spec(...) once into a local in the GET instead of twice.
No behaviour change beyond the 400 message wording (untested path). 368 pass.
Restores CLAUDE.md's canonical transaction rule ("the handler commits; helpers
flush; only quota owns its session") on the config path: sysconfig.set_value /
sysconfig.reset and state.set_sync_paused now flush instead of committing, so a
change lands in ONE transaction with the audit row the route records for it —
previously the helper's internal commit split them, and a crash between left the
mutation persisted with no audit trail.
All callers already commit (the config PATCH/DELETE + audit, the sync pause/resume
+ audit, the scheduler maintenance PATCH); the setup wizard, which relied on the
internal commit, gets an explicit db.commit() after its loop. apply_interval is
deliberately left committing — it performs a live APScheduler reschedule that must
follow the DB persist, so it owns its own unit of work (a legitimate exception,
like the download service transitions). New test locks the flush-not-commit
contract (a rollback before the caller commits discards the change). 368 pass.
The admin-tunable maintenance re-validation batch lived in its own app_state
column with a hand-rolled override-or-default reader in app.state — a fourth
config store duplicating the sysconfig SPECS pattern. Move it to a sysconfig key
(`maintenance` group, min=100 max=100_000): the reader (sync/maintenance.py) and
the Scheduler dashboard (routes/scheduler.py PATCH + GET) now go through
sysconfig.get_int / set_value / spec, and app.state loses get/set_maintenance_batch
+ the MIN/MAX constants + the settings import. Migration 0070 carries any existing
override into the system_config store before dropping the app_state column
(reversible; restores + copies back on downgrade). Reuses the existing
sysconfig.spec() for the dashboard's min/max, so the bounds have one source.
Full lane 365 pass (migration drift + reversibility exercised), ruff clean.
None are referenced by the SPA or tests; the Scheduler dashboard's "run now"
covers every manual trigger. Removed: POST /api/sync/{rss,backfill,enrich}
(+ the now-dead _user_channels helper and three sync.runner imports),
POST /api/tags/recompute (+ its admin_user / run_autotag_all imports), and
GET /auth/me (a lesser duplicate of the canonical /api/me in routes/me.py).
This also retires a latent bug: POST /api/sync/backfill reset backfill_done /
backfill_cursor on already-fully-backfilled channels (tipping them back to
"needs full history") for up to 25 channels per curl. Full lane 365 pass.
This codebase leans on "why" comments as its architecture doc, so a header that
claims the opposite of the code below it is worse than none. Fix the seven the
review flagged: channels.py (subscribe/unsubscribe shipped), playlists.py +
sync/playlists.py (YouTube two-way sync + write-back shipped), setup.py (drop the
"epic 6c" ref — the wizard steps exist), config.py ("auth._client_ip" ->
"routes.auth._client_ip"), db.py ("routes/plex.py" -> "routes/plex/" after the
package split), sysconfig.py (the "Phase C" note — the watch-sync job that reads
plex_watch_sync_interval_min has landed; clarified it's a boot default, not a knob).
Docs only.
Round-2 review of the R1 fixes found that the two-independent-limiter progress
writer issued a redundant heartbeat every interval during an ACTIVE download: a
percent change writes progress (which already refreshes updated_at, the stall
watchdog's liveness source), then the sub-integer calls that follow fire the
separate beat budget — ~2 DB writes/interval instead of the throttle's intended 1.
Replace the second limiter with a shared `last_touch` clock (last DB write OR
beat). Progress writes keep their own `allow_write` budget (so a beat still can't
delay a real write — the R1 concern), but the heartbeat now fires only when
nothing touched the row for a full interval. Active download → ~1 write/interval;
a genuinely stuck re-encode → 1 heartbeat/interval. New regression test:
an unchanged-percent call in the same interval as a write does NOT heartbeat.
ruff clean, full DB-lane 365 pass.
Round-1 review fixes on the E3 branch:
- prune partial-fetch (correctness): `_paginate` now reports whether it reached
the server's reported `total`; `_prune_stale` prunes only on a COMPLETE
non-empty listing. A truncated fetch (a mid-sequence empty/short page before
total — a transient Plex hiccup) no longer deletes the rows it merely failed
to list. New test: a listing that claims 5 but returns 1 prunes nothing.
