Merge PR #29: strip Aura .p-drawer-content padding in mobile drawer
Corrects the mobile drawer spacing: WorkspaceSwitcher stays bottom-anchored
(reverts the PR #28 grow change) and the Aura .p-drawer-content wrapper padding
(~20px sides+bottom) is removed with !p-0 so the block sits edge-to-edge like
the desktop sidebar. Reviewer PASS; 582 frontend tests green; visually
confirmed at 375px.
Correct the mobile drawer spacing fix. The real spacing was Aura's
.p-drawer-content padding (0 {modal.padding} {modal.padding} {modal.padding},
~20px sides+bottom) insetting the whole logo/menu/switcher block from the
drawer edges — plain p-0 lost to it in stylesheet order, so force it off with
!p-0 (same fix pattern as header !hidden). The WorkspaceSwitcher stays
bottom-anchored exactly like the desktop sidebar: revert the SidebarNav 'grow'
prop and the mobile grow=false (the nav is flex-1 again). Children keep their
own intrinsic padding; the drawer content is now edge-to-edge like desktop.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Merge PR #28: group mobile drawer sections (remove bottom-pinned void)
MOBILE-SHELL-PARITY follow-up: in the mobile drawer the nav's flex-1 pushed
WorkspaceSwitcher to the bottom, leaving a void between menu and switcher.
SidebarNav gains a `grow` prop (default true = desktop bottom-anchor); the
mobile drawer passes grow=false (flex-initial) so logo/menu/switcher group at
the top and a long menu still scrolls internally. Plus a Storybook router fix.
Reviewer PASS; 583 frontend tests green.
In the mobile drawer the nav's flex-1 filled the full-height panel, pushing
the WorkspaceSwitcher to the very bottom and leaving a large empty void between
the menu and the switcher. Add a 'grow' prop to SidebarNav (default true =
desktop bottom-anchor) and pass grow=false in the mobile drawer so the nav
takes only its natural height — logo, menu and switcher group together at the
top. Desktop sidebar unchanged. +2 Vitest assertions for the contract.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The story router only knew the pre-v2 route names, so SidebarNav's RouterLink
to the v2-prefixed 'v2-dashboard' threw 'No match for' and the Storybook error
boundary replaced every AppSidebar story. Add 'v2-dashboard' to the story
router's known names.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Merge PR #27: mobile drawer chrome parity with desktop sidebar (MOBILE-SHELL-PARITY)
MOBILE-SHELL-PARITY — mobile drawer chrome parity with the desktop sidebar:
- defect 1: header always expanded on mobile (effectiveCollapsed)
- defect 2: single non-overlapping close control; PrimeVue default X suppressed
(showCloseIcon=false) fixing a focus a11y trap; brand-row X is the control
- defect 3: full-height flex column verified + min-h-0 guard (device check pending)
Plus a tooling fix (.claude/worktrees hook exemption), a CLAUDE.md push/merge
authority policy change, and 9 new Vitest tests. Reviewer PASS; 581 frontend
tests green.
Per Bert's request (2026-06-03). Replaces the 'developer pushes manually'
rule with a Push & Merge Authority policy: Claude may push feature branches,
open Gitea PRs, and merge them without a separate approval click, provided
the merge gate is green (reviewer PASS, tests/lint/typecheck clean, backend
guards where applicable). Never push directly to main, never force-push,
always --no-ff via a reviewed branch, delete the branch post-merge. CLAUDE.md
supersedes the build-module skill's HUMAN GATE 2.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Review follow-up (crewli-reviewer). With header:'!hidden' the default close
button was still mounted (display:none); PrimeVue Drawer.focus() falls back to
focusing it (no [autofocus] in our slots), a no-op that strands keyboard focus
behind the overlay. Set :show-close-icon="false" so the button is never
created — focus() no-ops cleanly and v-focustrap focuses the visible brand-row
X. Also: flip the previously tautological showCloseIcon test to assert the
decision (toBe(false)); align the mobile close aria-label to English 'Close
menu' (matching 'Collapse sidebar') for locale consistency.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Defects 1 (logo) and 2 (close-X overlap) fixed and tested. Defect 3
(WorkspaceSwitcher) audited as already-correct full-height flex column with a
min-h-0 guard; not device-verified, dynamic-viewport-height flagged as the
remaining suspect. Records the resolution per the GRADIENT-BRAND-ALIGNMENT
convention.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
MOBILE-SHELL-PARITY subtask 5. Automated visual evidence via Playwright-CT is
not feasible — the CT harness loads only Vuetify styles (no Tailwind / PrimeVue
theme / PrimeVue plugin), so an AppSidebar screenshot renders unstyled. Add a
MobileDrawer Storybook story instead (Storybook DOES register PrimeVue + load
Tailwind via preview.ts) as the manual visual-verification target for all three
defects: single non-overlapping close X, expanded brand row/logo, and the
bottom-anchored WorkspaceSwitcher. Narrow the window <1024px to mount the drawer
(no viewport addon in this repo's Storybook).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
MOBILE-SHELL-PARITY. Adds 7 Vitest assertions for the three defects:
- AppSidebar: header pt is '!hidden' (default close-X suppressed), content pt
is a full-height flex column (flex-col/h-full/min-h-0), showCloseIcon is not
forced false, and WorkspaceSwitcher renders inside the drawer.
- SidebarHeader: the Icon stub now exposes data-icon; mobile brand-row control
is an explicit close (aria-label 'Sluit menu', tabler-x), desktop stays the
collapse chevron, and the header renders expanded on mobile even when
sidebarCollapsed is true (logo parity).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
MOBILE-SHELL-PARITY defect 1. SidebarHeader read shell.sidebarCollapsed
directly, so a desktop-collapsed state made the full-width (256px) mobile
drawer render a collapsed brand row (logo-only, no wordmark, expand-chevron
row) while AppSidebar forces SidebarNav + WorkspaceSwitcher to expanded —
the mismatch read as incorrect logo placement. Gate all collapse-dependent
branches on a new effectiveCollapsed computed (!isMobile && sidebarCollapsed):
desktop honours the collapse state, mobile is never collapsed so the drawer
header matches the expanded nav/switcher. Also keeps the subtask-1 close X
(on the expanded-row control) reliably visible on mobile.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
MOBILE-SHELL-PARITY defect 3. Static audit (code + @primeuix base CSS) shows
the mobile drawer content is ALREADY a correct full-height flex column:
.p-drawer-left .p-drawer-content gets height:100% + flex-grow:1 from base
styles, our pt makes it flex flex-col, SidebarNav (flex-1 min-h-0) claims the
slack and WorkspaceSwitcher (flex-shrink-0) anchors at the bottom — matching
the desktop aside. Add min-h-0 as flex hygiene (children shrink, nav scrolls
internally) and document the chain in-code, citing the @primeuix rules.
The 'broken height chain' premise did not match the code, so no redundant
height fix was fabricated. If WorkspaceSwitcher is still clipped on a real
mobile device, the remaining suspect is dynamic-viewport-height (a shrinking
address bar making 100% exceed the visual viewport) — flagged for a
real-device visual check at the merge gate.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
MOBILE-SHELL-PARITY defect 2. PrimeVue's default drawer header close-X was
leaking through 'header: hidden' (non-important 'hidden' lost to PrimeVue's
base .p-drawer-header display in stylesheet order) and overlapping the brand
row. Force-hide it with the important variant '!hidden' (matching the file's
existing !w-64), and provide the mobile close affordance in SidebarHeader's
brand row: on mobile the top-right control renders an X ('Sluit menu') and
closes the drawer (handleCollapseClick already calls setMobileOpen(false) on
mobile), mirroring the slot the desktop sidebar uses for its collapse chevron.
Single, non-overlapping close control; showCloseIcon is left at its default.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Isolated subagents (frontend/backend-implementer) run in a git worktree
created under .claude/worktrees/. The protect-files PreToolUse hook blocked
all edits to any .claude/ path, making every worktree-isolated agent unable
to edit its own checkout. Exempt .claude/worktrees/ (ephemeral agent scratch
space) while still protecting the real tooling config under .claude/.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Plan 2.5 final phase (P8). Closes the PrimeVue shell-migration workstream.
- RFC-WS-PRIMEVUE-PLAN-2-5: added Supersessions section recording the
governing-RFC divergences (§4 dark mode `<body>`→`<html>` per AD-2.5-D1;
§7.4 workspace sub option A reversed to placeholder after visual
review). Added closure summary (phases, ADs, brand-square recipe,
suite delta, lessons). Status → COMPLETE.
- BACKLOG: landed 8 items surfaced during Plan 2.5 (MOBILE-SHELL-PARITY,
WORKSPACE-DROPDOWN-SUB-CONTENT, DENSITY-AWARE-SPACING, TOPBAR-H-VAR-
DECLARE, CSP-FONT-SRC-LOCKDOWN, AUTO-IMPORTS-V2-SCAN, PNPM-RESOLUTIONS-
ROOT, SHELLUI-STALE-DATA-THEME-CLEANUP). Marked GRADIENT-BRAND-
ALIGNMENT as resolved.
- useShellUiStore.toggleDensity: removed redundant applyDomAttributes()
call (the AppShellV2 watch already covers density changes). Moved the
DOM-write assertion to AppShellV2 watcher-coverage specs.
Plan 2.5 status: COMPLETE.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds crewli-architect, backend/frontend-implementer, test-writer subagents,
the /build-module orchestrator command, the PR merge-gate template, and a
permissions allow-list in settings.json. Documents the layer as
CLAUDE_CODE_TOOLING.md section 10. Implementer Edit/Write is allow-listed;
git push deliberately omitted so merge/push stay human.
Co-Authored-By: Claude <noreply@anthropic.com>
Holistic parity pass after repeated piecemeal switcher tweaks kept
diverging from the header logo. Side-by-side audit found size, offset
and collapsed-centering already matched (both 32px at 16px); the real
divergences were (a) the inset-shadow opacity (logo 8% vs avatar 10%)
sourced from TWO separate scoped-CSS blocks (drift risk), and (b) the
hover-bg inset was only 8px ("stuck against the edge").
- Shared brand-square recipe: both the SidebarHeader logo and the
WorkspaceSwitcher avatar now use the SAME Tailwind utilities
`w-8 h-8 rounded-lg shadow-[inset_0_-2px_0_#00000014]`. Single
source (the utility classes) so size/radius/shadow can't drift
again. The two per-component scoped `.mark` / `.ws-logo-square`
box-shadow rules are deleted (the dropdown's larger
`.ws-logo-square-lg` stays scoped — out of scope). Only the gradient
differs by design (brand teal vs per-org).
