components/appointment/ had every colour hard-coded, so the flow rendered
identically in either theme even after the wiring was fixed. It now reads from
CSS custom properties.
These are not new colours. The :root values are byte-for-byte the hex codes
that were already in the components — twenty-one files, mapped one to one — so
light mode is unchanged. The dark values are derived from those same colours:
surfaces and borders darkened, text inverted, and the two brand colours (the
indigo and the orange) lightened rather than replaced, because both lose
contrast against a dark surface at their original values.
Verified in a headless browser with prefers-color-scheme forced dark:
data-theme lands on <html> and --ap-surface resolves to #1b1b20 rather than
white. Six one-off colours remain hard-coded — a success green, an error red
and similar — each used exactly once and none of them a surface.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Chasing the dark-mode wiring to the end turned up the actual cause. next-themes
was configured with attribute="data-", which sets an attribute whose literal
name is "data-": the page rendered <html data-="dark">. Both globals.css and
the project's own documentation assume data-theme, so the transition rule keyed
on [data-theme] never applied either, and no Tailwind selector could have
matched. It is data-theme now, verified in a headless browser with
prefers-color-scheme forced to dark.
That is one of two reasons the booking flow looks the same in either theme. The
other is simply that components/appointment/ contains zero dark: utilities —
nothing there was ever styled for dark. The wiring is fixed and the 66 dark
rules that do exist now compile against the real attribute; giving the booking
flow a dark palette is design work, not a wiring bug, and it is not something
this pass invents.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Tailwind was configured with darkMode: "class" while next-themes writes
data-theme="dark" on the public pages. Every dark: utility on the public
site — 111 of them — compiled to a selector that never matched, which is why
the booking flow stayed white in dark mode. The variant strategy now accepts
both .dark (the panel) and [data-theme="dark"] (the public pages), so neither
provider had to change and 66 dark rules now compile against the real
attribute.
The cancel and reschedule modals get tests, the first component tests in this
repo. They pin the things that would be silently wrong: the penalty comes from
the server before anything is cancelled, the free window says "no penalty"
rather than showing a zero, confirming actually sends the request (the old
dialog's confirm button only closed it), a failed preview does not block the
cancellation, slots are requested with exclude_appointment_uuid so the
patient's own hour is not shown as taken, and the reschedule sends only the
start time because the server owns the duration.
The amount assertion deliberately checks the number and unit rather than the
digit shape — numberToArStyle uses the ar-AE locale and its output depends on
the ICU data in the environment.
lib/getStateInfo.test.js had been failing since before this work: an unknown
host falls through to the representation API, so the test made a real network
call and timed out after five seconds. It mocks lib/req now. The suite is
fully green for the first time: 158 tests.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The site's own copy has been promising cancellation for a while —
lib/specialtyContent.js and components/appointment/information/Detail.js both
tell users they can cancel from "نوبتهای من" and get a refund up to five hours
before the visit. The UI never could. The cancel modal existed, but its confirm
button called handleClose: it closed the dialog and sent nothing. And the whole
button row it lived in (ButtonData.js) had been commented out since the PDF
download commit, so it was not even reachable.
Cancelling now goes through POST /appointment/{uuid}/cancel and shows the
penalty from /cancellation-preview before the confirm — the same calculation
the cancel itself runs, so the number the patient sees is the number they are
charged. If the preview fails, the dialog says so rather than blocking; the
cancellation is still allowed.
Rescheduling is new and service-aware. It asks for slots with
exclude_appointment_uuid, so the patient's own hour counts as free rather than
showing as taken, and it sends only the start time — the server computes the
duration. Sending a client-side duration would mean two parallel calculations,
and the day a service's minutes change the appointment would move with a stale
one. The services on the appointment are carried over unchanged.
Both actions only appear for an appointment that is still in the future and not
already cancelled, and both refresh the list afterwards so a just-cancelled
appointment stops showing as confirmed.