- `_prune_stale` (robustness): `stats.get("pruned", 0) + 1` — no KeyError when a
caller's stats dict omits the key.
- service transitions (conventions): pause/resume/redownload use the JobStatus
StrEnum instead of loose "queued"/"paused"/"error" literals, matching cancel.
- ffmpeg progress writer (correctness): a stuck-percent heartbeat now runs on its
OWN 1/interval budget, so it can't consume the progress-write slot and delay a
real percent change by up to an interval.
- backend/app/README.md (conventions): reword the transaction-ownership rule to
quote and extend CLAUDE.md's canonical ("handler/job-wrapper commits; helpers
flush; only quota owns its session") instead of stating a contradicting
"helpers MAY commit" carve-out.
ruff clean, full DB-lane 364 pass.
The layering rule ("request-only logic in routes/; scheduler+worker-shared logic
in the plex|downloads|sync packages; cross-cutting primitives as flat core
modules; extract at the SECOND consumer") was real but unwritten — it lived only
in three module docstrings and a one-line README mention. Collect it, plus the
transaction-ownership rule (a mid-operation helper stages, the caller commits),
the atomic-counter / JobStatus-StrEnum conventions, and the one-home-per-config
rule, into backend/app/README.md. Add docs/development.md (run locally + the gate
commands + branching), so a new session/contributor has one place to look instead
of copying whichever pattern they happen to find first.
worker.py (~1090 L) mixed the PARENT supervisor (claim + supervise + the process
loop) with the CHILD pipeline (the yt-dlp/ffmpeg download+edit that runs in the
spawned subprocess). Extract them:
worker.py 1090 -> 475 parent: claim/supervise/loop
downloads/pipeline.py (new 592) the child pipeline (_download/_process_edit +
hooks/progress/ffmpeg helpers + _Aborted)
downloads/jobdb.py (new 73) the short-lived-session helpers both sides
share (_set_job/_job_status/_touch_job/
_requeue_if_running/_reset_asset_pending +
the staging-dir paths)
The graph is one-directional (runner -> pipeline -> jobdb; worker -> jobdb; no
worker<->pipeline edge) — the parent never calls the child directly, it spawns
runner.py as a subprocess. Consumers rewired: runner import, and two tests follow
the moved symbols (_media_info/_produced_file -> pipeline; the ffmpeg-callback
patches target pipeline where the names are looked up).
Pure code motion, verbatim bodies. ruff clean, full lane 363 pass (unchanged).
request-access / demo / register / verify / verify-resend / password-login /
password-reset(request|confirm) / set-password each took an untyped `payload:
dict` and hand-picked fields with `.get(...)`. Replace with five small BaseModels
(EmailIn, EmailPasswordIn, TokenIn, TokenPasswordIn, SetPasswordIn) so FastAPI
validates the body shape and publishes the schema. Fields are `str | None` to
preserve exact behaviour — a missing/null field still coerces via the existing
`(... or "")`, no stricter 422s. Green under the auth-session suite (7), ruff clean.
pause/resume/cancel/delete/redownload each hand-rolled their asset bookkeeping
inline in the route (release, failed-asset reset, the 20-line fresh-asset
rebuild), duplicating logic the worker/GC also touch. Move each transition into a
service.* method (service.pause/resume/cancel/delete/redownload) that owns the
mutation + its commit; the routes keep only auth, the HTTP 409 guards, and
serialization. Drops the now-dead _set_status route helper and two stale imports.
Pure behaviour-preserving code motion (routes -56 lines). Green under the E3 S2
transition net + the full downloads suite (363 pass), ruff clean.
Pins the observable HTTP behaviour of pause/resume/cancel/delete/redownload
(status changes, asset release, failed-asset reset, fresh-asset rebuild) before
the inline route logic moves into service.* methods. Green on the current routes.
