New coach card replaces the icon+sentence empty state in the Transcribe
panel (edit mode). Three-step guide with 5-s SMIL drawing animation in
step 1 only. Animation freezes at the final frame when
prefers-reduced-motion is active. Footer links to Wikipedia Kurrent and
the Richtlinien page open in new tabs with visible '(öffnet in neuem Tab)'
annotations. 34 new i18n keys in de/en/es.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Final UI/UX spec for the /hilfe/transkription page referenced from
the Transcribe panel coach card. Card-grid layout with per-rule
Beispiel boxes, Wikipedia info-card, "Noch in Klärung" strip, and
closing invitation. Includes impl-ref tables, Paraglide keys for
de/en/es, print styles, and Gherkin acceptance criteria.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
ConversationThumbnail still imported the `$lib/thumbnails` helper that
a02f6cdc deleted, so every SSR render of /briefwechsel crashed with
"Cannot find module '$lib/thumbnails'". Finish that refactor by reading
`doc.thumbnailUrl` straight off the Document DTO (same shape
DocumentThumbnail already uses), and update the spec fixtures to match.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Seeds 120 UPLOADED docs with a deterministic date spread and runs
DocumentService.searchDocuments against a Testcontainers Postgres, not
a Mockito mock. Five cases:
1. First page returns exactly page_size items + correct totalElements
2. Last partial page returns the tail slice (offset 100 → 20 items)
3. Page beyond last returns empty content, totalElements still 120
4. SENDER sort path slices in-memory + reports correct total
5. Different pages return disjoint document id sets
Closes the integration-coverage gap between the Mockito unit tests and
the full Spec→Pageable→Page→DTO path that unit tests can't exercise.
Runs in ~87 s against the shared Testcontainers instance. (#316)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
<a aria-disabled="true"> is the documented pattern but screen readers
still announce "Previous, link, disabled" on pagination bounds — noise
users don't need because the disabled state is purely visual. Switching
to <span aria-hidden="true"> removes the bound control from the AT tree
entirely (Leonie's recommendation). Visual parity preserved via a
disabledBase Tailwind class (same layout + cursor-not-allowed + opacity-40).
Tests updated: "disabled prev/next" assertions now check for aria-hidden
and no href — the active-state href/aria-current assertions are
unchanged. (#316)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
triggerSearch (local state, filter change) and buildPageHref (server data,
page nav) were each iterating over the same ~10 filter params. Any new
filter would have had to land in two places. buildSearchParams is now the
single source of truth for which params the /documents URL understands;
both callers just pass their snapshot and an optional targetPage. (#316)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
PageRequest.of(0, 10_000) was inlined at ~12 sites across DocumentServiceTest
and DocumentServiceSortTest as an "effectively unpaged" sentinel for tests
that don't care about paging. Extracted to a named constant on each class
so the intent is visible at each callsite and we don't risk copy-paste
drift of the magic number. No behaviour change. (#316)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Frontend side of the /documents pagination work. The page.server.ts load
reads ?page= from the URL, forwards page+size=50 to the backend, and
exposes the new totalElements/pageNumber/pageSize/totalPages fields on
`data`. +page.svelte renders a <Pagination> component below the result
list; buildPageHref preserves every filter param and only updates page.
The existing triggerSearch debounce flow intentionally drops `page`
when any filter changes, so filter edits reset to page 0 automatically.
<Pagination> uses plain <a href> links (not goto) so SvelteKit's default
scroll restoration scrolls new pages to the top — the expected senior-UX
behaviour. Decorative chevrons wrapped in aria-hidden spans, 44px touch
targets, focus-visible ring, stacks vertically under 640px. The control
hides itself when totalPages ≤ 1.
Test coverage: 9 cases on Pagination (label, aria-current, prev/next
enable/disable, makeHref invocation, decorative chevron, touch target),
plus a filter-reset assertion on +page.svelte (page 5 → edit q →
goto URL must drop page=). Adds i18n keys in de/en/es. Manual edit to
api.ts pending a post-merge npm run generate:api against a rebuilt
dev backend. (#315)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds 5 dedicated controller cases — paging fields exposed on the JSON,
rejections for size>100 / size<1 / page<0 / page>100000, and a
captor assertion that the built PageRequest is forwarded to the service.
The size>100 case is the load-bearing guard on @Validated at
DocumentController — removing the annotation silently reopens the DoS
window this PR is meant to close.