- Breathing room: the avatar's horizontal offset is pinned at 16px by
the collapsed rail (64px = 32px square + 2×16px → the only offset
that centres the square AND matches the logo's px-4). Within that
fixed 16px, the budget is split inset + internal padding: wrapper
px-3 (12px hover-bg inset for breathing room, was 8px) + trigger
px-1 (4px internal). Vertical is unconstrained → py-2 both for a
generous hover-bg height. The offset stays 16px so logo↔avatar
parity and the no-jump invariant are preserved; only the hover-bg
inset grew from 8px to 12px.
Note: the prompt's 20px-offset option is incompatible with the fixed
64px collapsed rail (20+32+20≠64 → breaks centring + reintroduces a
jump), so the 16px-offset / 12px-inset path was taken per the brief's
stated alternative.
Specs: new cross-component parity spec mounts BOTH components and
asserts the avatar + logo share the exact w-8/h-8/rounded-lg/shadow
utilities; padding spec updated to px-3 wrapper + px-1 trigger.
Borderless + hover/open-bg + sub specs retained.
Suite delta: 571 → 572 (+1). vue-tsc clean. Scoped ESLint clean
(0 errors). Desktop only.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Matches the crewli-starter SoT and fixes the recurring collapse jump at
its root cause. The prior structures left a residual avatar shift:
- the original split put the avatar at 24px expanded (wrapper px-4 +
card p-2) vs 16px collapsed (bare square) — an 8px horizontal jump;
- the interim single-trigger variant used wrapper p-[10px] + trigger
px-[10px] expanded (~20px) vs justify-center collapsed (16px) — a
~4px residual horizontal shift.
Unified both states to a single symmetric structure:
avatar offset = wrapper px-2 (8px) + trigger p-2 (8px) = 16px
16px from the rail's left edge in BOTH states — identical to the
SidebarHeader brand logo. Because the padding is symmetric (8 + 8 each
side) and the collapsed rail is 64px = 16 + 32 + 16, the left-aligned
avatar is also visually centred when collapsed — no justify-center,
no px swap, no horizontal shift; constant vertical padding, no vertical
shift. The jump is gone at the root.
Borderless: the trigger has NO border in any state (the prior is-open
border is dropped per the starter screenshots). The only divider is the
wrapper's border-t between the switcher and the nav. The grey
background is the sole fill — transparent at rest, grey on hover, and
grey while the popover is open (isOpen wired to Popover @show/@hide).
The trigger's p-2 gives the grey background generous padding around the
avatar+text, matching the starter's hover treatment, and since it is
the button's own background it never moves the content.
Specs reworked: trigger p-2 identical across states (no px swap / no
justify-center — the no-jump lock), wrapper carries p-2, trigger is
borderless at rest AND while open, open-state grey background applies
on @show and clears on @hide. Single-.trigger / rounded-lg / collapsed-
hides-meta+chev / sub-line specs retained.
Suite delta: 571 → 571 (specs reworked, count unchanged). vue-tsc
clean. Scoped ESLint clean (0 errors). Desktop only.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Diagnosed against the crewli-starter SoT
(crewli-starter/src/components/layout/WorkspaceSwitcher.vue +
main.css .ws-switcher rules). Three issues Bert flagged in manual
smoke, one root cause:
ROOT CAUSE — the prior version swapped between TWO separate <button>
elements on collapse (a bare collapsed button vs a padded expanded
trigger) with different box models. Their vertical padding differed,
so the avatar's distance from the rail bottom changed by ~6px and
visibly jumped on collapse. crewli-starter instead renders ONE
`.trigger` button in both states and only hides `.meta` + `.chev`
(+ recentres) on collapse.
Fixes (all by adopting the single-trigger structure):
- Avatar no longer jumps: one `.trigger` button always; `py-2`
vertical padding constant across states; only horizontal
padding/justify changes on collapse. Avatar's vertical box is now
identical collapsed vs expanded.
- No resting border: trigger is `border-transparent` at rest and on
hover (hover only adds the grey bg), matching crewli-starter
`.trigger:hover { background }`. The wrapper has only the `border-t`
separator, no box border.
- is-open persistence: new `isOpen` ref synced from the PrimeVue
Popover `@show`/`@hide` events. While the dropdown is open the
trigger keeps the grey bg AND shows a visible border, matching
crewli-starter `.ws-switcher.is-open .trigger { background;
border-color }`. Persists until the popover closes (covers
programmatic hide via selectOrg + outside-click dismissal).
- Hover padding: trigger `px-[10px] py-2` inside a `p-[10px]` wrapper
reproduces crewli-starter's generous hover-card inset
(`.ws-switcher { padding:10px } .trigger { padding:8px 10px }`).
Collapsed alignment preserved: wrapper p-[10px] + trigger
`justify-center px-0` centres the 32px avatar at 16px from the rail's
left edge — still aligned with the SidebarHeader brand logo (px-4).
At rest the collapsed trigger is transparent, so it still reads as a
bare square mirroring the header logo; hover/open add the card.
Specs: replaced the now-obsolete "bare avatar button / no .trigger"
+ "collapsed wrapper px-4" locks with crewli-starter-structure specs:
single .trigger in both states, collapsed centres the lone avatar +
hides .meta/.chev, py-2 constant (the no-jump lock), and is-open
keeps bg+border. Sub-line + dropdown specs unchanged.
Suite delta: 569 → 571 (+2). vue-tsc clean. Scoped ESLint clean
(0 errors). Desktop only.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Reverses AD-2.5-W1 option A (no-sub) after visual review against the
crewli-starter SoT — the two-line layout reads better. Adds a
light-grey placeholder sub line under the workspace name (expanded
trigger + dropdown items); collapsed stays bare-avatar-only. No
backend: the placeholder is a neutral static string ('Organisatie'),
real org type + metrics still deferred under
WORKSPACE-DROPDOWN-SUB-CONTENT. The org object exposes no field that
reads well as a subtitle today (id/name/slug/role only; role is an
access identifier, not a description), so a neutral string is used
rather than fabricated metrics or the role string P4 originally
removed.
Fix A — avatar unified across collapsed/expanded. The collapsed
avatar styling previously lived directly on the <button>, letting
user-agent button rendering diverge subtly from the expanded <span>
avatar. The collapsed render now wraps the EXACT SAME avatar span
markup (same classes, gradient, .ws-logo-square inset-shadow) in a
bare transparent p-0 button — the 32px square is byte-identical
across states; only the surrounding context differs.
Fix B — sub line re-added to WorkspaceDisplay (cleanly typed as
`sub: string`, sourced from a SUB_PLACEHOLDER const with a TODO
pointing at the deferred backend). Rendered light-grey
(text-[var(--p-text-muted-color)], matching the in-component muted
text) at text-[11.5px] in the trigger and text-[12.5px] in dropdown
rows, mirroring the pre-P4 sizes. Collapsed renders no sub.
Specs: reversed the P4/P5 no-sub locks to sub-present assertions
(trigger sub present + honest-placeholder/not-role; dropdown sub on
every row + no role leak; collapsed-no-sub via `.meta .sub`).
Updated the collapsed-bare-avatar spec for the new span-in-button
structure (.ws-logo moved from button to inner span).
Suite delta: 566 → 569 (+3). vue-tsc clean. Scoped ESLint clean
(0 errors). No backend, no fabricated data.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
P7 token audit (8330e93f §10) found two off-brand color bugs:
- utils/v2/gradient.ts GRADIENT_PALETTE used Tailwind blue-green
anchors (teal-500/600 #0d9488, cyan, emerald, sky) clustered in a
single hue family. Two problems: the brand-anchor slot used Tailwind
teal #0d9488, NOT Crewli's #0D9394, AND orgs in multi-workspace
views all rendered as similar teal/green squares (poor
distinguishability). Replaced with the crewli-starter SoT palette:
[0] #0D9394 → #075F60 (Crewli teal — brand anchor)
[1] #7C3AED → #4C1D95 (purple)
[2] #EA580C → #9A3412 (orange)
[3] #16A34A → #14532D (green)
[4] #F59E0B → #92400E (amber)
[5] #EC4899 → #9D174D (pink)
[6] #4F46E5 → #312E81 (indigo — added for org-distinguishability)
[7] #E11D48 → #881337 (rose — added for org-distinguishability)
Palette stays 8-entry; only the values change. Indexing logic
(djb2 hash % 8) unchanged. Per-org avatar colors are not persisted
pre-launch, so the slot reshuffle is safe.
- AppTopbar.vue user-avatar gradient (two sites: the trigger Avatar +
the user-menu header Avatar). Fallback in the CSS var was #0d9488
(Tailwind teal-600), NOT Crewli #0D9394 — if the var ever fails to
resolve, the chrome would render off-brand. Fixed to #0D9394.
The hardcoded pink #f472b6 in the gradient's from-color was kept
intentionally: it matches the crewli-starter SoT (user avatars are
a pink/purple gradient distinct from workspace chrome's teal — the
visual contrast between "your account" and "your workspace" is
by design).
Regression locks:
- gradient.spec.ts +2 specs: brand-anchor slot is #0D9394 (and
defensively, #0d9488 must not appear anywhere in the palette);
palette spans diverse hue families (purple + orange present
beyond the teal anchor).
Suite delta: 564 → 566 (+2). vue-tsc clean. Scoped ESLint clean
(0 errors, pre-existing warnings only).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Manual smoke showed the expanded trigger flush against the rail edges:
the wrapper used px-4 when collapsed but only p-2 when expanded.
Unified the expanded wrapper to `px-4 py-2`, matching the collapsed
state and the header brand row, so the rounded trigger card is inset
16px from both rail edges with symmetric breathing room consistent
with the header's px-4 rhythm.
Trade-off documented in the wrapper comment: with px-4 wrapper + p-2
internal trigger padding, the expanded avatar sits ~24px from the
rail edge — ~8px deeper than the bare header logo at 16px. Accepted
as deliberate (the expanded trigger reads as a distinct rounded
button card). The bare-square / logo mirror lives in the collapsed
state, which is unchanged.
No spec changes — the prior assertions cover the collapsed wrapper
(px-4 + h-[56px]) and the expanded trigger's rounded-lg shape, both
still hold. Suite delta: 564 -> 564.
Desktop only. vue-tsc clean. Scoped ESLint clean (0 errors).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Manual smoke after the logo-anchor fix (63300e5f) showed the footer
WorkspaceSwitcher used different spacing and a boxed avatar vs the
header logo. Applied the identical alignment recipe so the workspace
avatar mirrors the brand logo in both states (Option B from the
brief — bare square when collapsed, padded trigger only when
expanded).
Changes (WorkspaceSwitcher.vue only):
- Wrapper toggles `h-[56px] flex items-center px-4` when collapsed
(mirrors the SidebarHeader brand row) and stays `p-2` when
expanded (room for the padded trigger).