The supervisor kills a job that writes nothing for download_stall_timeout_seconds
(150s) and labels it a "network" stall. But the ffmpeg progress writer only wrote
on an integer-percent CHANGE, so a re-encode that sits on one percent for minutes
(AV1/VP9 -> H.264 or a crop on the CPU-only prod host) produced no row write and
was killed mid-encode as "stalled" — and the retry stalled the same way.
Fix: the writer now heartbeats (`_touch_job`, bumping updated_at without touching
the progress bar) once per interval even when the whole percent is unchanged, so
a live-but-quiet re-encode keeps a fresh updated_at. Reorders the rate-limit gate
first so a burst still costs one DB op — either a progress write or a heartbeat.
Pure-logic test with an injected clock: three stuck-at-10% intervals write no new
progress but heartbeat each time. Full lane green, ruff clean.
Two jobs on the same shared asset (same video+format queued twice, or by two
users) starved the worker pool: worker A downloads the asset; worker B claims the
sibling job, finds the asset `downloading`, requeues it, sleep(2)s, and re-claims
it on the next tick — burning a whole worker slot behind one in-flight asset for
the entire download (at concurrency=2, the queue stalls).
Fix: the claim SQL now excludes any job whose asset is already `downloading`
(`NOT EXISTS … media_assets.status='downloading'`), so a competing sibling simply
waits until the asset settles and the workers pick up unrelated jobs meanwhile.
The `_process_job` busy/lost-race branches drop their `time.sleep(2)` — the
exclusion keeps a requeued job out of the claim window, so there's no spin.
TDD: the new test failed on the old claim (claimed the sibling), passes now.
Full lane green, ruff clean.
MediaAsset.ref_count (how many jobs hold a shared cache asset) was a Python
read-modify-write at four sites (enqueue, enqueue_edit, release_asset, the
redownload route). Two concurrent enqueues of the same (source,format) both read
the same value and write the same +1 → the count ends up one short → the first
release drives it to zero and deletes the file out from under a live holder.
Replace all four with atomic DB ops: incr_ref_count (`ref_count = ref_count + 1`)
and decr_ref_count (`greatest(ref_count - 1, 0)` RETURNING the new value, so the
release decides whether to free the asset from the post-decrement count while
holding the row's write lock — Postgres serialises concurrent releases on that
lock, closing the double-free). Pure DB-atomicity change, no semantic shift.
Green under the E3 S2 ref_count net (5) + full lane (356 pass), ruff clean.
Locks the observable single-threaded ref_count accounting before the atomic-SQL
conversion: two enqueues of one (source,format) share a single asset and stack
ref_count to 2; release keeps the asset until the last holder then frees the
ready row; a non-holding (canceled) job's release is a no-op; the decrement
clamps at 0; an errored asset is kept at zero (only ready is freed). Green on the
current RMW code (5 pass) — the fix must keep every assertion.
A sync was upsert-only: a movie/show/season/episode removed from Plex stayed
browsable forever and 404'd on playback. Each section now diffs the rating_keys
it just listed against the stored ones and deletes the stale rows (watch-state /
playlist links cascade via the existing FKs), mirroring the collections prune
that already shipped.
Safety: the prune is GUARDED on a non-empty listing (_prune_stale returns early
when `seen` is empty), so a transient Plex glitch that yields an empty page can
never mass-delete a library — the metadata is cheap to re-sync, an accidental
wipe is the only real risk. Shows prune leaf-first (episodes -> seasons -> shows)
so a stale row is deleted directly before its parent's cascade reaches it.
Out of scope (separate follow-up): library-level reconcile of a whole removed
section + retiring the dead `PlexLibrary.enabled` column.
Tests: movie prune + watch-state cascade, the empty-listing guard (mutation-
verified — removing the guard wipes everything), show/season/episode prune.
351 DB-lane pass, ruff clean.
routes/plex/browse.py was a 1693-line god-module (libraries + facets +
collections + playlists + show/item detail + people). Split it into focused
sibling routers under the same /api/plex prefix — pure code motion, no SQL /
route-path / response-shape changes:
browse.py 1693 -> 469 (list_libraries, unified_library, facets)
collections.py (new 253) browse + admin collection editor; owns _collection_source
playlists.py (new 405) the user's orderable playlists
detail.py (new 438) show/item detail + enrichment; _lib_key lives here
people.py (new 139) cast/crew chips + person filmography
Cross-module helpers _norm_lang + _content_libs move to _shared.py (their one
home); detail imports _collection_source from collections (one-way, no cycle).