Adds 5 service cases — fast path uses findAll(Spec, Pageable) (not Sort),
propagates page+size to the DB, carries totalElements/totalPages/
pageNumber/pageSize back on the result, and for SENDER sort slices in
memory and reports the pre-slice total. Page-beyond-last returns empty
content with a correct totalElements (JPA edge case). (#315)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Fast path (DATE/TITLE/UPLOAD_DATE) pushes sort + paging into the DB via
findAll(Specification, PageRequest) and enriches only the returned slice
— 30× cheaper than enriching all 1500 matches when the user is only
going to see 50. In-memory sort paths (SENDER/RECEIVER/RELEVANCE) keep
their LEFT JOIN-friendly sort but now slice in-memory too, so enrichment
still runs against the page slice only.
Controller passes PageRequest.of(page, size) built from @RequestParam
values. Plan-level "add @Validated" prerequisite comes in the next commit.
All existing tests updated mechanically to pass a pageable argument
(PageRequest.of(0, 10_000) as an "effectively unpaged" sentinel). Stubs
that previously matched findAll(Specification, Sort) for the fast path
now match findAll(Specification, Pageable) with PageImpl<>.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Rename `total` → `totalElements` for Spring-Page parity and add three new
required paging fields: pageNumber, pageSize, totalPages. Adds a `paged(
slice, pageable, totalElements)` factory alongside the existing single-page
`of(list)` shortcut. Enables offset pagination of /documents search (#315).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds two specs for extending issue #294 with bulk uploads:
- bulk-upload-concepts.html — three concepts (stack, split-panel
with file switcher, progressive accordion) with a decision
matrix and the Concept B recommendation.
- bulk-upload-split-panel-spec.html — refined final spec for
Concept B. Covers all three states (N=0 empty · N=1 single ·
N≥2 multi) across 320 / 375 / 768 / 1280 viewports in both
light and dark mode, using the real tokens from layout.css.
Includes impl-ref tables for every new surface, Paraglide keys
in de/en/es, component tree, and backend contract.
The polymorphic-state model means /documents/new is a single
route: N=1 is byte-identical to #294, N=0 shows a whole-panel
drop zone with bulk-first copy, N≥2 grows a file-switcher strip
under the PDF preview plus a two-card form split.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
h-16 w-16 looked undersized in the 180×252 strip container (~25% of
the height). h-24 w-24 gives ~38% visual weight, matching the ratio
DocumentThumbnail uses for its lg (120×168) fallback (#309).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Prior coverage only exercised getThumbnailUrl() as a Java method call.
The new case serialises via ObjectMapper and asserts the resulting JSON
contains "thumbnailUrl":"..." so we catch silent breakages in the wire
contract (getter rename, @JsonIgnore, visibility drop) — not just
regressions in the method's return value (#309).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Uses the same heroicon as DocumentThumbnail so the "no thumbnail yet"
signal reads identically across the app: one shape, one meaning. The
parchment SVG still lives on in the fully-empty state (no resume doc
at all), where it represents a different thing — we removed it only
from the "document exists, thumbnail not generated yet" branch (#309).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Replaces the generic parchment SVG placeholder with an <img> pointing at
the backend's thumbnail endpoint when the document has one. The 180×252
container matches DocumentThumbnail's 5:7 A4 convention so the
dashboard tile sits visually next to the list/person-sublist tiles
instead of looking squatter than they do. dark:mix-blend-multiply keeps
paper scans from glaring on a dark page background (#309).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The helper had a single consumer (DocumentThumbnail) and its only job
was to compose what the backend's Document.getThumbnailUrl() now
produces. Deleting it locks the single-source-of-truth invariant —
there is no longer a way to build a thumbnail URL on the client (#309).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The backend now exposes thumbnailUrl as a serialised computed property
on Document, so the component drops its dependency on the frontend
URL-builder. PersonDocumentList's inline Doc prop type follows the
same shift (#309).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Reflects the new @JsonProperty getter on Document. Kept as a minimal
manual edit rather than a full regen because the running dev backend
belongs to the main workspace and swapping JARs there would be a
side effect on a parallel worktree's state. `npm run generate:api`
will converge on the same shape.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
DashboardService now reads the URL from the Document's computed getter
instead of passing null, so the resume strip can display the real
thumbnail of whatever the user was last working on (#309).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
@JsonProperty makes the computed getter part of every Document response
Jackson produces, so any DTO returning a Document automatically carries
the thumbnail URL without per-controller plumbing. The accompanying
comment warns future readers that the cache-buster is load-bearing
for the endpoint's `immutable` cache header (CWE-525) (#309).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Matches the shape the frontend previously built via
encodeURIComponent(thumbnailGeneratedAt), so the backend is now the
single source of truth for the thumbnail URL convention (#309).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The no-cache-buster branch covers documents whose thumbnail key is set
but whose thumbnailGeneratedAt is still null — which only happens in
the narrow window between the key being persisted and the async worker
stamping the timestamp (#309).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
First TDD step for centralising the thumbnail URL convention on the
Document entity (#309). Adds a stub getter returning null and a test
that locks the "no key → no URL" branch.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
PR #311 dropped the right/left arrow icons that signalled whether a
letter was sent or received. Readers who don't decode the colored
left border (new users, color-blind users, users at a glance) had
no visual cue for direction. Restore a 20×20 arrow inline with the
title — right-arrow for outgoing, left-arrow for incoming — kept
decorative (aria-hidden) since the aria-label already announces
"Gesendet:" / "Empfangen:".
