Commit graph halkyon/AGENTS.md
Author SHA1 Message Date
c20578e421
feat(app): tags keep their case; dictionary tags read JLPT-N5, Noun
A tag is still its name, but no longer in lowercase: the name keeps the
case it was written in, and names that differ only in case are one tag
(ADR 0004, rule 2, amended). Two such tags from two devices fold into the
one with the smaller id, its case with it. Tags stored before stay
lowercase until renamed.

Words from the dictionary are tagged JLPT-N5 to JLPT-N1 and with their
word class (Noun, Verb, i-Adjective, na-Adjective, ...). Anki tags and
PDF chapter tags arrive as written.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MMm7usAo6q9ZZRoPkWjoxi
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:27 +02:00
add1c76842
feat(app): the dictionary in the app: look words up, add them as cards
A studyset's dictionary follows from its language, as decided: the
instance's dictionary of its front's language, with meanings in its
back's (else English). Nothing to attach or detach; nothing of it shows
where there is none.

- Add material › From the dictionary: search a word (kanji, kana,
  romaji, meaning) and add it as a flashcard, or add a whole level
  (JLPT N5…N1, common words) at once, asked first above 500 words, its
  progress shown. Words already in the studyset show as such and are
  left out. Each card is a vocabulary card (headword, reading, senses
  with their codes), tagged `jlpt-n5` and with its word class, and
  marked as imported from the dictionary.
- Look up beside a vocabulary card's front fills the card with the word
  chosen, keeping its hints, explanation and tags.
- A card keeps the dictionary entry it was made from (card schema 6:
  `dictionary` {source, id}), to be found again and not added twice.
- The dictionary's credits show at its foot and in About, as its licence
  asks.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MMm7usAo6q9ZZRoPkWjoxi
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:26 +02:00
7008fcac52
chore(server): generated code as the generator writes it
dart format had gone over the generated files too (a blank line between
import groups, other line breaks), which CI's check against a fresh
`serverpod generate` would have refused. AGENTS.md now says to format
only the files one wrote.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MMm7usAo6q9ZZRoPkWjoxi
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:26 +02:00
3013bc04b3
feat(server): dictionaries, loaded from files and searched
The server now holds dictionaries for everyone: `dictionary_entry` and
`dictionary_source`, reference data owned by no one, never synced. It
loads the dictionary files in its `dictionaries` folder at every start,
those whose SHA-256 changed: in the background, each source replaced in
one transaction, so searches meanwhile find the old entries; a file it
can't read is logged and left out. All of JMdict loads in under half a
minute. On an instance the folder is deploy/dictionaries, mounted
read-only (compose.yaml); the server's README says how.

DictionaryEndpoint (signed in): the dictionaries with their credits; a
search (exact key, exact loose key, key starting with the query, loose
key starting with it, each in the source's order; katakana as hiragana,
macrons written out), some 150-450 ms over all of JMdict; and the
entries with a label, counted and a page at a time (all JLPT N5 words).

The curator now keeps spellings that aren't quite right apart, as
`looseKeys` (romaji with long vowels typed once): "sushi" finds 寿司
before 数詞 (すうし).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MMm7usAo6q9ZZRoPkWjoxi
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:25 +02:00
b1ef9b9d3a
feat(app): a studyset's languages, and Chinese and Korean fonts
The studyset sheet gains Languages, as before: off until "Set
languages", then the language of the cards' fronts and of their backs
(22 languages, Chinese in its two scripts). A studyset without its own
takes those of the nearest studyset above it, both together; set for the
first time, they start from those it took. The payload held both fields
already.

They set its text: Noto Sans SC, TC and KR join Noto Sans JP (bundled,
never fetched; 34 MB more, fonts 42 MB in all, your call), and since
Han characters take the shapes of the first font that has them, every
screen inside a studyset (its screen, cards, notes, paths, studying it)
puts the font of its front's language first, else of its back's,
Japanese otherwise. Every studyset route goes through one helper now.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MMm7usAo6q9ZZRoPkWjoxi
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:24 +02:00
4345741646
feat(app): Noto Sans and Noto Sans JP, bundled, the same on every device
The app brought no fonts: text took each device's own, a minimal Linux
showed Japanese as boxes, and kanji could take Chinese shapes. Now Noto
Sans is the app's font everywhere (Latin, Greek with polytonic,
Cyrillic), and Noto Sans JP fills in kana and kanji in their Japanese
shapes. Noto Sans JP alone wouldn't do: it lacks accented and polytonic
Greek. Both come with the app and are never fetched; scripts neither
covers fall back to the device's own fonts.

Each family is one variable font holding every weight (Flutter sets the
weight axis from the requested weight; a test checks bold inks more):
0.7 MB Noto Sans, 0.7 MB its italic, 8.1 MB Noto Sans JP, from Noto's own
GitHub releases (assets/fonts/ATTRIBUTION.md). Their SIL Open Font
License texts join the app's licence list.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MMm7usAo6q9ZZRoPkWjoxi
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:23 +02:00
02efa2cdcf
feat(app): every empty studyset tab says how to add some, as Files did
Paths, flashcards and notes now show what Files showed when empty: an
icon for what the tab holds, "No … yet", and the menus that add some
("Add material → Create a path", "⋮ → Import content", …), through one
shared widget, EmptyListMessage. The icons are those the app uses for
each already: the route, the study screen's card stack, the note pen,
the folder.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MMm7usAo6q9ZZRoPkWjoxi
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:22 +02:00
7975586994
feat: files in an order of one's own, with Reorder beside the tabs
A PDF now has a position among its studyset's files, as notes and paths
have: a fractional index in a column of its own (so moving it on one
device and renaming it on another don't overwrite each other), last
writer wins. The migration adds it to pdf_source on the server and on
each device with 'a0', so PDFs from before stand in the order they were
added (ties sort by id), and new ones go at the end.