__init__.py includes the four new routers. Three DB-lane tests follow the moved
symbols (import path for _content_libs; monkeypatch targets for PlexClient /
_show_enrichment now on detail).
Behaviour-equivalent: ruff clean, 348 DB-lane pass (unchanged), parity net green.
The `show` scope (all|watched|in_progress|unwatched) filter and the status CASE
were hand-duplicated between unified_library (the grid) and facets.wsq — the
exact drift risk the review flagged (a hidden movie kept out of the grid must
not inflate facets). Extract them into four shared query.py helpers
(movie_status_case, show_status_case, apply_movie_state_filter,
apply_show_state_filter), model-agnostic (callers pass the joined state
expressions), and point all three call sites at them.
Pure behaviour-equivalence refactor: browse.py -67 lines, no query changes.
Green under the E3 S1 parity net + query_conversion/facets/field-scoped/lang
tests (348 DB-lane pass, ruff clean).
Coverage-first safety net for the E3 S1 Plex god-module split. The movie/show
watch-state predicates are currently hand-duplicated in unified_library (the
grid) and facets.wsq (the facet aggregation); this pins the invariant the
facets docstring promises — with no metadata filter active, a facet's genre /
content-rating counts equal the same counts over the grid's visible items — for
every (scope × show) combination, deriving the expected counts from the grid's
own response. Locks the two branch-sets together before they merge into one
shared helper, so the merge is provably behaviour-equivalent.
Mutation-verified: dropping the facets hidden-exclusion branch makes both tests
fail (Drama 3≠2, R 2≠1). Green on the current code (348 DB-lane pass).
Bump VERSION 0.67.0 -> 0.68.0 and add the user-facing release-notes entry for
the E2 epic (feed sort index, keyset/NOT NULL, search GIN + state inner-joins,
single /feed/summary pass, no card-action refetch storm, job partial indexes).
The via_search / shorts_probed partials used `WHERE via_search` / `WHERE NOT
shorts_probed`, but the queries filter with `.is_(True)`/`.is_(False)` (SQL
`IS TRUE`/`IS FALSE`), which Postgres's partial-index prover does NOT match to a
bare-boolean predicate — so both fell back to a full seq scan (the shorts-probe
job kept full-scanning for its work queue, the exact thing PERF-3 meant to fix).
Align the predicates to the query form: `WHERE via_search IS TRUE` and
`WHERE shorts_probed IS FALSE`. Verified on the prod-sized localdev: both now
Index Only Scan (via_search 44,683 -> 3 buffers; shorts_probed 44,683 -> 1).
Gate green (346 pytest).
Two scheduler jobs full-seq-scanned the 257k-row videos table every run:
- PERF-17 enrich_pending (`WHERE enriched_at IS NULL LIMIT N`, every few minutes):
44,680 buffers even with 0 pending -> a partial ix_videos_enrich_pending makes
it a 1-buffer index scan (empty when nothing pending).
- PERF-18 maintenance revalidate (`WHERE unavailable_since IS NULL ORDER BY
last_checked_at NULLS FIRST LIMIT N`, ~300 pages/run): 44,754 buffers + top-N
sort -> a partial ix_videos_recheck_order in that exact order = a 42-buffer
index scan. The order_by is rewritten to `asc().nulls_first()` (equivalent to
the old `(IS NULL) DESC, ASC`) so it rides the index; the unused plain
ix_videos_last_checked_at is dropped.
PERF-3: drop the never-picked boolean btrees (is_short, via_explore) and slim the
useful ones to partials over the queried subset (via_search=true, NOT
shorts_probed, unavailable_since IS NOT NULL) — less write amplification on the
biggest table, same lookups. Model: those 6 columns lose their inline index=True
(the partials are migration-defined, like the search_vector GIN); live_status
left alone.