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The 168px-tall thumbnail tile was dominating rows where the text
column only rendered at text-xs / text-sm — visually the right
column sat half-empty. Three changes:
- Title: text-sm → text-lg
- Summary: text-sm → text-base
- Meta + tag chips: text-xs → text-sm
And remove the "vor N Jahren" chip entirely. The documentDate
in the meta row already carries the temporal context and the
chip was adding visual noise without new information.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Makes `max` an optional prop with default 3 — the common row-layout
case doesn't need to name the cap explicitly. ThumbnailRow's callsite
drops to `<TagChipList tags={doc.tags ?? []} />`, consistent with how
other shared components in $lib/components expose sensible defaults.
Refs #305
Fixes @leonievoss round-2 follow-up from PR review
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds row_direction_sent / row_direction_received keys across the
three locale files (de: Gesendet/Empfangen, en: Sent/Received, es:
Enviada/Recibida) and routes ThumbnailRow's directionLabel through
Paraglide. An English or Spanish screen-reader user now hears
"Sent:" / "Enviada:" in their language, matching the DistributionBar
i18n pass.
Refs #305
Fixes @leonievoss round-2 follow-up from PR review
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Rewires briefwechsel-rows.visual.spec.ts against the shared fixture
(seedBilateralPair + cleanupBilateralPair), adds afterAll cleanup,
and folds the conv-person-bar visibility gate into openBilateral()
so both the structural test and the snapshot block fail loudly on
a hero-state regression — matching the a11y spec's safety net.
Refs #305
Fixes @saraholt follow-ups 1 + 2 + 3 from PR round-2 review
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Lifts the three-API-call seeding (create sender, create receiver,
create document) out of briefwechsel-a11y.spec.ts and into a
dedicated fixtures module. The spec now calls seedBilateralPair()
in beforeAll and cleanupBilateralPair() in afterAll so the test
DB doesn't accrue seeded rows across reruns.
Two caveats captured in the helper docstring: the backend has no
person-delete endpoint (only the document is purged), and the
timestamped last names make leftover persons collision-free.
Refs #305
Fixes @saraholt follow-up 1 + 2 from PR round-2 review
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Extends the seeding pattern from the a11y spec: beforeAll creates two
persons + one document so the page renders the row layout. The
structural test now asserts the ConversationThumbnail tile AND the
DistributionBar are present — a regression that drops to the hero
or breaks the row wiring fails here instead of silently passing a
hero-state check.
Snapshot block stays gated on VISUAL=1 (baselines captured during
review against a seeded backend) so the structural coverage ships
immediately and the pixel-diff coverage ships once baselines land.
Refs #305
Fixes @saraholt blocker 2 from PR review
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The previous version navigated to /briefwechsel with no params, which
renders the hero state — axe-core scanned the hero, not the new
ThumbnailRow / ConversationThumbnail / DistributionBar. This commit
seeds two persons + one document via the API in beforeAll, then
drives the URL with ?senderId=X&receiverId=Y so each of the
36 test runs (3 viewports × 2 themes × 2 assertions) actually scans
the intended DOM. Also asserts that conv-person-bar is visible first,
so a regression that drops the page back to hero fails explicitly
rather than silently passing an empty sweep.
Refs #305
Fixes @saraholt blocker 1 from PR review
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Lifts the three-chip-plus-"+N" tag row out of ThumbnailRow into a
standalone TagChipList component so the chip cap + overflow policy
lives in one place and can be reused on other surfaces (document
detail header is a candidate). ThumbnailRow drops from 110 to ~90
lines and no longer owns tag-slicing logic — it just asks for the
list with max=3.