The Files tab lists them in that order by default (Custom order, first
in the Sort menu; newest first was the default before). With two files
or more, Reorder beside the tabs shows the custom order whatever they
were sorted by, a grip in place of each file's options and the Sort menu
put away, until Done.

The next release changes synced tables anyway (card.buriedAt, since
v0.2.2), so older apps have to update for that already.

Also: card.position's doc no longer speaks of moving cards.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MMm7usAo6q9ZZRoPkWjoxi
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:22 +02:00
04cbf00e3c
feat(app): notes in an order of one's own, with Reorder beside the tabs
With two notes or more, Reorder beside the tabs gives each note a grip
in place of its options, as paths have, until Done; e.g. to follow a
reading order. A note's position was stored already; the repository
now moves it (only that column changes, and it syncs).

The notes' bloc sits with the studyset's other lists now, above the
tabs, so the button knows how many there are. The position helpers the
paths kept to themselves are shared (core/sync/fractional_position.dart),
and so is the grip (ReorderGrip).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MMm7usAo6q9ZZRoPkWjoxi
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:22 +02:00
6c9dfadf27
refactor(app): flashcards keep the order they came in, without Reorder
Studying breaks the order of due cards anyway; what follows the list
order is new cards and cramming, and that order comes from creating and
importing, with sub-studysets putting lessons in order. No one drags a
card among a thousand (decided 2026-10-08).

Gone: Reorder and the grip on flashcards, CardListEventMoved,
CardRepository.move and the gateway's updatePosition, with their tests.
Each card keeps its stored position, so the order stays the same on
every device. liftWhileDragged is Material's plain lift again, for the
lists that keep Reorder.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MMm7usAo6q9ZZRoPkWjoxi
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:21 +02:00
bb23ca2478
fix(app): studyset tabs built as they scroll, and a Reorder mode for flashcards
A studyset of 1,200 flashcards froze the window for seconds on opening
(7.4 s in a debug test; a release build is faster but still grows with
every card), long enough for the desktop to offer to kill the app: the
tabs held each list as one child of a scrolling view, shrink-wrapped, so
every row was built and laid out at once, and again on every change.

The tabs are now CustomScrollViews holding each list's slivers, built as
they scroll into view (0.07 s for the same 1,200 cards). What a list says
in place of its rows sits in SliverCentered.

On the desktop, Flutter's default drag handle lay over each card's "…",
so a press there began a drag rather than opening the options. Cards now
move in a Reorder mode, as paths do: Reorder beside the tabs (with two
cards or more, all showing), a grip in place of each card's options,
Done to end it. ReorderToggle and liftWhileDragged are shared with the
paths.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MMm7usAo6q9ZZRoPkWjoxi
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:21 +02:00
6c911e50bb
docs: FileSaver and the text origin in AGENTS.md
Also drops the Column around the Library import sheet's one entry
(avoid_single_child_in_multi_child_widgets).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MMm7usAo6q9ZZRoPkWjoxi
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:21 +02:00
aa69be47d1
feat(app): Home, the Library and the studyset sheets laid out as before
Home: plain cards on the surface, 16 dp apart, the goal first and the
Pomodoro timer after it in a new dashboard. The streak chip sits centred
in the app bar, and its sheet shows the days inside a big flame and the
week as marks over the weekdays' initials, with a word on today in a box.
The week names its day as "Thursday, Oct 8" and its days carry no
tooltip. The heatmap scrolls its weekdays with the weeks, outlines the
days to come, and names a day's activities with its date. The goal card
has a ring of 5 dp, a check once met and a pencil, and its sheet saves
across its width with Remove goal below. Arranging puts an eye on each
widget's corner and lifts the one picked up, with a tap the hand feels;
with every widget hidden, Customise brings arranging back. The stepper
moves to core, where the goal's sheet takes it too.

Library: the counts line includes files (PDFs); Refresh in the app bar
and pulling the list down sync now; a list that can't be read offers
Retry, which reads it again; the star keeps its square corners.

Studyset sheet: "Title *", the secondary title explained behind a
question mark, Inherit as a box beside the colour, and the description
in up to three lines with a chevron. Move sheet: round marks, the
studyset itself marked as current, the top level only while it is
inside another, archived studysets last under their heading, the
destination named in the footer, and a new studyset to move into, made
only once the move is.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:18 +02:00
770de51bff
feat(app): keep the tabs beside notes and paths, flat app bars, the sync dot on the cogs alone
A studyset's notes and learning paths are routes of the Library tab, as
its screen is, so the navigation stays beside them; studying, cards, PDFs
and settings still take the whole window. App bars stay flat when content
scrolls under them: Material 3's tint flickers as an overscroll crosses
the edge again and again. The sync trouble dot shows on the Settings cogs
only; Home has a cog of its own now, so the Profile tab needn't carry it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:17 +02:00
beb3a32c30
perf: keep a card's burial on its row rather than in the review log
Finding today's burials in the review log meant reading every review the
cards ever had, once more in the library's progress, the card list and
each round. card.buriedAt holds it instead, read with the cards those
screens read anyway. The sync engine merges a row's fields one by one, so
a burial and a move of the same card on two devices both stand (a sync
test checks it). The review log's bury mode is gone.