Measured on the prod-sized localdev reseed: enrich 44,680 -> 1 buffer (25.9 ->
0.02 ms), maintenance 44,754 -> 42 buffers (35.5 -> 0.17 ms); all 11 CONCURRENTLY
index ops on 257k rows in 0.75s. Gate green (346 pytest incl. drift + downgrade
round-trip).
The tags_by_category extraction dropped feed_summary's `if selected:` guard, so
a /feed/summary request with no tag filter (the common case) ran a pointless
`SELECT ... WHERE id IN ()` round-trip. Early-return {} for empty tag_ids —
restores the no-query hot path and covers filtered_query + future callers too.
Gate green (346 pytest).
Review round on the E2 stack (S1-S5). No correctness bugs; the surviving
low/medium findings addressed:
- altitude (#1): extract the shared `tags_by_category` helper used by BOTH
filtered_query (SQL tag WHERE) and feed_summary (Python facet/count), removing
the category-grouping duplication, and add bidirectional "keep in lockstep —
equivalence-tested by test_feed_summary_matches_count_plus_facets" comments so
the SQL<->Python tag rule can't silently drift.
- test-coverage (#2): the /feed/summary equivalence test seeded only system tags,
leaving the `ChannelTag.user_id == user.id` visibility branch untested. Add a
user-owned Tag + link + two combos; the suite is now 346 (was 344) and the new
cases pass (feed_summary agrees with /feed/count + /facets on user-visible tags).
- dead-code (#3): annotate get_feed_count, get_facets, and the FE api.feedCount/
facets as "SUPERSEDED by /feed/summary — kept only as the equivalence oracle;
retire together" so the intentionally-retained surface reads as deliberate.
Not changed (confirmed no-op): the search channel-name resolve (#4) is
once-per-request on a 633-row table and a deliberate pre-resolve for PERF-12's
BitmapOr plan; the facet loop (#5) already computes passes() once per
(channel, category) with a distinct exclude each, so there is no recomputation.
Gate green: ruff + 346 pytest; frontend tsc/knip/prettier (FE change is a comment).
Marking a card watched/hidden/reset and closing the player each fired a full
`invalidateQueries(qk.feed())`, refetching EVERY loaded infinite-query page — 10
pages x a few quick marks = ~50 feed refetches, each a full scan (x PERF-1). The
optimistic `overrides` layer + the matchesView filter already render the correct
state (a hidden/watched card drops out instantly), so the refetch was pure waste.
Switch these paths to `refetchType: "none"`: the queries are marked stale but not
refetched now; the next filter change / remount pulls fresh server data. Applies
to the two Feed card-action paths (setState/clearState) and the PlayerModal
close (feed + playlists — progress is already persisted server-side, only the
on-screen resume bar waits for the next natural refetch).
Left as immediate refetch on purpose: channel/tag/playlist mutations (one-off, on
pages where the main feed is unmounted anyway) and the onboarding / notification-
undo paths (which must reveal the changed content right away). Gate green:
frontend tsc/eslint/knip/prettier/222 vitest (backend unchanged, 344 pytest).
The FE fired /feed/count (1 pass) + /facets (3 passes) on every filter change =
4 full scans of the subscribed set. New /api/feed/summary does ONE grouped query
(the base filter with tags stripped, GROUP BY channel_id -> per-channel video
counts) and derives the total count + every tag's distinct-channel facet count in
Python on the ~few-hundred-row channel map, faithfully mirroring filtered_query's
channel-based tag logic + the /facets drill-down (own category dropped, unless a
conjunctive topic-AND). Returns {count, facets} — the same numbers the two old
endpoints returned.
FE: a single shared query (Feed reads .count, Sidebar reads .facets); same key +
filters, so React Query dedupes to ONE request. Invalidations repointed to
qk.feedSummary. The old /feed/count + /facets endpoints (and their now-unused FE
api/qk) are kept as the equivalence-test oracle + safety net; a later cleanup can
drop them.
New equivalence test (test_feed_views, 8 filter combos incl. cross-category,
disjunctive/conjunctive topic modes, search): /feed/summary count == /feed/count
and facets == /facets, byte-identical. Structural win: 2 requests / 4-6 queries
-> 1 request / 2-3 queries (the review measured 530 -> 144 ms). Gate green:
344 pytest, frontend tsc/eslint/knip/prettier/222 vitest.