The day strip in the reschedule modal is its own small component rather than
the booking flow's DatePicker: that one reads the doctor uuid from route params
and would be undefined inside the dashboard.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Records what the service flow actually guarantees: mode is per location (a doctor
can be slot-based in their office and service-based in a clinic), duration is
server data and must never be summed in the front, and shift boundaries are not
derived client-side because a flat start_times list cannot tell a break between
shifts from a gap left by a booked appointment.
Also notes that user-panel reads of service fields are guarded, since slot-mode
appointments carry none of them.
Task: clinicpro/docs/new_feture/taskes/task-00b-nobat724-service-mode/
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A patient who booked in service mode could not see which services they had
reserved or how long the appointment was. Both now appear on the list card and in
both detail layouts, reading service_items and service_total_minutes from
GET /api/v1/appointments/user.
Every field is behind an explicit guard. Slot-mode appointments carry none of
them, and an unguarded map would crash the card for every slot-mode appointment,
taking the whole panel with it. A reserve entry shows its services but not a
duration, because it has no time.
Covered by Card.test.jsx: slot-mode appointments render unchanged, absent fields
(older backend) do not crash, reserve hides the duration, and a zero duration
produces no row.
Task: clinicpro/docs/new_feture/taskes/task-00b-nobat724-service-mode/
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
adaptServiceSlots labelled its single session "زمانهای خالی" and set end_time to
the last slot's *start* time, so the range shown was shorter than reality. It now
uses the last slot's end_time (falling back to start + total_duration_minutes) and
labels the session with the actual range.
Shift separation was requested but is not implementable from this payload: with a
flat start_times list, a gap between shifts is indistinguishable from a gap left
by a booked appointment. The normal step is duration + buffer, so any threshold
that splits shifts either splits every slot of a long service (step above the
threshold) or invents tabs around booked appointments. A fabricated tab claims a
shift that does not exist, which is worse than one correct tab. Real grouping
belongs to the server-side endpoint task 06 adds.
Also repairs three tests that had been red since adaptSlots changed shape: they
still asserted the old { morning, evening } contract while the function returns an
array of sessions. Suite goes from 4 failures to 1 (an unrelated pre-existing
getStateInfo network timeout).
Task: clinicpro/docs/new_feture/taskes/task-00b-nobat724-service-mode/
Slot-mode contract: adaptSlots untouched
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The service picker summed duration_minutes itself while the backend already
returns total_duration_minutes. Two sources of truth: when the formula changes to
solo/additional minutes, the site would keep showing the old number and the
patient would see a duration that does not match their appointment.
The picker runs before the date step and appointment-service-slots needs a date,
so it is called with today. total_duration_minutes does not depend on the date —
the backend computes it before touching that day's shifts, so the number is right
even when today is closed and start_times comes back empty.
The label reads "مدت تقریبی" until the server number arrives, then "مدت کل".
A missing field or a failed request falls back to the client sum with a
console.warn rather than blanking the step.
Staleness is derived from the selection key instead of reset in the effect body,
which also clears the set-state-in-effect lint warning.
Task: clinicpro/docs/new_feture/taskes/task-00b-nobat724-service-mode/
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
- Added `getBlogTagFacets` API call to fetch blog tags based on city scope.
- Updated `BlogsPage` to read selected tag from URL and handle tag changes with URL updates.
- Modified `Title` component to display tags from the new API and reflect active tag state.
- Enhanced breadcrumb navigation to link categories to their respective pages.
- Adjusted related content section to display articles from the same category and fixed layout issues.
- Corrected heading hierarchy across various components for better SEO compliance.
- Ensured consistent styling and spacing in related content items.
- Updated modification times and AST hashes for several files in the manifest.json.
- Added new entries for `app/api/revalidate/route.js`, `components/blog/detail/Faq.js`, `components/blog/detail/Sources.js`, and other files related to blog functionality.
- Introduced a new AST cache file for the blog frontend sync and default cover prompt, including detailed node and edge relationships.
- Introduced a new script `make-blog-cover.mjs` to create a default cover image for blog posts.
- The image is generated in PNG format with dimensions 1200x630, suitable for Open Graph.
- Utilizes the site's branding colors and logo from `public/nobat724.svg`.