The migration adds one nullable column, on the server and on devices:
ALTER TABLE "card" ADD COLUMN "buriedAt" (timestamp without time zone on
the server, INTEGER in SQLite); existing rows read as not buried. Apps
built before it stop syncing until they are updated.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:17 +02:00
5a9441a717
feat(app): undo every step of a study session, deletions held until it ends
The header's Undo takes back ratings, burials and deletions, last first,
across the session's rounds and from the end of a round, and shows the
card again. A card deleted during a session is held (DelayedDeletions.hold)
until the session ends, then goes after the usual time to undo, so the
message offering Undo still works; if the app is killed first, the card
stays. Undo on a burial's or deletion's message takes back that step alone.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:17 +02:00
27a535b305
docs: note what cards carry beside their type, and the review log's other entries
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:17 +02:00
b0aaecf115
feat(app): tag, edit, delete and swipe cards while studying, with tidier session sheets
- Above every studied card (but a matching one): the PDF it came from,
  its tags on one line ("+2 tags" for the rest) or "Add tags", and a
  menu to tag, edit or delete it; the session follows (an edited card
  shows as saved, a deleted one leaves the round).
- On phones, a front/back or vocabulary card swiped right is rated
  Good, left Bad; it tilts over the colour it heads for, and springs
  back when let go early.
- The session's settings: "Settings", a 48 dp Save that warns mid-round
  ("Heads up!"), steppers in an outlined box ("20 / day", "5 / 50"),
  the order in an outlined dropdown, the chosen tags with the rest
  behind "Filter by tags" / "+3 more".
- A studyset's study options: Save above Reset to defaults, the
  description small, reviews a day in steps of 25.
- One stepper look everywhere (study, Pomodoro profiles), its buttons
  "Decrease" and "Increase".

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:12 +02:00
65976ae63a
feat(app): cite sources in APA, MLA or Chicago, and give studysets a bibliography
A source is now shared by learning paths and studysets (core/citation/):
authors as full names, year, title, what it is a part of, publisher,
ISBN, link, and a citation kept as a file brought it. formatCitation
cites it in APA 7, MLA 9 or Chicago 18: it inverts and abbreviates
names itself (particles, suffixes, names written as already cited stay),
sets a work of its own in italics and a part of something apart from
the whole, and writes a missing year as "n.d." in APA and Chicago.

- Settings: a Citations card per device, with a preview and a link to
  the style's own guide.
- Studysets: a Bibliography button opens the works cited, to read and,
  behind its pencil, to change (studyset schema 4).
- Steps: sources edited through the same sheet, opened from an
  "N sources" button in the runner (step schema 2; schema 1 reads on).
- Path import: steps' sources, text kept as written, the old format's
  "citation" read as what a source is part of.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-09 18:09:12 +02:00
2030e397cc
fix(linux): the Pomodoro timer shows in the top bar from the snap too
All checks were successful
CI / check (push) Successful in 9m41s
CI / deploy (push) Successful in 14s
From the snap, nothing showed in the top bar: AppArmor refused the
item's bus name (dbus_bind org.freedesktop.StatusNotifierItem-<pid>-1),
which snapd's desktop plug doesn't grant, contrary to what the research
found. The item is now laid out as libayatana-appindicator lays itself
out, within what snapd v2.77's desktop-legacy plug allows: at
/org/ayatana/NotificationItem/halkyon_learn, the only path from which
the label's signal (XAyatanaNewLabel) may come, and registered with the
watcher by that path under the connection's own name, binding no name
at all. Checked against GNOME's watcher here: it registers as the
system's own indicators do.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-07 13:33:49 +02:00
2deabcb5b0
fix(app): a sheet shows that more of it lies below the visible part
Found on the laptop: the Pomodoro profile's last switches (Do Not
Disturb among them) lay below the visible part of the sheet, and
nothing said so: the body scrolled right up to the pinned buttons.
Every sheet (SheetScaffold) now has, as Material 3 has scrolling
dialogs, a line above the footer while some of the body lies below and
one under the title while some lies above, and a scrollbar beside the
body, always shown on the desktop in place of the one Material shows
only while scrolling.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-07 12:28:07 +02:00
d9f7498cd4
feat(linux): suggest the top bar's extension instead of a notification
Decided 2026-10-07: where GNOME's top bar shows no icons (the
AppIndicator extension missing or off), the timer shows in the app
alone, and no notification stands in for it. The Pomodoro card says
that the timer can count down in the top bar, with a button that asks
GNOME Shell to switch the extension on where it is installed or to
install it (GNOME's own dialog), or opens its page on
extensions.gnome.org where Shell can't be asked; until the person says
not to suggest it again. An app can't install or switch on a Shell
extension by itself. The alert at a phase's end stays.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-07 12:28:06 +02:00
ad28a5d8c2
refactor(app): studysets and settings are written in the newest schema
Decided 2026-10-07: the codecs write the newest schema every time, as
before, rather than the oldest one that holds what is there. An older
app loses sight of a studyset or of the settings once a newer one has
written them, until it is updated too.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-07 12:28:05 +02:00
a822763df8
fix(app): the Pomodoro timer holds up with the app gone, asleep or edited
A review of the timer found:

- Stop in the notification with the app swiped away took the timer
  away there, but the app showed it again at its next start. The
  service now leaves a mark the app takes when it starts, and the timer
  it kept is forgotten.
- Pause, Resume and Skip did nothing with the app gone: since Android
  12 a service may not start an activity from a notification's button.
  The notification is posted again with buttons that open the app.
- A timer kept while the app was closed was shown as it was before it
  caught up, so its ended phase alerted a second time. It is now caught
  up before it first shows.
- Dart's timers sleep with the device, so a phase that ended while the
  phone or laptop slept went unnoticed until the app came back. While a
  phase runs the timer looks every 15 seconds, and on Android the
  alarm's receiver wakes it too.
- On Linux, changes that came at once overlapped on the bus (an
  indicator registered twice, notifications left over, a ticking label
  never stopped); they now go one at a time. The indicator registers
  again when GNOME's watcher restarts, and with a watcher that only
  turns up later.
- Changing the running profile could put the timer outside its cycle
  ("Focus 6 of 4", the Live Update's bar past its end); where it is
  stays within the new cycle, and a phase waiting at its full length
  takes the new length. A countdown past its end shows 00:00.
- Writes of the kept timer go one at a time, and a closed timer arms no
  more timers.
- The alert no longer renames the timer's notification channel, and a
  German hint said the alert came at the device's end.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-07 12:28:04 +02:00
1921043925
fix(app): a path import that fails says why in the app's language
The importer said what was wrong with a file in English sentences built
in the data layer, the last English left in a German app. It now names
the problem, the step it is in and what the problem names (a field, a
type, a number), and the studyset screen puts it into words: "Step 2:
unknown "type": quiz." or "Schritt 2: unbekannter „type“: quiz.". The
format's own field names stay as they are in the file.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-07 12:28:03 +02:00
918e10bb2b
feat(app): a Pomodoro timer on Home, in the notification and in the top bar
Home has a Pomodoro timer (decided 2026-10-07, after Goodtime): the
profile it runs, chosen there while no timer is set, the time left of
the phase, and Start, Pause or Resume, Skip and Stop. While studying, a
small countdown sits in the study screen's header, and a tap pauses it
or goes on. A phase that ends leads to the next, which starts by itself
where the profile says so and otherwise waits, paused at its full
length.

Profiles follow the person to every device, in their settings (schema
4, still written as schema 3 while the profiles are as they come, so
older apps go on reading the settings): Classic, 25 minutes of focus,
5 of break and 15 after four, and 52/17 without long breaks, built in;
both can be changed, more added and deleted, from Home's card or from
Settings. A profile can start breaks and focus by themselves, and keep
the device quiet while focusing. The running timer is the device's own,
kept in the person's folder: closed and opened again, the app catches
up with the phases that ran out meanwhile.

Android: the app's own foreground service (Kotlin, specialUse, as
Goodtime does) shows an ongoing notification counting the phase down,
with Pause or Resume, Skip and Stop; on Android 16 it is a Live Update,
its progress bar the profile's cycle, each phase a segment, focus in
cobalt and breaks in copper. An exact alarm (USE_EXACT_ALARM) alerts
with sound and vibration when the phase ends, also with the app
closed, and Do Not Disturb goes on while focusing once the person
allowed it, the filter before given back afterwards. No notification
plugin: flutter_local_notifications has Live Updates only in a
prerelease that needs compileSdk 37; androidx.core 1.18, which the
build resolved already, has every API needed.

Linux, over the session bus (package:dbus, already in the build): where
the desktop shows icons in its top bar (GNOME with the AppIndicator
extension, which Ubuntu has built in; KDE) the timer counts down there,
the buttons in the icon's menu, as a StatusNotifierItem written in Dart
on the names and paths the snap's desktop plug allows; elsewhere a
notification with the buttons says until when the phase runs. An alert
with sound ends each phase while the app runs, and GNOME's banners stay
off while focusing, through gsettings. xdg-desktop-portal's Background
status, the planned fallback, serves Flatpaks only.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-07 12:28:02 +02:00
d42afb078a
feat(app): the library's options move, import into and set how a studyset is studied
A studyset's options in the library now offer, as on its own screen,
Edit, Move to…, Study options, Import from Anki, Archive and Delete.
Moving goes through the same sheet and says where the studyset went;
importing reads an Anki deck into it as Add material does.

Study options say how a studyset is studied, on every device: how many
new cards it starts a day (by default as the person's study settings
say), how many reviews it brings a day (no limit by default; cards in
their learning steps come anyway), and whether reviews or new cards
come first, or the new ones spread evenly among the reviews. They live
in the studyset's payload, schema 3; a studyset without them is still
written in schema 2, so that older apps, which skip a payload in a
newer schema whole, keep seeing every studyset nobody gave options.
Editing a studyset keeps its options, and Reset to defaults takes them
away.