- Collapsed: a BARE rounded-lg avatar button at the same 16px left
offset as the header logo. No `.trigger` container, no rounded
hover-bg box wider than the avatar — the button IS the visible
square (`.ws-logo .ws-logo-square w-8 h-8 rounded-lg`). True
top/bottom mirror of the brand square.
- Expanded: unchanged padded `.trigger` button with avatar + name +
chevron + hover bg. Avatar's left offset stays at 16px from the
rail (wrapper p-2 + trigger p-2) so the expanded avatar also
lines up vertically with the header logo.
Same alignment equation as the header recipe:
rail_collapsed (64px) = square (32px) + 2 × px-4 (2 × 16px)
In both states the avatar's left edge sits at x=16px from the
rail's left — identical to the brand logo above. Vertical line
down the left side now reads as a single column of squares.
Desktop only. Mobile drawer chrome stays as MOBILE-SHELL-PARITY.
Tests adapted:
- `expanded trigger uses rounded-lg` (was tested in both states; the
collapsed render no longer has a `.trigger` container).
- `expanded trigger has no justify-center` (split from the
prior two-state assertion).
- New: `collapsed renders a bare avatar button (no .trigger
container, just .ws-logo)` — locks the bare-square contract.
- New: `collapsed wrapper uses px-4` — locks the
centring-equation invariant (rail=square+2×px-4) against
accidental wrapper-padding regressions.
Suite delta: 563 → 564 (+1 net: +2 new collapsed-shape asserts,
−1 redundant two-state assert).
vue-tsc clean. Scoped ESLint clean (0 errors, pre-existing
warnings only). Manual smoke pending Bert — draw a vertical line
down the rail's left edge and verify the brand square and the
workspace square left edges sit on it in both states; in collapsed
mode verify the avatar is a bare square (no boxed button), same
visual treatment as the bare logo above.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Refines the prior sidebar styling commit (8e166512) after manual smoke:
- Brand logo + workspace avatar: rounded-xl -> rounded-lg (crisper
square per design review; both stay unified at the same radius).
- Logo no longer jumps on collapse. The previous code toggled
`justify-between` ⇄ `justify-center` on the header row, which
re-centred the logo against the parent's width — and the parent
width animates from 256px to 64px over 200ms, so the logo slid
from x≈112px (centred in the expanded rail) to x=16px (centred in
the collapsed rail). Visible jump.
Fix: the brand row is now ALWAYS `px-4` and left-aligned. The
logo's horizontal offset (16px from the rail's left edge) is
identical in expanded and collapsed states. Why this still looks
centred when the rail collapses:
rail_collapsed (64px) = logo (32px) + 2 × px-4 (2 × 16px)
With those numbers aligned, a stationary left-aligned logo IS
visually centred in the 64px-wide collapsed rail. The width
transition then "slides the rail closed around" the anchored
logo. Wordmark + Beta badge sit to the RIGHT of the logo and
v-if-disappear on collapse; their absence doesn't shift the logo
because they were never to its left.
- Toggle chevron placement:
- Expanded: collapse chevron (◀) inline at the right of the brand
row, pushed by `ms-auto` (NOT by justify-between forcing the
layout to recentre the logo).
- Collapsed: a SECOND row below the brand row holds a centred
expand chevron (▶) button. Replaces the prior tucked-chip that
overlapped the logo. No overlap, no overhang needed against the
aside's `overflow-hidden`.
- WorkspaceSwitcher trigger: same anchor-on-the-left treatment
applied. The `justify-center` switch on collapse is gone (it
caused an identical avatar slide). Wrapper padding `p-[10px]` ->
`p-2` so the avatar's left offset (wrapper 8 + trigger 8 = 16px)
matches the SidebarHeader logo (px-4 = 16px) — the brand square
and the workspace square are now vertically aligned in the
collapsed rail.
Desktop only. Mobile drawer chrome tracked separately as
MOBILE-SHELL-PARITY.
Tests adapted:
- WorkspaceSwitcher.spec.ts: trigger-rounded assertion bumped
rounded-xl -> rounded-lg; +1 spec locks "trigger never carries
justify-center" (avatar-anchored invariant).
- SidebarHeader.spec.ts: collapsed-behaviour spec rewritten — was
asserting `justify-center`, now asserts the row carries `px-4`
WITHOUT either justify-center OR justify-between (the actual
anchor contract); +1 spec confirms the expand chevron lives in a
SIBLING row of the brand mark (no overlap).
Suite delta: 561 -> 563 (+2). vue-tsc clean. Scoped ESLint clean
(0 errors, pre-existing warnings only).
Manual smoke pending Bert: collapse/expand slowly and watch the
logo — must NOT move horizontally. Confirm rounded-lg looks crisp
(not pebble-soft). Confirm expand chevron sits in its own row
below the logo, no overlap.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Plan 2.5 P6 follow-up. Closes two desktop shell-parity gaps from P6
manual smoke against crewli-starter SoT:
- Sidebar logo decoupled from the collapse toggle. Expanded layout is
now justify-between (brand group left, collapse chevron right) and
collapsed layout is justify-center with the logo alone. The expand
affordance becomes a small absolute-positioned circular button at
the rail's right edge — solid background + border so the slight
overlap with the centred logo reads as a tucked-in chip rather than
a collision. Toggling collapsed no longer shifts the logo.
- Brand square (SidebarHeader logo) and workspace avatar
(WorkspaceSwitcher trigger) unified to the same rounded square
(h-8 w-8 rounded-xl). Existing sizes were already consistent at
32px — radius bumped from rounded-lg (8px) / var(--p-border-radius)
(~6px) to rounded-xl (12px) per the design direction. Collapsed
rail now reads as a vertical mirror: brand square at the top,
avatar square at the bottom, bracketing the nav icons.
- WorkspaceSwitcher trigger restyled: rounded-xl (was the sharper
var-radius), p-2 (was px-[10px] py-[8px]), hover background. The
collapsed-variant gating of name + chevron is unchanged from P5.
Edge-mounted overhang past the rail edge was not possible: the aside
carries `overflow-hidden` (intentional, for the w-64 ⇄ w-16 width
transition) which clips anything past the rail edge. The tucked-chip
pattern (24px circle at end-0, solid bg) is the visual compromise —
the affordance stays inside the rail, discoverable, and visually
decoupled from the logo.
Desktop only. Mobile drawer chrome (logo placement, drawer X button,
missing switcher) tracked separately as MOBILE-SHELL-PARITY.
Tests:
- +2 WorkspaceSwitcher.spec.ts: trigger uses rounded-xl; collapsed
trigger renders avatar only (hides .meta).
- +1 SidebarHeader.spec.ts: collapsed row hides the .brand-name
wordmark and toggles to justify-center (logo-stays-centred lock).
Suite delta: 558 → 561 (+3). vue-tsc clean. Scoped ESLint clean
(0 errors, pre-existing warnings only).
Manual smoke pending Bert: collapse the rail, verify logo stays put
and the expand chip appears at the right edge; verify the trigger
shows rounded corners + hover bg; verify the collapsed avatar
mirrors the brand square size/radius.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Per RFC-WS-PRIMEVUE-PLAN-2-5 §5.6–§5.10. Final code phase of Plan 2.5
before closure docs (P7 tokens, P8 closure).
Changes:
- Fix 9: sidebar full-height. The desktop <aside> now carries
`h-screen sticky top-0` so it fills the viewport vertically and
pins to top on body scroll. Without this the aside sized to its
children's intrinsic heights (~250-400px) and ended mid-viewport
even though the surrounding grid row stretched to 100vh. With
h-screen, SidebarNav's `flex-1` claims the remaining column space
and WorkspaceSwitcher anchors to the true viewport bottom — its
`border-t` (existing from P5) is now the divider above the
switcher per crewli-starter. Mobile Drawer untouched (PrimeVue's
internal pt classes already give it 100% panel height).
- Fix 10: density toggle promoted to the store. New
useShellUiStore.toggleDensity() flips comfortable ⇔ compact and
calls applyDomAttributes() synchronously. AppTopbar's local
toggleDensity wrapper deleted; the button now invokes
shell.toggleDensity() directly and carries a stable
data-testid="density-toggle" plus a `title` matching its
aria-label. Density icons swapped from generic flex-alignment
glyphs (tabler-layout-distribute-{vertical,horizontal}) to the
literal density metaphor (tabler-baseline-density-{small,medium}).
Both new icons verified present in the loaded
@iconify-json/tabler set. Topbar right-side order
(search → density → dark → notifications → user) was already
correct from P5; locked with a new ordering spec.
Verified (no code change):
- Fix 6 (§5.6): dark mode `.dark` on <html> confirmed in
useShellUiStore.applyDomAttributes (AD-2.5-D1, P3 complete).
Component-level dark coverage remains a separate backlog item
(DARKMODE-V2-COVERAGE).
- Fix 8 (§5.8): the ▼ arrow is the Vue DevTools v8.0.2 dev-only
toggle button injected by the devtools vite plugin, not Crewli
code — diagnosed, no action.
- Fix 7 (§5.7): non-reproducible at code level. Topbar is
`sticky top-0` and is a SIBLING flex item of <main> inside the
shell's flex-col right column; normal flow stacks <main> below
the topbar at first paint, so the title cannot fall behind a
sticky topbar in this composition. Documented as no-op; if
Bert reproduces it after Fix 9 lands, the symptom is something
else (likely a per-page negative margin or a separate scroll-
container interaction worth its own ticket).
Density enum corrected against runtime data-density: 'comfortable'
(not 'comfy' — the earlier RFC assumption is wrong; the store has
always typed `'comfortable' | 'compact'`).
Tests:
- +2 useShellUiStore.spec.ts: toggleDensity flips comfortable ⇔
compact AND writes data-density via applyDomAttributes;
toggleDensity from compact returns to comfortable on call 2.
- +2 AppTopbar.spec.ts: density button reachable by
data-testid="density-toggle"; topbar right-side order locked
via HTML index comparison (search → density → dark → notif →
user). Existing density-flip specs adapted to spy on
toggleDensity (the new direct call site).
Suite delta: 554 → 558 (+4). vue-tsc clean. Scoped ESLint clean
(0 errors, pre-existing warnings only).
Manual smoke pending Bert:
1. Sidebar full-height, switcher pinned to viewport bottom (Fix 9)
2. Page title clears topbar (Fix 7 — expected no change needed)
3. Density toggle visible between search and dark with the
density icon (Fix 10)
4. Click density toggle → spacing visibly changes, <html
data-density> flips between comfortable and compact (Fix 10)
5. Topbar order: search → density → dark → notifications →
avatar (Fix 10)
6. Dark mode still toggles (Fix 6 regression)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
P2 reverted CSS font-family to Inter but missed the JS font-loading
path: src/plugins/webfontloader.ts loaded Public Sans from Google
Fonts via WebFont.load(). The wf-publicsans-n4-active class on <html>
(found during P5 manual smoke) proved Public Sans was still loaded at
runtime, plus an external Google Fonts CDN request — both contrary to
AD-2.5-T1 (local @fontsource/inter, no CDN).
