Copy Calendula's GroupedRow (connected surface segments, press-morph corners)
and the circular colour chip into Floret's ui/common, and render the lists with
them: full-width rows, 72dp height, 2dp connect-gap, ListColorChip (pastelised
list colour + checklist glyph). Shared identity across the two apps.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Restore the smart cards to their original size/typography (they were right)
and put the expressive scale-up where it was actually wanted — the list rows:
titleLarge names, 16dp colour dots, taller padded rows in the surface card.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Bigger smart tiles (160dp, 28dp corners, displaySmall hero counts, larger
icons, 20dp padding, 12dp gaps); list group cards use extra-large (28dp)
corners and surfaceContainerHigh for more presence.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Per-account lists now render inside rounded surfaceContainer cards
(grouped-list pattern) with inset dividers and transparent rows, instead
of bare ListItems on the background — surfaces-first per design feedback.
- Add debug-only DemoSeeder (local, non-syncing 'Floret Demo' list) behind
BuildConfig.DEBUG + an intent extra, to populate the UI for design review.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
First real screen replacing the scaffold: LargeTopAppBar, smart-list cards
in a 2x2 grid with correct tonal pairings (Today=primary, Overdue=error,
Upcoming=tertiary, All=secondary containers), user lists grouped by account
with colour dots + open counts, and an Extended FAB. Consumes the live
ListsViewModel; navigation callbacks stubbed for the next screens.
Verified on device against real tasks.org/DAVx5 data.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
On-device smoke test (tasks.org, which bundles an older OpenTasks provider)
hit SQLITE_ERROR 'no such column: is_recurring' — that column and
distance_from_current don't exist in every provider version. Query all
columns (projection=null) and let the by-name mapper read what's present
(absent -> null); the repository re-sorts, so no provider sort column is
needed either. Verified live: reads the real tasks.org/DAVx5 provider
(1 list, 0 open tasks) with zero errors.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Non-visual stack (backoffice) complete and verified against the real
provider semantics (tasks.org's bundled dmfs TaskProvider, authority
org.tasks.opentasks, org.tasks.permission.* dangerous):
- TaskContract subset + ProviderResolver (runtime authority/permission)
- AndroidTasksDataSource (Instances query, ContentObserver) + TasksRepository
with live Flows; domain models, smart-list filtering, sorting, form validation
- Self-scheduled due-reminder engine (AlarmManager, boot/provider-change)
- Render-only ViewModels + UiState for every screen
- 24 JVM unit tests; assembleDebug + lintDebug + testDebugUnitTest green
- RootScreen is a functional scaffold over real data, to be replaced with the
Material 3 Expressive screens one by one
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Sibling to Calendula. Pure front-end posture; TaskContract data layer,
task screens and reminder engine to follow (see docs/PLAN.md).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
