Adding a secretary meant deciding all 17 permission resources in the same
dialog. The full-height capture showed the form running past 1300px with the
save button below 16 accordions, and the dynamic registry makes that worse: every
page added in future lengthens this one modal.
The add/edit modal now carries only doctors, profile and address, and fits on
screen with its footer visible. Permissions move to SecretaryPermissionsModal,
reachable from a row action and opened automatically right after a successful
add, since a new secretary starts on the role defaults and the owner usually
wants to set them.
Neither create nor update sends permissions any more — the backend seeds the role
defaults on create, and the permissions modal owns the writes, fanning out over
every link row so a secretary shared across doctors stays consistent.
PermissionAccordions moves to components/ui as a shared component. Sections now
start collapsed with a granted/total badge on each header, so the panel opens at
a fixed height and still says which sections are on.
Two design-system slips caught by re-screenshotting rather than by the audit:
- a text button as a third row action pushed the name column out of the table, so
the desktop row uses an icon with a title and the mobile card keeps the label
- .btn.secondary is not defined in styles.css (variants are primary/ghost/soft/
danger/accent), so it renders as a bare .btn. Used ghost here. 25 other files
have the same dead class; left alone as a separate sweep.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
PatientController::resolveScope and TenantTagController::guardTagView only ever
checked the secretary, while every write in both controllers already ran through
both checkers. So an invited clinic doctor with patients.view off got 200 with an
empty list where a secretary got 403 — one permission, two behaviours. No data
was exposed either way; tenant scoping emptied the result.
The fix is not canOrNonMember. That collapses two different situations: a
membership row switched to active=false means the collaboration ended, and
ClinicDoctorPermission::can() returns false for everything in that case too.
Routing it through the permission gate turned the existing 404 on a single record
into a 403, which confirms the record exists to someone who just lost access.
ClinicRecordAccessTest caught it.
isActiveMemberDenied() answers the narrower question — active member, permission
off — and leaves a deactivated row to the data scope, which closes it with a 404
and discloses nothing. A test now pins that distinction so it cannot be collapsed
again.
Tags keep the tags.view OR patients.view rule, now for both roles.
Verified live in three states: active with both off 403/403, deactivated not 403,
active with patients.view on 200/200.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A full role-by-role sweep (9 roles x 18 endpoints against the running app) showed
the addresses toggles in the owner's permission form controlled nothing. Grep
confirms it: no gate anywhere referenced 'addresses'. The panel's address list was
gated on appointment_settings.view instead — the same borrowed-permission pattern
already fixed for resources and treatment.
GET /api/v1/addresses now gates on addresses.view.
The resource drops to view-only. Creating, updating and deleting an address in
ClinicController is explicitly owner-or-admin
($clinic->getUser()->getId() !== $user->getId()), so those three actions could
never be delegated to a secretary or an invited doctor no matter what the form
said. Both role defaults narrow to ['view' => true] to match, and stored JSON
keeps its old keys harmlessly since merge only reads registry keys.
This widens secretary access: addresses.view defaults to true while
appointment_settings.view defaults to false, so secretaries who could not list
addresses now can. That is deliberate and costs no confidentiality — the same
addresses are already served anonymously from
GET /api/v1/clinic/{uuid}/addresses, which is whitelisted in security.yaml.
Verified live in three states: default 200, addresses.view off 403, and
addresses off with appointment_settings on still 403, proving the borrow is gone.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The backend gate is now per-action, so a single canUpdate driving add, edit and
delete would show buttons the server answers 403 to. Each button now checks its
own action, and the includes modal takes canCreate/canDelete so its add select
and per-edge remove button follow the same split.
ServiceCategoryTab is deliberately left alone: its save patches
/api/v1/service-item/{uuid}, which is ClinicServiceController and already gated on
services.update, so canEdit was already the right permission. Only its read of the
category tree moved behind services.view, and the page it lives on already
requires that.
The page test now drives a configurable can(), covering view-only, create-only,
update-only and delete-only.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The controller carried only IS_AUTHENTICATED_FULLY on the class and none of its
15 routes checked a permission. A secretary whose owner had turned `services`
fully off could still create, rename and delete service categories, build item
groups, replace group members, and rewrite service relations and per-branch
overrides.
Scope is intra-tenant privilege escalation, not IDOR: owned() and requireItem()
already resolve every uuid against the caller's active environment, so no data
crossed tenants.