Audit context: the plugin was auto-registered via the Vuexy
registerPlugins() glob (src/@core/utils/plugins.ts walks
plugins/*.{ts,js}). No explicit import / call site to delete — file
removal is enough. The plugin only ever loaded Public Sans (no other
families), so full deletion is correct.
Changes:
- Removed src/plugins/webfontloader.ts (auto-registration falls away
with the file itself; no manual unregister needed).
- Removed webfontloader (1.6.28) + @types/webfontloader (1.6.38) from
package.json / pnpm-lock.yaml.
- Strengthened tests/unit/styles/typography.spec.ts with a new
describe block that scans every src/plugins/*.ts for: any
webfontloader reference, any WebFont.load call, any "Public Sans"
spelling, any fonts.googleapis.com URL. Plus a regression-lock
spec asserting webfontloader.ts itself stays deleted.
Suite delta: 552 → 554 (+2 new JS-path specs). vue-tsc clean.
Scoped ESLint clean (0 errors).
Manual smoke pending (Bert): hard-reload /v2/dashboard, confirm
- wf-publicsans-* and wf-active classes are gone from <html>
- computed font-family on body text starts with "Inter"
- Network tab has no fonts.googleapis.com request
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Follow-up to P5 (commit ac36dfe9). Vue warned about unresolved
AppBreadcrumb component in AppTopbar's #start slot — auto-import via
unplugin-vue-components did not register it because
components-v2/ is outside the scan path (Components({ dirs: [...] })
covers src/@core/components, src/views/demos, src/components only).
The original P5 edit did include this import line, but a formatter
pass appears to have pruned it as "unused" before runtime parsed the
template; the symbol was unresolved at render and the warning
surfaced. Restored explicitly so any future formatter pass keeps it.
Fix 2 (AppTopbar #start = AppBreadcrumb) now functions visually,
not just structurally. Manual smoke pending (Bert).
Follow-up backlog: AUTO-IMPORTS-V2-SCAN — extend Components({ dirs })
to include src/components-v2/ so the v2 chrome can rely on the same
auto-import ergonomics as v1. Not done here to keep the fix surgical.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Per RFC-WS-PRIMEVUE-PLAN-2-5 §5.1–§5.5 plus the AD-2.5-W1 option-A
supersession (no sub on dropdown items either, accepted divergence).
Atomic changes:
- AppTopbar: brand block (gradient "C" mark + Crewli wordmark) removed
per Fix 1; the #start slot now renders <AppBreadcrumb /> per Fix 2.
Legacy meta-based useBreadcrumb consumption (breadcrumbModel computed,
vue-router useRouter import, command-based PrimeVue Breadcrumb model)
is gone; AppBreadcrumb owns the registry-driven path. Dead
topbar-mark-shadow scoped CSS rule deleted.
- AppBreadcrumb: import updated to the renamed useBreadcrumb.
- AppSidebar: docstring updated to make the Fix 3 vertical order
(Header → Nav → Switcher, switcher bottom-anchored) explicit. No
template change needed: SidebarNav's root <nav class="flex-1"> already
fills available column space, naturally pushing WorkspaceSwitcher to
the bottom (two flex-1 siblings would split the column 50/50 and
compress the nav — a separate spacer element is structurally wrong).
- WorkspaceSwitcher: dropdown panel restructured per crewli-starter
reference. Semantic class markers (.popover-head/.title/.link/.list/
.opt/.is-current/.ws-logo/.name/.check-mark/.foot) added alongside
Tailwind utilities so specs assert structure with stable selectors.
Footer buttons wired to placeholder createWorkspace / inviteUser
handlers (console.warn + TODO) until the flows ship. Manage link
stays a non-navigating label (no v2-workspaces-manage route yet).
No sub line on any dropdown row (AD-2.5-W1 option A).
Atomic legacy useBreadcrumb retirement (planned since P1):
- Legacy route-meta-driven useBreadcrumb + toBreadcrumbItems +
BreadcrumbRouteRecord types deleted entirely (only AppTopbar
consumed it, and that consumption is gone after Fix 2).
- useNavBreadcrumb → useBreadcrumb (single SoT for breadcrumb chain).
- NavBreadcrumbItem → BreadcrumbItem.
- AppBreadcrumb.vue import updated to the new name.
- SidebarNav.vue docstring reference scrubbed to the new name.
- useBreadcrumb.spec.ts: 10 legacy toBreadcrumbItems specs removed;
4 walkNavTree specs retained.
AppTopbar.spec.ts:
- vue-router mock simplified (route.matched no longer relevant).
- AppBreadcrumb stubbed in #start; legacy command-vs-route assertion
removed; new spec verifies AppBreadcrumb is rendered.
WorkspaceSwitcher.spec.ts: 5 new dropdown specs (header / row count /
current-row checkmark / footer buttons / no-sub on rows).
Suite delta: 557 → 552 (−5 net: −10 legacy toBreadcrumbItems specs,
+5 Fix 5 dropdown specs, −1 obsolete AppTopbar breadcrumb-model spec,
+1 new AppTopbar AppBreadcrumb-presence spec).
vue-tsc clean. Scoped ESLint clean (0 errors). All 3 re-grep checks
returned 0 hits (useNavBreadcrumb/NavBreadcrumbItem, topbar brand
selectors, standalone "sub" identifier in WorkspaceSwitcher — only
documentation comments referencing the no-sub state remain, which
describe absence by design).
Manual smoke skipped (Auto Mode); coverage from the post-edit specs
includes AppBreadcrumb-in-#start, dropdown structure, and trigger
no-sub. Recommend Bert run `pnpm --filter crewli-app dev` and verify
the 6 checks listed in the prompt before merging.
Known divergence from crewli-starter (accepted):
- Dropdown rows are ~16px shorter than crewli-starter (no sub line).
Tracked as WORKSPACE-DROPDOWN-SUB-CONTENT for a future RFC with
the required backend scope (organisations.type enum + metrics).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
P4 (Plan 2.5, AD-2.5-W1 + AD-2.5-B1) refactored SidebarNav to read
APP_NAVIGATION directly, retiring the OrganizerLayoutV2 → useV2Nav →
AppSidebar :groups → SidebarNav :groups props chain. Five artifacts
were deliberately left in place to keep the P4 diff focused — this
commit removes them.
Deleted:
- src/composables/useV2Nav.ts (+ spec) — v1→v2 nav fold adapter, no
production consumer post-P4
- src/types/v2/nav.ts — V2NavGroup / V2NavItem types, only consumed
by the deleted composables above. types/v2/ directory removed (empty)
- src/components-v2/layout/sidebarNavActive.ts (+ spec) — pure helper,
SidebarNav now uses inlined active check against NavItem.routeName
- navFixture export + V2NavGroup import from stories/v2/_helpers.ts
Also: stale "useV2Nav(orgNavItems)" reference scrubbed from
OrganizerLayoutV2.vue docstring (the function no longer exists; the
comment now describes the retired plumbing generically).
Suite delta: 575 → 557 (−18 specs). The drop is correct — the removed
specs tested deleted dead code (sidebarNavActive: 8 specs, useV2Nav:
10 specs), not contract behaviour.
vue-tsc clean. Scoped ESLint clean (0 errors). Final re-grep on all
deleted symbols (useV2Nav, V2NavGroup, V2NavItem, sidebarNavActive,
navFixture) returns zero hits across apps/app/src/.
Per zero-compromise gap 5 (delete > adapt): orphans don't stay.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Per RFC-WS-PRIMEVUE-PLAN-2-5 §4 AD-2.5-W1 and AD-2.5-B1, §5.4 Fix 4.
Changes:
- WorkspaceSwitcher: sub field removed from template, WorkspaceDisplay
type, and buildDisplay derivation. Stories did not carry sub args
(auto-derived from seeded org.role); no WithSub story existed. New
regression spec (WorkspaceSwitcher.spec.ts) locks the no-sub render.
- SidebarNav: now consumes APP_NAVIGATION from src/config/navigation.ts
as the single source of truth (shared with breadcrumb derivation in
useNavBreadcrumb). The groups: V2NavGroup[] prop is removed; render
walks top-level NavItems (branch nodes render label-heading + children;
leaf nodes render as rows; items without routeName render as
non-clickable dormant placeholders). Previous nav data source:
groups prop fed by useV2Nav(orgNavItems) in OrganizerLayoutV2.
- APP_NAVIGATION expanded with 7 entries to preserve visual sidebar
continuity (Evenementen at top-level + Beheer branch with 5 children).
All new entries use routeName: undefined until the corresponding v2
page lands (TODOs noted per entry); only Dashboard maps to v2-dashboard.
- AppSidebar: groups prop removed; passes only :collapsed to SidebarNav.
- OrganizerLayoutV2: useV2Nav(orgNavItems) plumbing retired; the layout
now renders <AppSidebar /> with no nav-data wiring.
- Tests: AppSidebar.spec drops the "passes groups prop to SidebarNav"
assertion; OrganizerLayoutV2.spec drops the "forwards orgNavItems"
assertion. New WorkspaceSwitcher no-sub regression spec (+2 tests).
- Storybook: SidebarNav.stories and AppSidebar.stories updated to no
longer thread navFixture/groups; WithActiveItem pushes v2-dashboard.
Position of WorkspaceSwitcher (Fix 3), workspace dropdown panel (Fix 5),
and AppBreadcrumb wiring (Fix 2) remain unchanged in P4 — both lands in
P5. The legacy useBreadcrumb composable also remains untouched until P5
(atomic with AppTopbar refactor).
Orphans flagged for follow-up cleanup (intentionally not deleted in P4):
useV2Nav composable + spec, V2NavGroup/V2NavItem types, sidebarNavActive
helper + spec, navFixture in stories/v2/_helpers.ts.
Suite delta: 575 → 575 (+2 WorkspaceSwitcher no-sub spec, -1 AppSidebar
groups-prop assertion, -1 OrganizerLayoutV2 groups-forward assertion).
vue-tsc clean. Scoped ESLint clean (0 errors).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Per RFC-WS-PRIMEVUE-PLAN-2-5 §4 AD-2.5-D1 and §5.6 Fix 6. Single class
on <html> drives both PrimeVue darkModeSelector and Tailwind v4
@custom-variant dark — one toggle, two ecosystems react.
Audit findings (pre-change):
- applyDomAttributes was writing BOTH data-theme="dark" AND .dark on
documentElement. The historic data-theme write is the design-doc §4
mechanism that AD-2.5-D1 supersedes; the .dark toggle was already
correct (and is already paired with PrimeVue darkModeSelector: '.dark'
in plugins/primevue/index.ts:31, verified in P1).