- Includes custom font styling using the Vazirmatn font.
- The generated image is saved to `public/assets/images/blog-default-cover.png`.
- Updated modification times and AST hashes for several JavaScript files in the app/blog, app/component, app/doctor, and components/common directories.
- Added a new AST cache file for a data JSON that was skipped due to a non-object root.
Deactivated doctors are now absent from the public list, but their
/doctor/{uuid} page still loaded. getDoctor() now treats the backend's
raw `is_active === false` as not-found, so the profile 404s like the
list — matching the site's "hide inactive doctors" behavior.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The public list GET /api/v1/doctors only excluded inactive doctors
when an explicit `active` filter was passed; with no param it returned
everyone (deactivated doctors just ranked lower). Deactivated doctors
(admin toggled active_doctor_appointment off) leaked onto nobat724.
- DoctorRepository::findWithFilters: default (no `active` param) now
filters activeDoctorAppointment = true. The active=1 (bookable) and
active=0 (admin, inactive-only) escape hatches are unchanged.
- Doctor::toDetailArray: expose raw `is_active` (= activeDoctorAppointment,
independent of schedule) so public clients can 404 a deactivated
doctor's profile page; distinct from `active` (flag && has_schedule).
- Tests + docs/api/doctor.md updated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Updated mtime and ast_hash for several files in manifest.json
- Added new AST cache files for the clinic contact info fix and maintenance mode client prompts
The backend now answers 503 with code MAINTENANCE_MODE while maintenance is
on. Without this change a visitor got a red error toast over a broken page
client-side, and a silently empty page server-side, because fetchReq discards
the status and returns null on any failure.
- lib/maintenance.js detects the state by BOTH status 503 and the error code;
a bare 503 can come from a reverse proxy and is not maintenance
- The axios interceptor checks it before the 401 branch, so a maintenance
response never triggers the refresh-token path or logs the user out
- fetchReq redirects to /maintenance, with a silentMaintenance opt-out used by
getStateInfo: that one runs inside generateMetadata and while rendering the
maintenance page itself, where a redirect is either ineffective or loops
- redirect() works by throwing, so the try/catch blocks in the doctors,
clinics and specialties pages now rethrow NEXT_REDIRECT instead of
swallowing it
- clinicApi.js handles 503 too; it previously rendered maintenance as a clinic
with zero doctors
- The page reuses the existing 404 design and is marked noindex
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Updated community IDs for various components in graph.json to reflect correct associations.
- Added new function `getPublishedDoctors()` in app/sitemap.js and established relationships with existing functions.
- Adjusted source locations for several functions in app/sitemap.js to ensure accurate mapping.
- Enhanced canonical URL handling in lib/getCanonicalUrl.js to prevent incorrect canonicalization for paginated specialty pages.
- Updated manifest.json with new modification times and AST hashes for affected files.
- Created a new JSON file for the blog city scoping activation prompt, including nodes and edges representing the document structure.
- Added a new JSON file for the SEO post-deploy verification prompt, detailing nodes and edges related to SEO verification tasks.
P1-P11 of the live SEO audit are implemented and verified against a local
production build, so the audit file becomes a reference rather than a
task list: it now records what shipped, the root causes that differed
from the original hypotheses, and the deliberate trade-offs.
Remaining work is split into smaller prompts, ordered by dependency:
- seo-post-deploy-verification: the acceptance criteria were "curl on
production" but were only run against a local build
- blog-city-scoping-activate: blocked on the backend blog city column
- sitemap-simplify-with-city: drops the 35-sweep workaround once the
doctors list exposes city
Each names its blocking dependency and carries reference numbers so a
regression is visible.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Implemented canonical URL strategies for city-specific domains and entities.
- Added helper functions for domain and city resolution.
- Created tests for canonical URL generation and domain resolution.
- Introduced entity quality checks for doctors and clinics to ensure meaningful content.
- Developed unique introductory texts for listing pages to avoid duplicate content.
- Established robots.txt policies for listing pages to manage indexing based on user filters.
- Enhanced specialty content with dynamic introductions and FAQs to improve SEO.