The study queue takes each studyset's limits for its own cards and the
order of the studyset studied; a card's schedule now knows when it was
last reviewed, to count the reviews of the day. Sessions, the round's
end and the library's badges all count the same way, the badges now
with the person's own number of new cards a day too.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-07 12:28:01 +02:00
3ee9bce0c5
feat(app): the app speaks British and American English and German
Settings offers English (UK), English (US) and Deutsch after the
device's language, each named in its own language. app_en.arb, the
template, is now British English (colour, centre) and the fallback for
any language the app doesn't speak; app_en_US.arb overrides the five
messages American English spells differently, and app_en_GB.arb is
empty, there so that Material's texts and dates are British too. The
German translation says "du" and keeps one set of terms throughout
(Lernset, Karteikarte, Lernpfad; Anki's own Stapel and Unterstapel in
the Anki import). arb_files_test checks that German has every message
of the template and that every translation keeps its placeholders.

resolveAppLocale picks the first language the device prefers that the
app speaks, in its country's variant where there is one; other English
is British, and a country alone never picks a language (Spanish in the
United States isn't American English), which Flutter's own resolution
would do.

Both MaterialApps take material_ui's GlobalMaterialLocalizations
delegates: gen-l10n's list brings flutter_localizations', which serve
only package:flutter/material.dart, so material_ui's widgets found no
Material texts in German and threw.

The editor's own text blocks show our placeholders while empty (list
item, to-do, quote, heading, and the desktop's / hint): the editor's
own follow a locale of their own, process-wide, in few languages. The
last texts written into widgets (the person's data failing to open,
the cards failing to read, the About dialog) moved into the arb files.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-07 12:28:01 +02:00
55a0a7fdf0
feat(app): appearance, language and version in Settings, and Settings from Home
Settings gets a card for how this device shows the app: its appearance
(as the device, light or dark) and its language (the device's, or one
the app speaks, each named in its own language), side by side from
medium windows on, as two outlined dropdowns. Both stay on the device
rather than following the person: they are kept in
<app support>/device_settings.json (written to a temporary file and
renamed into place), read before the app shows, so its first frame
already looks as chosen, and a file that can't be read falls back to
what the device does. DeviceSettingsBloc sits above SessionGate's keyed
subtree, since the settings outlive every person, and both the plain
app and the person's app take their theme mode and locale from it.

At the bottom of Settings, the release and its build, read once at
start together with the build the server is asked about.

Home has a Settings cog beside the streak, with the sync trouble dot as
on Profile. The library's create button reads "Create studyset", and
Delete is red in every menu that offers it: the studyset's card, screen
and rail, and the composer's draft.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-07 12:27:56 +02:00
667180416a
ci: a commit that passed the checks isn't checked again
All checks were successful
CI / check (push) Successful in 5s
CI / deploy (push) Has been skipped
A fast-forward merge brings the commit a branch's push checked to main
unchanged, and CI ran every check a second time before deploying it. The
check job now marks every commit that passed in the runner's home
(checked/, kept 90 days), and a commit marked already skips straight to
its deploy; the sandbox check still runs every time. Only pushes run CI,
so only those who may push can leave a mark.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-05 22:17:12 +02:00
4d3d4d9592
feat: releases work out their version from the commits
All checks were successful
CI / check (push) Successful in 9m17s
CI / deploy (push) Successful in 14s
tool/release.sh no longer takes a version: it works it out from the
commits since the last release, following semantic versioning (feat
raises the minor number, fix or perf the patch number, a breaking change
the major number from 1.0 on and the minor one before; other types
release nothing), or takes major, minor or patch as chosen. It shows the
version and its commits and asks first, checks that CI is green on the
newest commit with code, writes the version and a raised build number
into the pubspec in a signed release commit with a signed tag, and
pushes only once the APK and the snap passed their checks. The release
notes list the features and fixes. A run that stopped after its release
commit and tag carries on with them.

Channels for now: the release on code.halkyon.app (APK, snap, checksums)
and the F-Droid repository; the Snap Store upload is off until
snap_store_channel names a channel.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-05 21:22:50 +02:00
d149d4a5e3
feat: the F-Droid repository moves to halkyon.app
All checks were successful
CI / check (push) Successful in 9m23s
CI / deploy (push) Successful in 15s
The repository is Halkyon's, for all of its apps, so its address is the
brand's: https://halkyon.app/fdroid/repo. learn.halkyon.app stays Halkyon
Learn's own address, for its website later. Moved before anyone but the
maintainer had added it; the address is permanent once people have.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-05 20:41:54 +02:00
ae2496335d
feat: an F-Droid repository at learn.halkyon.app
All checks were successful
CI / check (push) Successful in 9m16s
CI / deploy (push) Has been skipped
Releases also publish the APK in the project's own F-Droid repository,
https://learn.halkyon.app/fdroid/repo. tool/fdroid_repo.sh adds the APK,
signs the indexes with the repository's own key and uploads them to
ramiel as the user fdroid, whose key can only write into the folder
Caddy serves (rrsync), the APKs first and the indexes last.
tool/release.sh checks that key and its password before building and
publishes after the Forgejo release.

- tool/release/fdroid/ holds the repository's settings (Halkyon, the
  snap's icon) and the app's listing (the snap's text), with the
  anti-features TetheredNet (learn.halkyon.eu only, for now) and
  NonFreeNet (YouTube, the server's SSPL) declared with their reasons,
  in F-Droid's own definitions and with its icons (fdroiddata, AGPL v3).
- The index key's certificate is pinned: a new key would make the F-Droid
  app refuse the repository.
- fdroidserver v2.4.5, since Ubuntu's v2.2.1 knows neither TetheredNet
  nor reasons.

README and AGENTS.md describe it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-05 20:25:02 +02:00
a966080d99
docs: the README's aims, developer setup and releases
All checks were successful
CI / check (push) Successful in 10m0s
CI / deploy (push) Successful in 11s
The README opens with what the app aims at: creating, sharing and
collaborating on study material, FSRS for retention, study material
organised with its external references, offline-first, and a transparent
stack anyone can host.