Gating is per-action (view/create/update/delete) and reuses denyServices() from
ClinicServiceController in the same domain, so a secretary with `update` cannot
create or delete. The call is the first statement in every action, before
requireCategory/requireItem — placed after, an unknown uuid would answer 404 and
leak whether the record exists.
An earlier note claimed these endpoints were consumed by the booking flow and so
could not be closed. That was wrong. service-selection/validate, the group routes
and the relation routes have no consumer in any of the three API clients, and the
sibling controller already puts every service read behind services.view — the
booking modal reads service-items through it — so any flow needing services
already needed the permission.
The docs claimed appointment_settings.* for the includes routes, which was never
enforced either; corrected to services.*.
The test loops the whole route list rather than sampling, and a guard asserts the
count of #[Route( equals the count of denyServices( so a future ungated route
fails here.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The previous note said GET /api/v1/subscription/my has no gate because it
returns 200 with every permission off. That was wrong. It gates deliberately
with a degraded payload instead of a 403: without subscription.view the
response drops the active subscription and used_trial, and effective_plan
keeps only features, max_secretaries and max_resources — no plan identity,
no billing. Verified against the running app both ways.
The 403 it does not throw is the point: FeatureGate and useSubscription need
capability flags on every page, so a 403 would break the whole panel.
Also adds .claude/prompt/service-catalog-permission-gate.md for the one real
gap, with the per-route analysis that was previously deferred.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
clinic.md's default envelope is regenerated from the running app, so it now shows
all 17 resources instead of 13, including services with its full create/delete
actions. Both role docs point at permission.md for the shared registry and spell
out the merge rule that makes new resources work on existing rows: deleting a key
means "take the default", not "deny" — denying requires an explicit false.
A systematic sweep over every gated route with all permissions off found two
places where the docs claimed enforcement that does not exist:
- GET /api/v1/subscription/my returns 200 with every permission off. Only trial
is gated.
- ServiceCatalogController has no gate at all.
Both are pre-existing and both are left as-is rather than half-fixed: their
endpoints are also consumed by the booking and subscription flows, where a hard
gate would break secretaries who legitimately need them. The docs now say so.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The three hardcoded resource lists in the admin panel are gone. MySecretariesPage,
SecretariesPage and DoctorPermissionsModal now render from
GET /api/v1/permission-catalog, so a resource added to the backend registry shows
up in all of them with no frontend change. Each has a test that proves exactly
that by adding a resource to the mock and asserting it renders.
SecretaryPermissions was an interface with a field per resource, which made
"dynamic" impossible in TypeScript — every new resource would have been a compile
error. It is now an open map. Only two files consumed it.
The borrowed gates are corrected:
- five resource pages moved off appointment_settings onto their own 'resources'
- treatment-cases moved off appointments onto 'treatment'
- service-categories moved onto 'services', which is what ServiceCatalogController
actually manages (categories, item groups, service relations) — not resources
TreatmentCaseController had no permission gate at all, only IS_AUTHENTICATED_FULLY,
so any secretary could read and edit treatment cases. All seven of its actions are
now gated on treatment view/update.
ResourcePermissionTrait takes the resource from an overridable method instead of
hardcoding appointment_settings. HolidayController overrides it back, since the
holidays page really is appointment settings. The booking gate keeps its
appointments.view fallback so a secretary who may book is not blocked by a
resource-config permission.
Defaults were picked to preserve today's effective access, so no role gains or
loses a page from this move.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Both permission forms in the admin panel can now render from the backend
registry instead of their own hardcoded lists. Resources come back as an array
so display order is part of the contract, each carrying its Persian label, its
actions, and the clinic_only flag that used to live in the frontend.
contextPermissions() normalizes the no-row branch through the registry too, so
a doctor whose permission row was never provisioned sees the same shape as one
who has it.
Two existing assertions compared the API response against DEFAULT_PERMISSIONS
by identity. The values are unchanged; only key order moved to the registry's,
so both now compare through PermissionCatalog::merge.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The list of permissionable resources was duplicated in six places that had
already diverged: both permission entities, three admin UI files and the
SecretaryPermissions TypeScript interface. Adding a resource meant editing all
of them, so new pages borrowed an unrelated resource instead — five resource
pages sat on appointment_settings.view and treatment-cases on appointments.view.