Behavior is byte-identical: same data-testid, same max cap, same
"+N" overflow indicator. All ThumbnailRow row-level tag tests
continue to pass against the new composition.
Refs #305
Fixes @felixbrandt suggestion 1 from PR review
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
persistThumbnailMetadata was a four-arg method signature that mixed
three conceptually related values. Wrapping them in a private
ThumbnailResult record drops the signature to (Document, result),
mirrors the existing SourcePreview record one step earlier in the
pipeline, and keeps generate() reading as a narrative of small
named outputs rather than positional arguments.
Refs #305
Fixes @felixbrandt suggestion 2 from PR review
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Defaults `now` in $props() destructure so each row instance freezes
its reference time at mount, instead of calling new Date() inside
the $derived every reactivity tick. No behavioural change — the
date math is stable across re-renders for a given row — but drops
the nullish-coalesce dance and is cleaner under Storybook-style
testing where a deterministic `now` is injected.
Refs #305
Fixes @felixbrandt suggestion 3 from PR review
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
relativeYearsDe already returns "" for future dates (covered in its
own spec), but the integration wiring inside ThumbnailRow was
untested. Adds a regression that a doc with documentDate in the
future produces no "vor N Jahren" or "vor weniger als 1 Jahr" chip.
Refs #305
Fixes @saraholt concern 5 from PR review
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Consolidates the hansPerson / annaPerson fixture into a makePerson()
factory matching the makeDoc convention, adds an assertion that
the bilateral list renders one ConversationThumbnail tile per
document (catches a broken {#each} keying wired around the
DistributionBar), and decouples the DistributionBar aria-label
assertion from the German locale now that i18n lands via Paraglide.
Refs #305
Fixes @saraholt concerns 3 + 4 from PR review
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Bumps the multi-page badge from text-xs (12px) / px-1.5 py-0.5 to
text-sm (14px) / px-2 py-1. Meets senior-legibility on a 320px phone
without crowding the 120-wide tile — the badge stays tucked in the
top-right corner.
Refs #305
Fixes @leonievoss senior-accessibility concern from PR review
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Drops the hardcoded German strings ("Briefverteilung in diesem Zeitraum",
"{n} von {name}") and routes every visible + assistive-tech string
through dist_bar_aria and dist_bar_segment message keys. An English
or Spanish user now sees "from" / "de" instead of "von" both on
screen and in the aria-label their screen reader announces.
Refs #305
Fixes @leonievoss i18n concern from PR review
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Without this prefix, a color-blind user or screen-reader user has no
indication of correspondence direction — the colored left border is
information but not announced, and the arrow glyphs were removed in
the earlier layout pass. Prepending "Gesendet:" or "Empfangen:" to
the aria-label gives assistive-tech users the direction first so the
row identity is unambiguous even without color perception.
Refs #305
Fixes @leonievoss WCAG 1.4.1 concern from PR review
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Captures the reasoning behind persisting two scalar columns on
documents rather than deriving aspect client-side or standing up a
thumbnail_metadata table. Also documents the 1.1 landscape threshold,
the null-during-rollout state, and the ordering invariants inside
ThumbnailService.generate().
Refs #305
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a dedicated axe-core sweep for /briefwechsel so contrast or
semantic regressions on the new row layout fail independently of
the catch-all accessibility suite. Scoped to the main landmark so
shared chrome violations (if any) aren't double-reported.
Refs #305
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a Playwright spec gated on VISUAL=1 with one snapshot per
(mobile/tablet/desktop × light/dark) = 6 baselines. Snapshots stay
skipped in CI until the baseline set is captured and committed —
running `playwright test --update-snapshots briefwechsel-rows`
against a seeded backend generates them.
Structural check runs unconditionally so the file is wired into CI
today rather than waiting for the baseline capture step.
Refs #305
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds two new assertions for the extracted DistributionBar — it must
appear in bilateral mode and stay hidden in single-person mode — and
repairs the shared makeDoc fixture: the embedded Person now carries
personType + displayName so the fixture matches the regenerated
Document schema without TypeScript complaints.
Refs #305
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Drops the inline row markup, arrow icons, status-dot helper, and the
otherPartyName helper that only fed it. Each visible row is now a
ThumbnailRow, which owns its own aria-label, border color, meta and
tag rendering. The year-divider and "new document" footer are
untouched — they were always intended to stay as timeline chrome.
Also widens the documents prop shape to include the summary, tags
and thumbnail metadata that ThumbnailRow consumes; the backend
already returns these fields via the Document schema so no server
change was required.
Refs #305
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