- tailwind.css had NO @custom-variant dark directive — Tailwind v4
default is `prefers-color-scheme` (OS-controlled), so utility
`dark:` variants would have ignored the topbar toggle entirely.
- One stray .dark subtree wrapper in AppTopbar.stories.ts:56
(DarkTheme story) — deliberate Storybook isolation per its comment,
but in violation of AD-2.5-D1's single-source-of-truth rule.
Changes:
- useShellUiStore.applyDomAttributes(): removed data-theme write,
kept .dark class toggle on document.documentElement, kept
data-density (P6 wires density-toggle UI; density is an
orthogonal axis and unaffected). File-header comment updated to
cite AD-2.5-D1 + reference the Tailwind & PrimeVue mirror sites.
- assets/styles/tailwind.css: added
`@custom-variant dark (&:where(.dark, .dark *))` so utility
`dark:` classes resolve via the same .dark trigger.
- components-v2/layout/AppTopbar.stories.ts: stripped class="dark"
from the DarkTheme story's render wrapper. Story comment updated
to flag that visual confirmation now comes via parity-batch
Playwright (after Plan 2.5 closes), not Storybook autodocs. A
proper documentElement-mutating decorator is a backlog item.
- stores/__tests__/useShellUiStore.spec.ts: updated the existing
applyDomAttributes assertion to drop the data-theme expectation
(the write is gone); added a new `describe('applyDomAttributes
— dark mode (AD-2.5-D1)', …)` block with 2 specs (class toggle
reactive, no data-theme attribute written).
Re-grep verification — all three return 0 hits:
- stray .dark in v2 (excluding `dark:` utility prefixes)
- data-theme setAttribute calls in stores/
- [data-theme=…] CSS selectors anywhere
Suite delta: 573 → 575 (+2). vue-tsc clean. Scoped ESLint clean.
Note: darkModeSelector: '.dark' was already set in
plugins/primevue/index.ts:31 (verified in P1 audit) — the config
dimension of AD-2.5-D1 was satisfied before this commit; P3 closes
the store-side, Tailwind-side, and stray-class dimensions.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per RFC-WS-PRIMEVUE-PLAN-2-5 §4 AD-2.5-T1. Establishes Inter as the
canonical Crewli body font via @fontsource/inter (local package, no
Google Fonts CDN).
Audit findings (pre-change):
- No @fontsource/public-sans package was installed.
- No <link> tag in index.html loaded Public Sans.
- Only one Public Sans reference existed in source: the vendored
Vuexy SCSS variable $font-family-custom at
src/@core/scss/template/libs/vuetify/_variables.scss, which drives
Vuetify's $body-font-family on legacy surfaces during F4.
- No src/main.css exists; the Tailwind v4 entry lives at
src/assets/styles/tailwind.css with no @theme block yet.
Changes:
- @fontsource/inter@^5.2.8 added to dependencies; weights
400/500/600/700 imported at main.ts ahead of tailwind.css.
- src/assets/styles/tailwind.css: new @theme block declaring
--font-sans Inter-first, plus :root --crewli-font-family and
html/body font-family applying that variable cascade-wide.
- src/@core/scss/template/libs/vuetify/_variables.scss:
$font-family-custom switched from the historical body font to
Inter (vendored edit, narrowly scoped, F6 removes @core/ entirely).
- tests/unit/styles/typography.spec.ts: 3-spec regression lock
(Tailwind direct stacks, Vuexy SCSS variable, zero historical
references in either file). File-content inspection — jsdom does
not cascade from imported stylesheets, so getComputedStyle would
always pass.
Suite delta: 570 → 573 (+3; the prompt's template was +2 but the
audit revealed two distinct font-config files, so each gets its own
assertion per the prompt's "cover all sites" rule). vue-tsc clean.
Scoped ESLint clean.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per RFC-WS-PRIMEVUE-PLAN-2-5 §8 step 1. Foundation scaffolding only —
no shell fixes, no Public Sans removal, no useShellUiStore changes
(P2–P6 scope).
Implements:
- theme darkModeSelector verified at '.dark' (already correct in
plugins/primevue/index.ts — config site is here, not theme.ts).
- src/config/navigation.ts: APP_NAVIGATION registry per AD-2.5-B1
(Dashboard entry only — v2-dashboard is the only v2 route today).
- src/composables/useBreadcrumb.ts: walkNavTree pure helper +
useNavBreadcrumb composable per AD-2.5-B1. The legacy meta-based
useBreadcrumb is preserved (consumed by AppTopbar, P1 may not
touch AppTopbar); P4 retires it and renames useNavBreadcrumb.
- src/components-v2/layout/AppBreadcrumb.vue: layout primitive
wrapping PrimeVue Breadcrumb, consuming useNavBreadcrumb.
- Tests: walkNavTree (4 specs, co-located), AppBreadcrumb mount
(2 specs, tests/component/layouts/).
Suite 564 → 570 (+6, all new specs green). vue-tsc clean. Scoped
ESLint clean.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds dev-docs/superpowers/plans/2026-05-17-gui-redesign-tier1-primitives.md
to .claude-sync.conf so the Plan 3 doc is visible to Claude Chat for
drift detection and review during Plan 3 execution.
Glob support remains the right BACKLOG item (TECH-SYNC-001 candidate)
to eliminate manual conf maintenance for future plans.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Eight components under apps/app/src/components-v2/shared/ plus a
severity-map utility, seeded from amended spec §8 and enforced via the
bidirectional Vitest consistency test required by §8.X.
Carries forward two Plan 2-deviation cleanup tasks from spec-amend
commit ae0bd2da:
- (a) migrate 6 centralized stories from src/stories/v2/ to co-located
- (b) refactor AppTopbar to wrap PrimeVue Menubar per RFC AD-3
Also deletes the X.vue boundary-test stub and repoints
boundaries-v2.spec.ts at a real shared component.
Plan format follows Plan 1 precedent (REQUIRED SUB-SKILL header,
- [ ] task syntax, Definition of Done, Plans 4-5 outline). Execution
will happen in a fresh Claude Code session via
superpowers:subagent-driven-development.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- §8 severity table now covers all 21 values of the 5 mirrored enums
(ShiftAssignment/ArtistEngagement/Payment/Person/MatchStatus); drops
3 phantom rows (active/inactive/expired, in no enum). Closes the
silent grey-fallback gap for 11 production values found in the
Plan 3 brainstorm self-audit.
- §6 stories-placement reworded: custom/wrapper components (incl.
PrimeVue wrappers) co-locate; only the ~80 PrimeVue catalog +
Foundations centralize. Plan 2 misread this.
- New §8.X: bidirectional Vitest consistency test
(apps/app/tests/unit/utils/statusSeverity.consistency.spec.ts),
added in Plan 3 — fails on any unmapped enum value OR orphan key.
Plan 3 cleanup tasks (tracked in the Plan 3 plan doc, not here):
(a) migrate Plan 2's 6 centralized stories to co-located.
(b) refactor AppTopbar to wrap PrimeVue Menubar per RFC AD-3.
Background: discovered during Plan 3 (Tier-1 primitives) brainstorm
self-audit. Fixes spec-vs-reality drift and two Plan 2 deviations
from binding spec/RFC text; prevents recurrence for future enums.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds to .claude-sync/ scope (.claude-sync.conf):
- dev-docs/RFC-WS-GUI-REDESIGN-CREWLI-STARTER.md
- dev-docs/superpowers/specs/2026-05-15-crewli-starter-gui-redesign-design.md
- dev-docs/superpowers/plans/2026-05-16-gui-redesign-foundation.md
These docs are referenced by RFC-WS-FRONTEND-PRIMEVUE.md and
PRIMEVUE_COMPONENTS.md but were absent from sync coverage, preventing
Claude Chat from drift-checking or reviewing Plan 3 work against
actual spec content.
The sync script has no glob support, so each new
dev-docs/superpowers/plans/*.md must be added to .claude-sync.conf by
hand (noted inline in the conf).
Vitest hoists vi.mock()/vi.hoisted() above all imports, so the
component import can sit with the other imports (import/first satisfied)
without the eslint-disable-next-line directives — the mock factories
only deref their refs at mount time. Honors the no-eslint-disable rule.
28/28 affected specs green.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The project's no-unused-vars only ignores all-underscore names (/^_+$/u);
`_bp` in the @vueuse/core useBreakpoints mock failed it. Latent since
Task 3 — masked because the whole-codebase `pnpm lint` stylish formatter
OOMs (RangeError on the legacy-code message volume) and emitted no
results. Scoped errors-only lint surfaced it. 21/21 specs still green.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Code-review follow-up. AppTopbar DarkTheme mutated <html>.dark which
leaked into Default/CompactDensity stacked on the same autodocs page;
scope dark to the story subtree via a `.dark` wrapper (Aura
darkModeSelector is the `.dark` class) — verified isolated on the docs
page. Also factor the duplicated render scaffolds in AppDialog (shared
dialogStory factory) and WorkspaceSwitcher (meta-level render).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Replaces the Plan-1 skeleton stubs: OrganizerLayoutV2 now fills
AppShellV2's #sidebar/#topbar/#drawer slots with the ported AppSidebar /
AppTopbar / RightDrawer and sources orgNavItems via useV2Nav() (legal
now that OrganizerLayoutV2.vue is the layouts-v2 zone). AppShellV2 is
unchanged; its contract test stays green. New component test locks the
composition (right component per slot, :groups forwarded, no skeleton).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
AD-G2 ("OrganizerLayoutV2 wraps AppShellV2") was in tension with AD-G5:
src/layouts/OrganizerLayoutV2.vue classified as the v1 `layouts` zone,
which is deliberately barred from components-v2. New `layouts-v2` zone
(src/layouts/*V2*.vue, mode:file) gets pages-v2-equivalent v2 capability;
the v1 `layouts` zone is unchanged so v2 isolation is preserved. RFC
AD-G5 amended; locked by 3 boundaries-v2.spec.ts regression tests (7/7).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
vue/no-reserved-component-names is error-level; <dialog> is native HTML.
Matching is via the stubs key + findComponent reference, not the stub's
own name, so the rename is behaviour-neutral (14/14 tests still pass).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Replaces crewli-starter's hand-rolled Teleport/scrim/keydown Escape
pattern with a typed PrimeVue Dialog wrapper. v-model:open via writable
computed; slots #tabs/#footer gated; 12 unit tests all passing.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds :key="component" to <component :is="Body"> so opening drawer B
while A is open fully remounts (A's instance/state can't leak into B) —
a real defect for a primary shell overlay. Same-name reopen still
relies on the body reacting to prop changes (documented inline).