PERF-12: the search ORed `unaccent(channels.title) ILIKE '%q%'` straight into
the per-video predicate, which forced a full scan + detoast of every
search_vector and a per-candidate channel fetch. Pre-resolve the matching
channel ids (a few-hundred-row scan) and match `Video.channel_id.in_(ids)`
instead, so the planner keeps the search_vector GIN as a BitmapOr branch
(alongside ix_videos_channel_pub_desc). Empty match -> false() collapses the OR.
PERF-16: show=watched|hidden|in_progress used an outer join + coalesce(status,
'new'), so the planner couldn't start from the user's small video_states set.
Those views only return rows that HAVE a state, so inner-join video_states and
filter the raw column (status is NOT NULL on a present row, so it's identical).
'unwatched'/'all' keep the outer join. Dropped the now-dead position_expr.
New regression tests (test_feed_views): search returns both text- and channel-
name matches; the four watch-state views return the exact expected sets.
Measured on the prod-sized localdev reseed (prod planner, 0067/0068 applied):
- search /feed/count q=magyar: 663,463 -> 19,206 buffers (1,406 -> 31 ms; OLD
seq-scan+detoast -> NEW BitmapOr GIN+channel index)
- search feed page: 2,908 -> 1,479 buffers (32.6 -> 1.4 ms)
- show=watched (user with 64 watched): 40,661 -> 264 buffers (~154x)
Gate green (336 pytest, incl. the 2 new equivalence tests).
A nullable published_at forced the feed keyset into a non-sargable
`... OR published_at IS NULL` disjunction: with the 0067 index the planner still
applied it as a FILTER and scanned from the top of the index, so a deep `newest`
page discarded every earlier row (O(depth)).
- Migration 0068: backfill any NULLs (0 in prod) -> SET NOT NULL + DEFAULT now().
- Model: Video.published_at is now non-Optional, nullable=False, server_default now().
- feedquery: the `pub` key is nullable=False, and keyset_predicate is rebuilt
recursively so a non-nullable leading key gets a non-strict range bound
(`published_at <= v`) as an AND conjunct — an index range Cond instead of a
filter. Nullable keys keep the NULL-aware OR form (a `<= v` bound would drop
their NULL group). Equivalence proven by test_feed_keyset (all 13 sorts,
cursor-walk == unpaged).
- Ingest paths (stub_from_playlist_item, apply_video_details, playlists sync)
fall back to now() so they never write an explicit NULL.
Measured on the prod-sized localdev reseed at a ~50k-deep cursor (0067 index +
prod planner): OLD `Rows Removed by Filter: 50001`, 25,687 buffers / 9.16 ms ->
NEW `Index Cond: published_at <= v`, 1 row filtered, 47 buffers / 0.037 ms. SET
NOT NULL on 257k rows: 0.32s. Gate green (334 pytest).
The feed's default `newest` sort is `published_at DESC NULLS LAST, id ASC`, but
the only published_at index was a plain ASC btree — read backwards it satisfies
DESC NULLS FIRST, not NULLS LAST — so every feed page fell back to a full
`videos` seq scan + top-N sort. Migration 0067 adds the two matching indexes
(global + per-channel) CONCURRENTLY and drops the now-redundant single-column
ix_videos_channel_id (the channel composite leads with channel_id, a strict
superset).
Measured on the prod-sized localdev reseed (257,312 videos, prod planner):
- global newest: 44,747 buffers / 154 ms -> 60 buffers / 0.32 ms
- per-channel(12.6k): 7,604 buffers / 96 ms -> 54 buffers / 0.14 ms
CONCURRENTLY build of both on 257k rows: 1.7s total (~0 downtime). This is the
likely root cause of the "prod feed is slow" symptom.
Model: drop the inline index=True on Video.channel_id (the composite supersedes
it); the composites are migration-defined like the search_vector GIN. Gate green
(334 pytest incl. the migration drift + downgrade/re-upgrade round-trip).
peter