"Setting up" lists what developers need: Flutter as CI pins it, with
Flutter's own guide per platform; the Serverpod CLI; the Linux packages,
with what each plugin needs them for; Developer Mode on Windows, and why;
no Docker for development; Python for the scripts in tool/.

"Releases" says where releases are published, how to check a download
(the signed checksums, the APK certificate's fingerprint), how versions
are numbered, and how a maintainer makes one with tool/release.sh.

AGENTS.md: Windows needs Developer Mode for Flutter's plugin links too,
not only for pdfrx's build.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-05 17:00:27 +02:00
88b82cda4b
feat: a release script for the APK and the snap
All checks were successful
CI / check (push) Successful in 9m8s
CI / deploy (push) Successful in 1m5s
tool/release.sh builds a release on the maintainer's laptop: one APK for
every phone, from a checkout of the signed tag and signed with the app's
own key, and the snap, both connecting to learn.halkyon.eu, the project's
instance; their checksums signed with GPG; the release on code.halkyon.app
through Forgejo's API, and the snap in the Store's edge channel.

Before building it checks that main is clean and pushed, that the tag is
signed by a key deploys accept, the version and a raised build number,
the pinned Flutter, and that the keystore holds the key whose certificate
tool/release/android-certificate.sha256 names. After building it checks
the APK's signer, package, version, that it isn't debuggable and that it
names the server, and the snap's version. The key's password and the
Forgejo token are asked for at each run and never stored.

Release builds take the app's key from the environment and otherwise fall
back to the debug key, so `flutter run --release` works anywhere. The
native libraries go into the APK compressed: an APK for three
architectures was 110 MB with them stored as they are, 48 MB now. The
list of dependencies Gradle adds for Google Play is left out.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-05 13:00:39 +02:00
e35ce5c81d
feat: the app tells when sync fails, and when it is too old
Published apps can't be made to update, so the app now notices when it
no longer matches the server: the server names the oldest app build it
works with (instance.info's minimumAppBuild), asked before the first
sync of each launch, and the sync engine's refusal of differing synced
tables counts the same. Either way the app says "update the app; if it
is up to date, the server needs an update" in a banner on Home and in
the new Sync card in Settings, and what the person does meanwhile stays
on the device.

The Sync card shows where sync stands since the app started (syncing,
everything synced, sync failing, with the last completion and failure)
and offers Sync now. While the last sync failed, a dot sits on the
Profile tab and the Settings cog. One sync that goes through clears it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-05 13:00:38 +02:00
999a3ce0b0
feat(deploy): hourly off-site backups with restic
All checks were successful
CI / check (push) Successful in 9m0s
CI / deploy (push) Successful in 7s
learn.halkyon.eu dumps its database every hour (people's files are in it
too) and restic stores the dump encrypted in a Hetzner Storage Box,
keeping 24 hourly, 7 daily, 4 weekly and 3 monthly snapshots. The box's
own snapshots, removable only in Hetzner's Console, protect the backups
from a taken-over server. A systemd timer runs it as deploy; once a day
it also prunes and checks the repository. Failures go to the journal.

Hourly, because a restore costs people who started after the backup
what their device held: the app no longer finds its session and starts
anew. deploy/backups.md describes the setup, the restore and what a
restore loses; both were rehearsed on the local Compose stack.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-03 16:45:40 +02:00
b96db731c1
feat: PDF sources: read, take notes and cards from pages, deadlines
Some checks failed
CI / check (push) Failing after 26s
CI / deploy (push) Has been skipped
PDFs live in a studyset's new Files tab: uploaded from Add material
(a file the studyset holds already asks first), each row with its
pages, cards, notes and size, how far it is read and its reading
deadline; renamed, given a deadline (a page or a chapter by a day),
marked as read, started afresh, deleted with a moment to undo.

The reader (/pdfs/:id) opens where reading left off: on wide windows
a rail of tools (search, fit, zoom, fullscreen), a page counter with
the chapter, and a sidebar; on phones a top bar, a page scrubber and
a sheet. The sidebar holds the PDF's flashcards, by page or chapter,
made from selected text or written there, each tied to its page and
tagged with its chapter; and the page's note, an ordinary note tied
to the page. Screenshots of a region become images that know their
page. Cards and screenshots lead back to their page; reading on past
a chapter with cards suggests studying it. Home shows the deadlines
running and to come, and a timeline of them.

Server: pdf_source rows (content in the payload, last writer wins)
and an append-only pdf_reading log, the furthest page derived from
it so devices never conflict. Files up to 50 MB for PDFs (20 MB for
others; a larger file that isn't a PDF is deleted once stored). A
deleted PDF takes its file's asset row along unless another PDF has
the same file. Card payload 4 and note payload 2 keep the page.