Companion test asserts the cross-component switch swaps the body
cleanly (A unmounts, B mounts). Addresses the Task 5 code-review
Important finding.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
unplugin-auto-import scans src/composables/; the new drawerRegistry
exports added global + vue-module declarations. auto-imports.d.ts is
tracked — keep it in sync (same precedent as prior composable syncs).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds composables/drawerRegistry.ts (boundary-safe register-by-call map:
register/resolve, zero static component imports — composables zone may
not import components, RFC-WS-GUI-REDESIGN AD-G5). Extends useRightDrawer
with resolveDrawerComponent (thin facade, prior API/tests preserved).
RightDrawer.vue: PrimeVue <Drawer position=right>, v-model:visible via a
writable computed ↔ useRightDrawer isOpen/close; title/flush read from the
open() props object (A4); dynamic <component :is> via resolveDrawerComponent
with a graceful empty state on null; #actions header slot retained. 18
unit/component tests.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- FIX A (IMPORTANT): PrimeVue Breadcrumb ignores `route` key; map non-last
items with `command: () => router.push(item.to)` for real client-side nav
- FIX B: add type="button" to all 6 native <button> chrome elements
- FIX C: authStore.logout() bare call matches project no-void pattern
- FIX D: document param-route edge case in toBreadcrumbItems
- FIX E: regression test asserts command+push on non-last, no command on last,
no `route` key on any item
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- useBreadcrumb composable: pure toBreadcrumbItems() helper + thin
useRoute() wrapper; route-driven, no prop coupling
- AppTopbar: hamburger→setMobileOpen, theme/density toggles→shell store,
PrimeVue Breadcrumb/OverlayBadge/Popover/Avatar/Menu; replaces all
manual document.mousedown listeners with PrimeVue built-in dismissal;
notifications stubbed (useNotificationStore is a toast queue, not a
feed — TODO TECH-WS-GUI-REDESIGN); sign-out→authStore.logout()
- Unit tests: 10 breadcrumb + 6 AppTopbar assertions (16 total, all pass)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
CRITICAL: replace `lg:hidden` on PrimeVue Drawer with `v-if="isMobile"` so the
teleported portal/overlay is never created on desktop viewports regardless of
mobileOpen state. Replace useMediaQuery raw string in SidebarHeader with
useBreakpoints(breakpointsTailwind).smaller('lg') shared by both components.
Add desktop/mobile comments; adapt tests to useBreakpoints mock; add
Drawer-absent-on-desktop and aside w-16/w-64 width-class assertions (21 tests).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Ports crewli-starter's monolithic AppSidebar.vue into two typed production
components: SidebarHeader (the .brand block) and AppSidebar (composing
SidebarHeader + SidebarNav + WorkspaceSwitcher). AppSidebar renders a
permanent <aside> on desktop (lg+) and a PrimeVue Drawer on mobile, both
wired to useShellUiStore for collapse/mobile state.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Ports crewli-starter WorkspaceSwitcher into the Crewli SPA as production
TypeScript: PrimeVue Popover replaces the manual click-outside listener,
data is derived from useAuthStore/useOrganisationStore (no new store), gradient
pairs are deterministic via a new pure util with full Vitest coverage.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- FIX 1: Replace <button @click="router.push"> with <RouterLink custom>
+ <a> for real link semantics (middle-click, ⌘-click, screen-reader);
custom isNavItemActive prefix-match stays the active source of truth;
adds :aria-current="page" on active items; drops useRouter/router.push.
RouterLink to prop cast via itemTo() helper (RouteLocationRaw from
unplugin-vue-router) to satisfy typed RouterLinkTyped<RouteNamedMap>.
- FIX 2: Align .nav-item comment to actual template values (py-[9px]
rounded-md, not CSS vars); replace inaccurate Tailwind v3/v4 before:
composability justification in <style scoped> with the real reason
(accent bar at left:-10px is clipped by the overflow-y-auto nav).
- FIX 3: text-left → text-start (logical property, RTL-safe).
- FIX 4: Document id=route-name assumption in useV2Nav.ts with a
one-line comment at the id: assignment.
- FIX 5: Reword misleading "dotted names" spec description to state
the real invariant (id = v1 route name, already kebab-case).
- FIX 6: Add 2 tests — useV2Nav wrapper .value equality, and
consecutive-headings edge case (empty-items group produced).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
unplugin-auto-import scans src/composables/, so the new useV2Nav added
a global + vue-module declaration. auto-imports.d.ts is tracked; keep
it in sync (same precedent as Plan 1's useRightDrawer sync).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Ports crewli-starter's sidebar nav into the SPA as production TS:
V2NavGroup/V2NavItem types, a pure toV2NavGroups adapter wrapped by
useV2Nav(items) (composables zone can't import @/navigation, so the
v1 nav array is passed in — the layout supplies orgNavItems in Task 7),
a pure isNavItemActive helper, and SidebarNav.vue (props-only,
router-driven nav, route-based active state, collapsed mode, main.css
translated to Tailwind inline). 16 unit tests. Icon import is
allowed via the components-foundation bridge (no eslint-disable).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Plan-1 Task-4 added { type:'components-foundation', pattern:
'src/components/Icon.vue' } without mode:'file'. eslint-plugin-boundaries
defaults to folder mode, so the single-file pattern never matched and
Icon.vue fell through to the generic `components` catch-all — breaking
the sanctioned components-v2 -> Icon bridge (RFC AD-G5) for every v2
shell component. Plan-1's boundary test only exercised the forms/**
folder-glob edge so the gap was latent. Adds mode:'file' + a regression
test locking the components-v2 -> Icon.vue edge.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Storybook failed to boot ("module does not provide an export named
'default'"): .storybook/preview.ts did a default import while the
PrimeVue installer is a named-only export (deliberate — f218ac6e
switched to named to stop Vuexy registerPlugins() double-registration).
That commit missed updating preview.ts, so Storybook has been broken on
main since. Align the consumer with the producer (named import, same as
main.ts) rather than re-adding a default export. Pre-existing bug,
unrelated to the GUI-redesign merge.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
unplugin-auto-import scans src/composables/, so the new useRightDrawer
(Task 7) added a global + vue-module declaration. auto-imports.d.ts is
tracked (committed for editor/CI type resolution without a build), so
keep it in sync — same precedent as committing typed-router.d.ts. No
new symbol from useShellUiStore: src/stores is not in the auto-import
dirs (stores are imported explicitly), which is correct.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Replaces the Task 3 stub in pages-v2/dashboard.vue with the boot-proof
page (data-testid v2-dashboard, correct definePage meta). Adds the
Playwright CT smoke (appshell-boot.spec.ts) that mounts AppShellV2 in
Chromium and asserts both the shell root and slot content are visible;
uses page scope for the root-element assertion (CT component locator
only matches descendants, not the root itself). Full Plan-1 gate green:
typecheck 0 new errors, eslint clean, 5 vitest files / 21 tests + 2
component tests, vite build succeeded, typed-router has v2-dashboard.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The vite-plugin-vue-meta-layouts layout name is a hand-maintained union
in env.d.ts (not auto-generated). Task 8 created OrganizerLayoutV2 and
Task 3's boot-proof page sets meta.layout: 'OrganizerLayoutV2', but the
union lacked the v2 entries so a typed definePage failed vue-tsc.
Registers both v2 layouts (PortalLayoutV2 added now to pre-empt the same
gap when portal v2 lands). Completes Task 8 (RFC-WS-GUI-REDESIGN AD-G2).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Tailwind-grid shell skeleton with named slot regions (sidebar, topbar,
default, drawer). OrganizerLayoutV2 wires the skeleton with RouterView,
selectable via definePage meta. Vitest component mount test: 2 tests pass.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a valid case (PortalLayoutV2 under pages-v2/portal) and an invalid
case (OrganizerLayoutV2 under pages-v2/portal -> wrongLayout) so the
rule's portal-path branch has positive + negative coverage. Closes the
Task 5 code-review Important finding.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a custom ESLint rule (local-rules/require-v2-layout-meta) that
fails any src/pages-v2/**.vue page missing
definePage({ meta: { layout: 'OrganizerLayoutV2' } }) (or PortalLayoutV2
under pages-v2/portal), preventing a silent wrong-shell fallback to the
default layout (RFC-WS-GUI-REDESIGN AD-G2). Wires eslint-plugin-local-rules
+ a pages-v2 override. The RuleTester spec is called at top level (ESLint
RuleTester self-manages describe/it under Vitest) and vitest.config.ts
gains the eslint-rules test glob so the spec is discovered.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Documents that the test filter and the .eslintrc rule key both use the
v5-era 'boundaries/element-types' alias; a future eslint-plugin-boundaries
bump that drops the alias must update both together or the filter silently
matches nothing. Addresses the Task 4 code-review Minor.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds three new eslint-plugin-boundaries element zones and their matrix
rows so the GUI-redesign v2 surface is structurally isolated: v1 code
cannot import from v2 (back-porting forbidden), v2 can reach the
narrow FormField/Icon bridge via the components-foundation zone, and
pages-v2 can import from components-v2. Backed by a Vitest spec
running via the ESLint Node API (node environment; happy-dom's
document object breaks the case-police resolver). Adds a placeholder
src/components-v2/shared/X.vue so the resolver can classify the
import target during the test (unresolvable imports are not boundary-
checked by the plugin).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Moves the `v2-` de-dup (needed because getPascalCaseRouteName folds the
v2/ URL segment into the base) into the unit-tested v2RouteName helper
and simplifies the vite.config.ts call site to v2RouteName(raw, nodePath).
Removes the duplicated isV2 detection. No behavioural change: /v2/dashboard
still resolves to route name v2-dashboard; v1 names unchanged. Addresses
the Task 3 code-review Important finding.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a second routesFolder (src/pages-v2 -> /v2/) and extends
getRouteName so v2 routes get a v2- NAME prefix, preventing collisions
with same-named v1 pages. getPascalCaseRouteName already folds the v2/
URL segment into the base name, so the leading v2- is stripped before
v2RouteName re-adds the canonical prefix (avoids v2-v2-dashboard).
Includes the regenerated typed-router.d.ts and a boot-proof
pages-v2/dashboard.vue placeholder.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Add blank line between the new pointer blockquote and the **Aligned to:**
paragraph so the blockquote closes cleanly across markdown renderers.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
All corrections audited against the codebase:
- §7.4 useWorkspaceStore ghost removed (computed over auth/org stores)
- §12 portal /portal/* verified in repo; observability is meta-based,
/api/v1/p/* is separate backend layer — no cross-doc conflict
- §3 getRouteName v2- name-prefix convention (route-name collision)
- §4 theme parallel-mode AD + useRightDrawer in useShellUiStore
- §8/§9 DraggableBlock is foundation, not Tier-4
- §3 single ESLint enforcement for definePage meta-key
- §8 StatusTag severity map; §14 brace-glob fallback; §13 CT/Storybook
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Brainstorming outcome: pivot the PrimeVue redesign to use crewli-starter
as the design source of truth, parallel /v2/ routes, PrimeVue-first
fidelity, page-by-page cutover. Supersedes F4a-F4d of
RFC-WS-FRONTEND-PRIMEVUE.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Tailwind's lg:hidden loses to PrimeVue's .p-button { display: inline-flex }
due to equal specificity but later cascade order. Resulted in the mobile
hamburger remaining visible on desktop, allowing the Drawer to open over
the already-visible permanent sidebar.