PermissionCatalog is now the only place that says which resources and actions
exist. Each entity keeps its own DEFAULT_PERMISSIONS, but as role policy only;
a test asserts those defaults never name a resource the registry doesn't have.
getPermissions() merges the stored JSON over the role defaults, so a resource
added to the registry later resolves to the role default instead of silently
false for every existing row. Explicitly stored values are never overwritten,
and no data migration is needed.
Two asymmetries fixed along the way:
- ClinicDoctorPermission validated writes against its own DEFAULT_PERMISSIONS,
so services.create/delete could never be stored for an invited doctor.
- DoctorSecretary had no validation at all and would store any key, and it only
read $patch['resources'] — the admin SecretariesPage sends a flat map, so its
permission edit silently did nothing. Both entities now accept either shape
and filter through the registry.
New resources 'resources' and 'treatment' are registered with defaults chosen to
preserve today's effective access, since both pages are currently gated on a
borrowed resource.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A finished session prints three areas with their device readings, so a course
with any history pushed the sessions that still need work off the screen.
Each session is a collapsible card now, with the scannable facts kept in the
head — number, status, date, staff, and an area count so opening is a decision
rather than a guess. History starts collapsed; a session that still needs
booking starts open, because its button is the reason it is on the page and
should not sit behind an extra click. A session with neither areas nor an
action has no body and renders as a plain row rather than an empty toggle.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The tab stacked every course's full session list on one page. A protocol allows
sixty steps, so one open course was enough to bury the others. Courses are cards
now — service, progress, supervising doctor, staff — and opening one replaces
the list with its detail: a back button, two tabs (the whole course, or only
what is still to come), search and paging inside each. The choice lives in the
URL so browser-back returns to the same course.
Booking stays where the work is: the button sits on the session card inside the
upcoming tab, not on a separate page.
The patient banner's 'نوبت بعدی' read '—' for anyone mid-course, because it only
looked at booked appointments and a course's later sessions have none yet. It
now falls back to the next session of the active course and relabels itself
'جلسهٔ بعدی' when it does — a planned session is not a booking, and the banner
should not call it one.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The modal took its date from the caller and had no way to change it. Opened from
a session card, that meant the projected date or nothing: if the device had no
free time that day the user had to close a form they had already filled with a
patient and a service, go elsewhere, and start again.
The date is now a field in step one, seeded from the prop and reset when the
prop changes. ServiceSlotPicker already clears the picked slot when its date
changes, so the times reload for the new day on their own.
The header chip drops the date whenever that field is on screen; showing the
same value twice, one editable and one not, invites the reader to trust the
wrong one.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The tab was called 'نوبتهای بعدی' but shows the whole course — finished sessions
with their recorded readings as much as upcoming ones. It is 'دورههای درمان' now.
Booking from a session still made the user search for a patient the page already
had open. The plan response carries the patient's national code (from the
profile, falling back to the user — the same COALESCE PatientController uses,
because users.national_code is routinely empty), and the modal takes a patient
prop that seeds the lookup and hides the search step. The old 'بیمار یافت شد'
card is suppressed in that mode; saying it twice is noise.
Sessions are now searchable and paged. A protocol allows up to 60 steps and a
patient can hold several courses, so an unbounded list was only ever going to
work for the small cases. Search filters on what the card actually shows —
service, staff, status, session number, area names — and runs in the page,
since /plan already returns the whole course and a round trip would add latency
and nothing else.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The staff field started empty even when the service's treatment protocol already
named who may perform it — a decision made once in the service settings and then
asked again on every booking. The modal now reads that protocol and preselects
its first staff member, but only until the user touches the field; otherwise a
manual choice would be wiped on the next service change. A protocol with no
staff leaves it empty and does not block submission.
Renames the four user-facing 'اپراتور' strings to 'پرسنل', matching the record
in /admin/staff that they all refer to. ResourcesPage keeps the word: there it
names a kind of bookable resource (doctor, operator, room, device), not a
ClinicStaff row.
Drops a test whose premise the default invalidated; the two new ones cover both
sides — protocol with staff sends staff_uuid, protocol without staff sends none.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A patient's file had no view of their multi-session courses: which ones they
have, when the remaining sessions fall, or what was recorded in the sessions
already done. All of it lived on a tenant-wide page.
The tab lists the patient's courses and, per course, a card for every session
with its date, its status, and — for finished ones — the areas treated with the
device readings a staff member logged. Estimated dates are labelled as such, so
a projection is never read as a booking.
