Two things the shop asked for, and one it will notice.
**A message on an exception.** A closure or a late opening can now carry
a reason, and it reaches the panel. It shows while the door is shut and
disappears the moment the shop opens — a notice explaining a late
opening is worse than useless once the door is open.
It outranks a free message on purpose: "closed this afternoon, back
tomorrow" is what someone standing outside needs, and "new collection in
store" can wait. Say so if you would rather it were the other way.
The first version keyed this off the day's `isOpen`, which means "this
day has opening hours" — so a day opening at 14:00 counted as open all
morning, exactly when the reason is needed. A test caught it; the rule
now reads the state at this minute.
**Reusable phrases.** The same handful of notices get written over and
over. /admin/modeles keeps them, translated once, and offers them
wherever a message is composed — exceptions, closure periods, the
banner. Picking one costs no translation at all: the saved English is
reused directly, where the service takes the better part of ten seconds.
Exception notes and holiday labels are translated too, which they were
not before. After the save rather than during it, so a slow service
never costs the shop its dates.
**Previewing the future.** The dashboard can render the panel at any
moment within about a year: "what will the window say while I'm away?"
is worth answering before someone is standing in front of a locked door.
The whole pipeline was already a function of "now", so this costs
passing a different instant. An unparseable or absurd value falls back
to the present rather than confidently rendering nonsense.
Checked against a three-week holiday, which surfaced something worth
knowing: the automatic "opens on…" line stays empty, because the
resolver's search is bounded to fourteen days. During a long closure the
message is the only thing that tells customers when the shop is back —
which is a good reason for this feature to exist.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012cSY9pVhZmJUKNN7wf1Myd
"Opens later at" and "closes earlier at" were prefilled with 14:00 and
16:00 — arbitrary values, wrong on most days, which left the preview
showing hours nobody had asked for until both fields were corrected.
They now start from the day's real opening and closing, so any
adjustment is relative to what that day actually is.
A day with a lunch break is bounded by its first opening and its last
closing, which is what these two actions act on.
Changing the date changes the hours the times are relative to, so a
value typed for Tuesday must not survive into Thursday. The fields live
in a component keyed on the date and React resets them by remounting —
the React compiler rightly refuses an effect that writes state back, and
it is the clearer arrangement anyway.
The end-to-end test picks a trading day rather than assuming today is
one. The first version guarded the assertions behind a visibility check,
which would have passed in silence on exactly the days the controls
matter least.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012cSY9pVhZmJUKNN7wf1Myd
Opening later and closing earlier were fixed to today, which was an
assumption rather than a requirement: closing next Thursday afternoon is
the same decision as closing this one, and the shop knows about it in
advance more often than not.
The date is a single control and still defaults to today, so the
two-click path for the most frequent case is unchanged.
The usual hours differ from one weekday to the next, and "opens later"
is late relative to them, so the whole sixty-day window is resolved on
the server and handed over at once. Sixty small entries cost less than a
round trip per change of date.
The date control's accessible name is explicit: "du" alone collided with
the "Du" field of the range form below it, which the end-to-end run
caught as an ambiguous locator — and an ambiguous locator is an
ambiguous label for anyone using a screen reader.
The end-to-end suite ran before this was pushed, unlike last time.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012cSY9pVhZmJUKNN7wf1Myd
Twenty-six tests across two viewports, covering what the spec asks for:
changing today's hours reaches the panel, a message survives the
translation service being unavailable, a closure period closes the days
it covers — plus the authorisation paths and the sign-out regression.
They run against the standalone build served the way the container
serves it, not `next start`, which refuses to work with standalone
output anyway. The suite therefore exercises the artifact that ships
rather than a second arrangement that could drift from it.
Sign-in mints the session cookie Auth.js would have issued rather than
driving Authentik. What is under test is the application's behaviour for
a given role; the handshake itself is verified against the live provider
separately, and standing up an identity provider per run would trade a
lot of machinery for coverage of somebody else's code. The secret lives
in one module imported by both the config and the fixtures — when it
differed, every signed-in test failed at once while looking like an
authorisation bug.
Database access goes through plain SQL rather than the Prisma client,
whose generated module format Playwright's loader and Next's bundler
disagree about. That traded one problem for a subtler one: node-postgres
parses a DATE column into a local-midnight Date, so reading it back
shifted the day at UTC+2. Dates are read as text now.
The mobile profile runs on Chromium: WebKit needs system packages only
root can install, and a suite nobody can run locally is a suite nobody
runs. The config says how to switch to the real engine.
Two real defects surfaced, both found by the tests rather than by
reading. The seven "Ouvert" checkboxes on the hours page were
indistinguishable to a screen reader; each now names its day. And on a
phone the signed-in address appeared nowhere at all — the header hides
it to save room — so nobody could tell which account was about to sign
an audit entry on a device the shop shares. It is on the dashboard now.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012cSY9pVhZmJUKNN7wf1Myd
vliaudatandClaude Opus 5