The reader draws with pdfrx (PDFium) behind PdfEngine in
Dependencies; widget tests use a fake, and
integration_test/pdf_engine_test.dart checks the real engine on a
device or the desktop.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-03 14:40:08 +02:00
ec34bb2334
feat(app): editor blocks: callout, spoiler, formulas, files, video, study button
All checks were successful
CI / check (push) Successful in 8m34s
CI / deploy (push) Successful in 1m1s
Blocks beyond the editor's own, each built once for editor and view:
- callout: eight kinds in theme colours, kind, title and "folded at
  first" from its header; written in place, left with Enter on its
  empty last line; never inside another
- spoiler: written in plain sight, hidden behind Reveal read-only
- formula on its own line, written on its own screen with a preview
  and snippet buttons, and formulas within a line (Σ in the format
  bar); set by flutter_math_fork behind MathView, LaTeX's numbered
  environments read as their aligned forms
- image and sound picked from the device, kept as synced files
- YouTube video from its link, with its start and a caption: a card
  until played, nothing fetched before; played in place where a web
  view exists (youtube_player_iframe, privacy-enhanced), else opened
  in the browser, behind YoutubeVideos in Dependencies
- study button: a studyset, a tag or all cards, with or without its
  topics, back first; its StudyFocus rides on the study route and
  narrows that session only
- tables, already in the editor, now insertable everywhere

An Insert button opens on the format bar, so blocks can be inserted on
phones too; the desktop's / menu lists the same blocks, translated.
Card faces now dock the format bar while written, so cards get blocks
and formatting on phones as well. The blocks are listed in
EditorBlocks: the core's, plus the features' through
EditorBlocksScope. Blocks of a type the app doesn't know say they need
a newer version and stay as they are, so formatVersion stays 1.

Their types and keys are in content_blocks.dart, which RichContent
reads: a block's text counts as text, and a video or study button
isn't blank. The path import now makes formulas real ones.

integration_test/youtube_player_test.dart plays a real video through
the app's player on a device, which no widget test can: YouTube
refuses embeds that don't say where they are embedded.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-03 12:20:41 +02:00
43834ef42b
docs: restart serverpod start after a migration of synced tables
The sync engine (serverpod_offline_sync v0.0.8) keeps rebuilding the
projections of every space on every session of a server whose first
reconciliation changed its registry of synced tables
(serverpod_offline_sync#150, open). `serverpod start` migrates in the
running server, so after such a migration the development server is
slow until restarted; deploys reconcile in their maintenance run and
aren't affected.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-02 23:01:54 +02:00
79bc2fc618
Build and run the app on Android
The editor's fork moves to keyboard_height_plugin v0.3.0 (83bdb2a3): v0.1.5
compiled against API 31, while the AndroidX libraries of Flutter v3.47 need
34, so Gradle's AAR metadata check failed. Release builds get the internet
permission, which Flutter's template grants to debug and profile builds
only. Gradle's own build folder in android/ (its problems report) is
ignored. AGENTS.md: the server's test/integration/test_tools/ is generated
too, so test files are formatted one by one.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-02 19:52:22 +02:00
ca69f74fd9
Limit how much room people's files take
All checks were successful
CI / check (push) Successful in 6m41s
CI / deploy (push) Successful in 1m4s
An anonymous person may keep 100 MB of images and sounds, an identified one
500 MB, and one person more by a file_allowance row the operator adds; all
of them together 15 GB, since every file is in the database and in each dump
the server keeps (decided 2026-10-02). Each comes from the environment
(FILE_LIMIT_*_MB), with these defaults, documented in deploy/.env.example.

The server checks when an upload starts, counting unfinished uploads so they
can't fill the disk unnoticed, and again when it finishes, counting what is
kept, so uploads started side by side can't overshoot; a refusal is a
FileLimitExceededException, and an upload refused as it finishes is deleted.
deletePerson deletes the allowance with the person. Migration: the new
table file_allowance, only added.

The app keeps a refused file waiting on the device and stops that round of
uploads; the Anki import warns when a deck's files may not fit in the room
left (asset.usage()).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-02 06:02:15 +02:00
7b933b2951
Document files and the Anki import
All checks were successful
CI / check (push) Successful in 6m30s
CI / deploy (push) Successful in 1m3s
AGENTS.md: the file endpoints and the app's files, asset URLs in content,
imported cards, platform plugins behind interfaces, server tests on the sync
engine's database, and GStreamer for Linux builds. ADR 0004: content names
files as asset:<id> URLs, and how stored files are deleted.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-02 05:11:51 +02:00
5a5b2feead
feat(app): map questions on Japan's prefectures
A map question is a multiple-choice card drawn on a map: the composer's
Map kind takes the region tapped as its answer, and studying it means
tapping a region and checking, the right one green and a wrong pick red,
each named. The map of Japan's 47 prefectures comes from Natural Earth
(public domain), built by tool/maps/japan_prefectures.py from release
v5.1.2 (its download checked by hash): simplified, small islands
dropped, rounded; about 107 KB. Maps are drawn to fit, tappable and
zoomable for small regions.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-01 21:10:07 +02:00
f66595a1d6
feat(app): study by tag, chosen in synced settings
All checks were successful
CI / check (push) Successful in 5m44s
CI / deploy (push) Successful in 1m2s
The study screen's session settings choose tags to study a studyset by:
its cards with any of them. The choice lives in the person's settings
row, synced, by tag id and name, so a renamed tag stays chosen and so
does one folded into another of its name. Two settings rows made offline
fold into the one with the smallest id, which gets what only the other
has.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-01 17:40:11 +02:00
9dc9b6a5ed
feat(app): choose, manage and filter by tags
The card editor offers its tags: a picker that searches by the start of
a name, creates a tag (name and colour) and chooses it, and saves the
choice with the card. A sheet lists every tag with its count, to add,
rename (merging into a tag of that name), recolour and delete them, with
an undo. The card list shows each card's tags and narrows to one tag.