Each unbooked session carries a button that opens the same NewAppointmentModal
used elsewhere, seeded with that session's date, and now binds the resulting
appointment to that exact session via a new treatmentSessionUuid prop — a
patient can have several open courses, and without it the attachment falls back
to guessing from the service.
The modal opens in resource mode, not doctor mode: a course's service is booked
against the device's calendar, so useDoctorBookingServices returns nothing for
the supervising doctor and the picker would render 'no bookable services'. The
plan response now carries the course's device for exactly this. A course with no
device yet says so instead of offering a button that cannot work.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
GET /api/v1/treatment-case/{uuid}/plan returns every session with a date and an
is_estimate flag, plus each session's area records — the device readings a staff
member actually logged. Until now nothing exposed either: due_at existed only
for the next session, and TreatmentCase::toArray() serialised sessions without
their areas, so 'what was done' was unreachable outside the staff panel.
Kept separate from GET /treatment-case/{uuid}; that response feeds the edit
modal, which needs neither the calendar nor the areas.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
TreatmentScheduler deliberately writes only the next session's due_at, because a
date not yet anchored to anything real is a false claim about the future and has
to be rewritten every time a patient runs late. But the panel still needs to
show the whole course.
TreatmentPlanProjector builds that chain at display time and writes nothing.
Each date carries is_estimate so a projection is never mistaken for a fact. A
session's anchor is, in order: when it was finished, when its appointment is, or
its written due_at; a case with none falls back to when it was opened, so a
course that has not been booked yet still shows dates instead of blanks.
Read and write stay in separate classes — mixing them risks storing an estimate.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The list could be narrowed by status, search and open-date, but not by patient
— so a patient's own file had no way to ask which courses belong to them.
`?record=` adds that bound.
patientRecord is joined once and shared with the search branch; joining it twice
under the same alias is a DQL error, and search already needed it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Two holes in how a course's later appointments were made.
The link from the unbooked queue carried nothing — `/admin/appointments/new`
with no parameters — so the secretary retyped the patient and the service, and
which case the appointment joined was inferred from the service they happened to
pick. A patient with two open courses had no way to say which one they meant,
and picking the wrong service silently opened a third case. (The suggestion link
did pass slot_start and resource_uuid, but the create page never read either.)
POST /api/v1/my/appointment now takes an optional treatment_session_uuid.
SessionBookingLink validates it — same tenant, still unbooked, case open, same
patient — and reserves that session. Confirm-time attachment steps aside when
the appointment already holds a session. The booking form states in words which
session, which course and which patient it is about to book, read from a new
GET /api/v1/treatment-session/{uuid}.
Nothing ever detached a session from its appointment, so a cancelled booking
left the session `booked` forever, and since findNextUnbooked requires
"has no appointment", it could never return to the queue. Cancellation and
no-show now release it back to `planned`. A finished session is history and is
left alone.
The system still never books the next appointment by itself — it only suggests.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The list was cards only, so comparing courses across patients meant reading four
stacked blocks instead of scanning columns. It now has the same table/card
toggle the patients page uses, reusing that page's toggle icons and keeping the
choice in the URL so back and refresh hold it. Table is the default here: the
operator, status and session progress are the columns a manager scans.
Both views derive the operator the same way, through one helper — history
(performed_by) when a session has been done, otherwise the plan
(assigned_staff), labelled as not yet performed.
Also fixes a three-states slip in the operator picker I added last time: it
rendered "no staff defined" while the staff list was still loading, which is
what an empty list looks like from the user's side. Loading now says so.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A treatment case said which doctor supervised it but never who actually did the
work, so the list could not answer the first question a manager asks about a
course: who performed it.
Two separate things now travel with the case. `performed_by` is history —
derived from the sessions' performedBy, so it only ever reports what happened.
`assigned_staff` is plan — a new treatment_case_staff table, editable from the
modal, saying who is meant to handle this patient's course. The card shows the
first and falls back to the second while nothing has been performed yet.
Search matches both. A manager typing an operator's name wants that person's
work, and work already done is part of it.
Assignment also narrows the operator queue: a case with assigned staff shows its
sessions only to those people, because a patient who started a multi-session
course with one operator should keep them. An unassigned case keeps the existing
protocol rule, and an empty list means "anyone the protocol allows" rather than
"nobody" — the same "no rows is not a restriction" convention used elsewhere.