Fix: wrap mobile-only cluster (hamburger + title) in a plain <div lg:hidden>
so the wrapper owns the visibility toggle. The wrapper is not a PrimeVue
component, so no specificity competition.
The Drawer itself had the same anti-pattern (class="lg:hidden") and is
worse, because PrimeVue Drawer teleports to body — a wrapping div on the
parent does not isolate the teleported overlay, and a class on the Drawer
root loses to .p-drawer { display: flex } when visible. Converted to
v-if="!isLg" driven by useMediaQuery('(min-width: 1024px)'). Vue simply
does not render the component on lg+, so no display rule competes.
Audited all 5 layouts for the same anti-pattern:
- AppShell.vue — fixed (Button + Drawer described above)
- default.vue / OrganizerLayout.vue / PortalLayout.vue — delegate to
AppShell; no PrimeVue elements with responsive classes
- blank.vue — plain <div>, no PrimeVue
- PublicLayout.vue — plain <main>, no PrimeVue
useMediaQuery is auto-imported via unplugin-auto-import's @vueuse/core
entry in vite.config.ts; explicit imports get stripped by the post-edit
ESLint --fix hook as redundant.
F3-introduced bug (commit 43915501); surfaced during F3.5 testing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
main.ts explicitly calls installPrimeVue(app) AFTER registerPlugins(app)
per the comment in main.ts ("so PrimeVue lives outside the Vuexy @core
machine"). The intent was a single registration site outside the
auto-discovery loop.
Bug: registerPlugins (src/@core/utils/plugins.ts) globs
plugins/*/index.{ts,js} eagerly and invokes the `default` export of
each match. plugins/primevue/index.ts was exporting installPrimeVue
as the default, so registerPlugins also picked it up and called it.
End result: PrimeVue and its three services (Toast, Confirmation,
Dialog) were each registered twice on every app boot. Visible
symptoms: duplicate Toast emissions on a single Toast.add() call,
and ConfirmationService callbacks firing twice for one user
confirmation.
Fix: convert `export default function installPrimeVue` to a NAMED
export, and update main.ts's import to `{ installPrimeVue }`. The
registerPlugins glob still picks up the module path but the
`pluginImportModule.default?.(app)` invocation becomes a no-op via
optional chaining (no default export to call). main.ts remains the
single registration site.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The PrimeVue side of the parallel-mode app renders icons via
@iconify/vue's <Icon> component (src/components/Icon.vue). At runtime
@iconify/vue resolves an icon name like "tabler-eye" by looking up
its data in the in-memory icon registry; on a miss it falls back to
fetching https://api.iconify.design/tabler/eye.json. The CSP blocks
that origin, so every Tabler icon used in AppShell, SidebarHeader,
SidebarUserCard, and the migrated login form rendered as an empty
<svg viewBox="0 0 16 16"></svg>.
New plugins/iconify.ts loads the full Tabler set
(@iconify-json/tabler/icons.json, already in package.json as 1.2.23)
and registers it via addCollection() at module-load time. main.ts
side-effect-imports it before any other import so the registry is
warm before the first Icon mounts.
This is a NEW concern, separate from the existing plugins/iconify/
(index.ts + icons.css) which generates Vuexy-style i-tabler-* CSS
classes for Vuetify's VIcon adapter. The two systems must coexist
during F3–F6 parallel mode; the legacy directory can be deleted
alongside Vuetify when F6 lands.
Bundle cost: ~1.9 MB uncompressed JSON, ~400 KB gzipped in the main
chunk. Per-icon imports are a future optimisation.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
unplugin-vue-components' Components({ dirs }) in vite.config.ts only
scans src/components, src/@core/components, and src/views/demos. The
sub-components introduced in B1/B3 live under src/layouts/components/,
which is NOT in the auto-import scan path. Without an explicit script
import, Vue renders <SidebarHeader> and <SidebarUserCard> as unknown
HTML elements (no DOM output, no errors), which is why the topbar and
sidebar-bottom cards looked empty in browser inspection.
Adding the two imports inline with the existing Icon import keeps the
component graph explicit. The post-edit eslint --fix hook preserves
the imports because the template usages (already present from B1 and
B3) make vue-eslint-parser see them as used.
The original B1/B3 commits had the imports stripped by the hook
because the imports were added in a separate Edit *before* the
template usages — eslint --fix correctly removed them as unused at
that moment, and the next Edit added the template usage but not the
import.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Left side gains a desktop-only breadcrumb "Organisation / Page title"
using the current organisation from useAuthStore and a page title
resolved by:
1. route.meta.title (if a page sets it explicitly), then
2. matching the active route name against the navItems prop, then
3. humanizing the route name as a last-resort fallback.
The chevron separator is suppressed when either side is empty, so
portal and pre-org users see just the page title. Mobile preserves
the existing hamburger + title text (the breadcrumb is hidden on
<lg to keep the topbar single-row).
Right side gains a notification bell and a help icon. The bell is a
visual placeholder (no badge) — clicking shows a PrimeVue Toast
"Notificaties komen binnenkort beschikbaar" until the notification
framework lands as a separate sprint.
The help icon would normally open https://docs.crewli.app in a new
tab, but the host currently serves with a TLS cert that does not
cover the name (ERR_TLS_CERT_ALTNAME_INVALID), so the click handler
falls back to a Toast. A TODO comment in the source records the
target URL and the one-line switch to make once the cert is fixed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Consolidates the user menu into a single sidebar-bottom location.
SidebarUserCard.vue shows avatar (initial), full name, role (Dutch
label mapped from org pivot role or 'Super Admin' fallback) and a
chevron-up that opens a PrimeVue Menu with "Mijn Profiel" and
"Uitloggen". The Menu uses popup mode; PrimeVue v4's absolutePosition
logic auto-flips above the trigger when the panel would overflow the
viewport bottom — verify in Phase C.
AppShell loses the topbar avatar Button + Menu and the associated
state (userMenuRef, userInitial, userMenuItems, toggleUserMenu) plus
its imports (Avatar, Menu, useAuthStore, computed). The component is
now a pure layout shell with no auth-store coupling. The topbar's
right side is intentionally empty in this commit; B4 fills it with
breadcrumb / notification bell / help icon.
Layout: nav uses min-h-0 flex-1 overflow-y-auto so it shrinks under
viewport pressure and lets the user card stay pinned at the bottom
of the sidebar. Mobile Drawer's content pt-override sets the same
flex-column behaviour so the user card sits flush at the bottom of
the drawer overlay.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Section headings ("Beheer" / organisation name, "Platform") were
already uppercase + muted but read as bold paragraph dividers more
than as quiet group markers. Tighten letter-spacing, drop weight
from semibold to medium, lighten the color one step (surface-500 →
surface-400), and shrink text to 11px so the headings recede and
let the nav items themselves carry the visual weight.
Spacing nudged from mt-4/mb-2/px-2 → mt-6/mb-1/px-3: more breathing
room above each group, less below (the items already have py-2 on
top), and the heading left-edge now lines up with the icons of the
nav items beneath it (both at px-3).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Introduces SidebarHeader.vue — a PrimeVue-only org-switcher that
replaces the centered Crewli wordmark at the top of the sidebar. The
component mirrors the legacy Vuetify OrganisationSwitcher (avatar with
org initials, organisation name, plan-tier placeholder, dropdown
chevron, PrimeVue Menu of available orgs) but cannot reuse it
directly per the R-10 layout-shell-isolation invariant.
Plan-tier shows a hardcoded "Pro" placeholder until the backend
Organisation resource exposes a plan field — tracked separately, not
in F3.5 scope. When the user has no active organisation (portal
users, fresh super_admin), the component degrades to the original
title block so PortalLayout continues to read "Crewli Portal".
Desktop sidebar width bumped w-64 → w-72 (256 → 288 px) to give the
org-switcher card breathing room and accommodate the user-info card
arriving in B3. Mobile Drawer width bumped 16rem → 18rem to match.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Installs Storybook 10.4 in apps/app/ as a component-development and
autodoc tool. Configures viteFinal with all seven SPA aliases so
stories resolve imports identically to the dev/build pipeline.
preview.ts reuses @/plugins/primevue's installPrimeVue() so Storybook
stays in lock-step with main.ts whenever the PrimeVue config changes.
Only the addons we need are wired: addon-docs (autodocs) and
addon-a11y (axe-core checks). addon-interactions is intentionally
omitted — interaction testing stays in Playwright CT per the testing
architecture.
Seed stories: PrimeVue Button (Primary/Secondary/Danger), Tailwind
utility box, and FormField (Default/WithError/Disabled) wrapped in
@primevue/forms Form + Zod resolver.
Adds make storybook target alongside make app / make docs.
Three RFC drift corrections discovered during F3 implementation:
1. AD-5 icon rendering: corrected from "<i :class='i-tabler-X'>"
utility-class pattern (which would require UnoCSS, not installed)
to "@iconify/vue's <Icon> component with name='tabler-X' prop"
(existing Crewli pattern producing real SVG output). The thin wrapper
shipped in F3 B6 as apps/app/src/components/Icon.vue accordingly.
2. AD-2 theme architecture: corrected package reference from
@primevue/themes@^4.5 (deprecated by PrimeFaces) to
@primeuix/themes@^2 (the path now prescribed by PrimeVue 4's
official install docs at primevue.org/vite/). Same maintainers,
same API surface (definePreset, Aura preset, semantic tokens).
F3 commit B1 already uses the corrected package.
3. Appendix B Aura theme token plan: updated import-path examples to
@primeuix/themes and @primeuix/themes/aura accordingly.
Also updated:
- §6 F3 deliverables list: dependency line now reads @primeuix/themes@^2
with a footnote linking to the B1 rationale.
- Appendix C Version Pinning Policy: separated @primeuix/themes from
the primevue/^primevue/forms lockstep pin (independent release cadence).
- dev-docs/PRIMEVUE_COMPONENTS.md §3 (Data display): VIcon row updated
to <Icon name="tabler-..." />; surrounding migration-spirit paragraph
rewritten; §10 external-resources link relabeled to @primeuix/themes.
These are RFC drift corrections — the implementation in F3 (commits
B1, B2, B6 of this sprint) already uses the corrected packages and
import paths. This commit aligns the spec with reality so future
contributors don't reach for the deprecated/inaccurate documentation.