The editor closed twice once tags changed after saving; it now closes
on becoming done only.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-01 17:40:10 +02:00
31d87e8415
feat(app): the studyset screen with its topics, the tree's own screen
All checks were successful
CI / check (push) Successful in 5m28s
CI / deploy (push) Successful in 7s
A studyset's screen, as decided with the tree (2026-09-30/10-01):

- A breadcrumb from the library down to the studyset; its steps go up
  (replacing the location, so back leads further up), and steps that
  don't fit fold into a "…" listing them, nearest first.
- From expanded windows on, two columns: the description over the topics,
  the flashcards beside them, Study and Add material in the app bar. Below
  that, the topics are in a drawer, opened from a bar at the bottom that
  also holds Study and Add material and slides away while reading onwards;
  the back button stays where a drawer would have taken its place.
- The topics: "All topics" for the whole studyset, then the studysets
  inside it in their order, each with the study it invites, what it holds
  and its options (edit, delete), opening on top of this screen; a switch
  rearranges them by dragging; "New sub-studyset" creates one inside.
- Study covers the studyset with everything inside it ("Learn 12" while
  something is due). Add material offers flashcards for now.
- Options: Edit (the sheet, now with the description, edited in the whole
  window, and for a studyset inside another the colour of the one it is
  in, which a new one takes at first), Move to… (the tree and the top
  level, without the studyset itself and what is inside it), Archive or
  Unarchive with Undo, Delete with Undo, which closes the screen.
- The description cut short when long, the whole a tap away; the licence
  as the last line, linking to its text where that is known.

The list's bloc, scoped to a parent, lists its children, works out the
studyset's own progress, creates inside it, reorders and moves. The card
list leaves adding and studying to the screen. The inline editing of the
title and description, and its bloc, are gone.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-01 15:14:57 +02:00
41f536bbe2
feat(app): deleting a studyset can be undone for a moment
Some checks failed
CI / check (push) Failing after 24s
CI / deploy (push) Has been skipped
Decided 2026-10-01. Deleting still asks first; once confirmed, the studyset
leaves the library at once and "Studyset deleted" offers Undo for three
seconds. Only when the time to undo is over (four seconds: the message's
three and a moment for a late tap) are its cards and then the studyset
blanked and deleted, since blanking can't be undone (ADR 0002, rule 2). A
deletion still waiting when the library closes (the app quits, the person
changes) is dropped, and the studyset stays.

The library's bloc keeps the waiting deletions; the event that ends the
time to undo is private, so no screen can skip it. AGENTS.md now describes
the library's shared widgets, the sheets (and why a sheet that asks before
discarding can't be swiped away), the order of studysets, and this Undo.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-01 09:59:47 +02:00
8e0f5cf538
feat(app): leaving the profile tab cancels a waiting GitHub sign-in
Decided 2026-10-01, in place of "closing the account dialog cancels it",
which had no dialog left once the account moved into the Profile tab:
choosing another tab cancels a sign-in waiting for GitHub in the browser,
since the person has left it behind. Settings, opened over the profile, is
not leaving it.

Screens learn whether their tab shows from TabVisibility, which the tabs'
container now provides: go_router's indexed stack as before (every tab
alive, hidden ones offstage with their animations paused), each tab wrapped
in it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-01 09:59:46 +02:00
c21ce6232b
feat(app): the UI foundation: theme, tabs and routes, strings
The app gets its look and navigation before more features land on them.

- Theme (core/theme/): hand-written light and dark Material 3 colour schemes around
  cobalt and copper; ThemeExtras for the colours Material has no role for (states, the
  four FSRS ratings); the 20 colours a person can give content. Tests hold text to WCAG
  AA contrast on each palette colour and on the dark theme's primary colour and surface.
- Layout and shared widgets: Material 3's window-size classes (core/layout/);
  SectionCard, AppLoadingIndicator, confirmDestructiveAction and FoundById
  (core/widget/). Lucide icons.
- Navigation with go_router: four tabs (Home, Library, Explore, Profile), each keeping
  its own stack of screens, as a navigation rail with the app's icon from 600 dp and as
  a navigation bar below; choosing the open tab again returns to its first screen. A
  studyset's screen sits in the library; studying, cards and settings cover the tabs.
- Routes carry ids: FoundById finds the item in the repository's live list, so a
  studyset deleted on another device, or an old link, says it no longer exists. A
  screen that deletes its own item keeps showing it while it closes, rather than
  flashing that message.
- Every person gets an app of their own (PersonApp, lib/src/app/): a
  MaterialApp.router with its own GoRouter and PersonScope around the navigator, built
  by SessionGate inside the person's keyed subtree. Screens read the person's
  repositories wherever they are built, so the static route() helpers and the listener
  that closed every pushed screen on a change of person are gone: a change of person
  now disposes the whole router and starts again at home.
- The account dialog becomes the Profile tab, with Settings (sign out, delete the
  account, About) behind a gear. A waiting GitHub sign-in now keeps waiting while
  another tab shows; a change of person still cancels it.
- The new screens' strings come from Flutter's gen-l10n (English for now). The
  generated files are committed, left out of analysis and formatting, and written
  anew by `flutter pub get`, so CI's check of generated code also catches stale ones.

Widget tests now start the whole app (test/support/pump_app.dart) at the screen's path
and move through it as a person would, by tabs and taps.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>
2026-10-01 09:59:44 +02:00