Unlike areas, removing an operator erases nothing: a finished session carries its
real operator on itself and never consults this list.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The assigned-services card and its stat card are gone. The staff role has one
job — today's sessions — and with /admin/my-services already removed, a list of
services the operator cannot act on was the last thing on the page competing
with the work itself.
The remaining stat card no longer stretches the page: `.stat-grid` is built for
four cards and collapses to one full-width column with a single child, so the
staff grid is capped.
The endpoint still returns `services` and `stats.services`; the field is
documented and cheap, and nothing else needed to change to stop rendering it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Skipping an area recorded only that it was skipped. Why it was skipped is
clinical history — the next session needs to read it — so `skip` now takes an
optional note, the same way completing an area already did, and the panel asks
for it inline instead of firing on the first click.
An operator finds out mid-laser that they closed the wrong area, and until now
had to carry that mistake to the end of the session. `reopen` puts a settled
area — completed or skipped — back to in_progress and clears finished_at,
keeping the recorded parameters and note so they can be seen and overwritten.
It stops at the same boundary everything else in this domain stops at: once the
session is finished the record is history, and reopening it is 409.
Also drops /admin/my-services. The staff role has one job — today's sessions —
and the dashboard already lists the services they may perform, so the page was
a second place to read the same list. Route, page, sidebar entry and the two
links to it are gone; the services stat card is no longer a link because it no
longer has a destination.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Every number on it was zero. "نوبتهای امروز من" counted appointments where
appointments.staff_id matches — a column no booking path fills by default, and
which is NULL on every row in the database. "سرویسهای من" read only direct
service assignment, so an operator whose whole job comes from a treatment
protocol was told they had no services. The landing page of the only role that
has one data page said, in effect, that they had nothing to do — while they had
two sessions booked that day.
Today's work now comes from TreatmentSessionRepository::findTodayForStaff, the
same queue rule the sessions page uses, so there is one definition of "my work
today" rather than two that disagree. Services are the union of direct
assignment and protocol authorisation.
The two stat cards are links to the pages they name; a number with no
destination made the user hunt the sidebar for a page the card had just
mentioned. Each row of the work list opens that session. The avatar moves from
its own full-width card into the header — two lines of text were costing a card
and pushing the day's work below the fold on mobile. The assigned-appointments
table renders only when it has rows.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The list could be narrowed by status and by search but not by when a case
opened, which is the one axis a clinic actually reports on. `from` and `to`
(YYYY-MM-DD) now bound `opened_at`, using the same strtotime day-boundary
convention the appointment date filters already use under the app's global
Tehran timezone. A malformed value is ignored rather than erroring — this is a
filter, not a form field.
Both bounds live in the URL via useUrlState, so back and refresh keep the range.
The two date inputs and the "تا" between them are one nowrap unit; letting them
wrap separately orphaned the word from its field on a 390px screen.
The card's "شروع" showed only the Jalali date, so several cases opened on the
same day were indistinguishable on that line too. It now uses formatDateTime.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The list had no way to tell two cases apart. TreatmentCase::toArray() carried
no patient, so four cases of the same service rendered as four identical
cards — same service, same supervisor, same date, same areas. Search would have
been meaningless without fixing that first, so the payload now carries the
patient (name, mobile, record number) and the card leads with the name.
Search: `?q=` on the list endpoint, matching patient name, mobile, national
code, record number and service name — the same keys a secretary already types
into the booking form. It lives in the URL via useUrlState, debounced, so back
and refresh keep the view.
Edit: PATCH /api/v1/treatment-case/{uuid} covering status, supervising doctor,
areas and session count, driven from a modal on the list. Rules live in
TreatmentCaseEditor, not the controller, around one boundary: no edit may
overwrite work already done. An area with session records cannot be removed, and
the session count cannot drop below the sessions that are booked or finished —
both 409, both tested. Reopening a closed case clears closed_at.
`areas[]` now also exposes `category_uuid`; the edit form selects catalog
categories, while `uuid` identifies the snapshot row.
Page fixes from the redesign checklist: the status filter was a hand-rolled
primary/secondary button pair, now `.seg` with `.on`; the raw `<progress>` bar
took the browser's own appearance and ignored the theme tokens, now a token-
styled bar with an explicit progressbar role; session counts go through
formatNumber; a failed request rendered as "no cases found", which reads as an
empty clinic rather than a broken one, and an empty search now says so in its
own words.
Adds the test files neither the page nor the case editor had.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
hamed