.claude-sync/ regenerates automatically post-commit via lefthook.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the Vuetify VForm + AppTextField + VBtn stack with the F3
form pattern: @primevue/forms' <Form> with a Zod resolver, the
project-owned <FormField> wrapper from B5, and PrimeVue InputText /
Password / Checkbox / Button at the input layer. Surrounding chrome
(VRow / VCol illustration column, VCard, VAlert reset-success banner,
auth-logo link, MfaChallengeCard) stays Vuetify until F4b migrates
the auth surface in full.
Zod schema:
- email: required, valid email format
- password: required
Both messages are Dutch (per F3 sprint plan convention).
422 error handling routes through useFormError() from B5. The Laravel
response shape (errors.<field>: string[]) feeds applyApiErrors directly.
rate_limited and other reason-only failures are synthesized into the
email field's error map so they surface visually under the email input,
preserving the existing UX.
The remember-me checkbox is rendered with PrimeVue Checkbox (no schema
coverage — it's UI state, not validated input). The password visibility
toggle is delegated to PrimeVue's Password component's built-in
toggle-mask prop (replaces the previous manual isPasswordVisible ref
and append-inner-icon plumbing).
Verification:
- pnpm typecheck — clean.
- pnpm test — 402 tests pass unchanged.
- pnpm build — succeeds; login chunk grew from ~21 KB to ~84 KB raw
due to @primevue/forms + Password/Checkbox component code (gzip 22 KB).
Will normalize during F4 as more pages share these modules.
- Manual browser test deferred to Phase C brand-review screenshot
capture.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Layout-shell rewrite per RFC AD-3, B7-option-B. R-10 isolation invariant
honored — this single commit is revertible to roll back the layout
change without losing B1–B6 progress.
New component (PrimeVue-only, no Vuetify imports per F3 hard constraint):
- apps/app/src/layouts/components/AppShell.vue (~210 lines)
- Desktop sidebar (Tailwind grid, lg+ breakpoint) renders nav items
as PrimeVue Buttons + Icons. Mobile (<lg) hides sidebar; PrimeVue
Drawer slides in on hamburger toggle.
- Top bar (Tailwind) has hamburger + title (mobile) and an Avatar +
Menu (PrimeVue) for the user dropdown with "Mijn Profiel" and
"Uitloggen" actions.
- Nav items accept the existing { title, to: { name }, icon: { icon } }
shape from src/navigation/vertical so call-sites stay terse.
Five top-level layouts delegate to AppShell (filename preserved per
AD-3 so vite-plugin-vue-meta-layouts continues to resolve routes
unchanged):
- default.vue — org + (super-admin) platform nav
- OrganizerLayout — same nav as default; matches authenticated org UX
- PortalLayout — portal-specific 2-item nav ("Mijn evenementen",
"Mijn Profiel")
- blank.vue — minimal chrome-less wrapper for login etc.
- PublicLayout — minimal wrapper for public form-fill routes;
uses <main> for semantic structure
F3 functional regressions (intentional — F4 sub-packages reintroduce
each item through PrimeVue):
- NavSearchBar (Vuetify-heavy combobox/overlay) — absent from top bar
- ContextSwitcher (Vuetify VBtn + VMenu) — absent
- NavbarThemeSwitcher (Vuetify IconBtn) — absent; dark mode driven by
PrimeVue's darkModeSelector: '.dark' continues to work via the
existing @core skin classes until F6 cleanup
- NavbarShortcuts (Vuetify-heavy) — absent
- NavBarNotifications (Vuetify-heavy) — absent
- UserProfile from @/layouts/components/ (Vuetify-heavy menu) — replaced
with the minimal Avatar + Menu dropdown described above; rich profile
panel returns in F4
- ImpersonationBanner — absent; super-admin impersonation UX is F4 work
- PortalLayout event-mode vs platform-mode topbar (route.meta.navMode
driven) — absent; F4 reintroduces via AppShell prop or slot
- Suspense + AppLoadingIndicator wrapping pages — dropped; pages handle
their own loading via PrimeVue ProgressSpinner
VApp at App.vue level still wraps everything, so Vuetify components
inside still-Vuetify pages continue to render correctly during the
parallel-mode window.
Test updates (no Vuetify in layout structure to assert against anymore):
- OrganizerLayout.spec.ts — mocks AppShell instead of the deleted
DefaultLayoutWithVerticalNav reference; provides Pinia.
- PortalLayout.spec.ts — same mock pattern; new structural assertions
go through AppShell stub; the new third test verifies
PortalLayout forwards portal nav items + title to AppShell.
- PublicLayout.vue — uses <main> for semantics; PublicLayout.spec.ts
still passes unchanged.
Auto-generated component/auto-import dts files refreshed for the new
AppShell component (committed for stable dev workflow).
Verification:
- pnpm typecheck — clean.
- pnpm test — 402 tests pass (test count unchanged after spec rewrites).
- pnpm build — succeeds in 14.05s; AppShell chunk is ~57 KB raw.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two additions complete the F3 runtime scaffolding:
apps/app/src/components/Icon.vue — generic Iconify renderer wrapping
@iconify/vue's <Icon> component. F4 migration substitutes
<VIcon icon="tabler-X" /> with <Icon name="tabler-X" /> at call-sites,
producing real SVG output and using the existing Crewli "tabler-*"
naming convention. Props: name (required, e.g. "tabler-eye"), optional
size. The component avoids @iconify/vue's auto-import for clarity at
call-sites.
apps/app/src/App.vue — mounts <Toast /> and <ConfirmDialog /> at the
template root inside VLocaleProvider. Both render alongside the
existing VSnackbar and VDialog confirm patterns during the F3–F4
parallel-mode window. F4 sub-packages migrate call-sites to PrimeVue's
useToast() / useConfirm() composables.
UnoCSS-style i-tabler-* utility-class rendering (RFC AD-5 v1.0 wording)
is not adopted — UnoCSS is not installed in the Crewli stack. The
RFC will be aligned in B9.
Verification:
- pnpm typecheck — clean.
- pnpm test — 402 tests pass unchanged.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two new artifacts that together provide the F4 form-migration target:
apps/app/src/components/forms/FormField.vue — project-owned wrapper
around @primevue/forms' built-in FormField. The default slot accepts
the actual input (e.g. <InputText name="email" />); the wrapper renders
label (with optional required asterisk), error Message, and hint chrome
around it. Reads field state from the parent <Form> via the built-in
FormField's scoped slot, so call-sites do not need to thread $form
manually.
apps/app/src/composables/useFormError.ts — the API 422 bridge. Parent
component calls useFormError() once; the composable provides an
apiErrors ref through Vue inject. Each FormField in the component
reads its own field name from that map. applyApiErrors() reads the
Crewli backend's { errors: { field: string[] } } shape and surfaces
the first message per field; clearApiErrors() resets between submits.
Error precedence per RFC Appendix A: explicit apiError prop > inject
apiErrors map > Zod resolver error from $field.
Signature note: RFC's useFormError(formRef) is implemented as
useFormError() — the formRef parameter is unused in the provide/inject
implementation, and Crewli convention avoids unused parameters. RFC
will be aligned in B9 if it remains a meaningful spec gap during F4.
Verification:
- pnpm typecheck — clean.
- pnpm test — 402 tests pass unchanged.
- B8 will exercise the components end-to-end on the login page; F4d
validates against the public-registration multi-step form.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
installPrimeVue(app) runs AFTER registerPlugins(app) (which registers
Vuetify + router + Pinia via the Vuexy @core machine). Placing the
PrimeVue install outside @core/utils/plugins is deliberate — it keeps
PrimeVue free of the Vuexy plugin loader so F6 can remove @core/
without disturbing PrimeVue registration.
Both frameworks are now active at runtime. Existing Vuetify pages
continue to render unchanged; PrimeVue components become available
for the layout-shell rewrite (B7) and the FormField wrapper (B5).
Verification:
- pnpm typecheck — clean.
- pnpm build — succeeds in 14.26s, no PrimeVue or theme-related errors.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three additions wire Tailwind v4 into the SPA without disturbing the
existing Vuetify pipeline:
- apps/app/src/assets/styles/tailwind.css — Tailwind v4 CSS-first entry.
Uses @import "tailwindcss"; @plugin "tailwindcss-primeui"; and
@source pointing at apps/app/src/ to scan template content.
- apps/app/vite.config.ts — adds the @tailwindcss/vite plugin between
vue() and vuetify(). After vue() so it sees compiled template
content; before vuetify() so Vuetify's SCSS pipeline runs unimpeded.
- apps/app/src/main.ts — imports tailwind.css before the Vuetify/Vuexy
SCSS so utility classes are available alongside Vuetify's cascade.
optimizeDeps.exclude remains ['vuetify'] (no PrimeVue addition) — HMR
behaves correctly in dev with the current config; revisit if needed.
Verification:
- pnpm typecheck — clean.
- pnpm build — succeeds in 13.97s; CSS emitted per-route as expected.
- pnpm test — 402 tests pass unchanged.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Scaffolds apps/app/src/plugins/primevue/ as three files mirroring the
Vuetify plugin structure at apps/app/src/plugins/vuetify/:
- theme.ts — CrewliPreset extends Aura via definePreset(). Primary
palette (50–950) is the exact token plan from RFC Appendix B,
centered on Crewli teal #0D9394 (light primary, primary.500) and
#0B7F80 (dark primary, primary.600). Surface tokens use Aura
defaults. colorScheme.light/dark map primary.color, hover, active,
and contrastColor per Appendix B.
- defaults.ts — empty pt (PassThrough) defaults object. F3 ships this
scaffold; F4 sub-packages populate component-level defaults as each
Vuetify surface migrates.
- index.ts — installPrimeVue(app) registers PrimeVue with the preset,
Dutch locale (primelocale/nl.json → nl.nl), darkModeSelector: '.dark'
(matches Vuexy convention per AD-2), and the three services Toast,
Confirmation, Dialog.
Theme imports use @primeuix/themes (the maintained successor PrimeVue's
official docs prescribe), not RFC v1.0's @primevue/themes. See B1 commit
for substitution rationale. RFC will be aligned in B9.
The plugin is not yet registered in main.ts — that lands in B4 after
Tailwind v4 wiring (B3).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Packages installed:
- primevue@4.5.5
- @primeuix/themes@2.0.3 (substitutes @primevue/themes per ecosystem
state — see rationale below)
- @primevue/forms@4.5.5
- primelocale@1.6.0 (pinned to ^1 per RFC)
- tailwindcss@4.3.0
- @tailwindcss/vite@4.3.0
- tailwindcss-primeui@0.6.1
Package substitution: @primevue/themes → @primeuix/themes
RFC v1.0 §6 F3 specifies @primevue/themes@^4.5, but during install pnpm
reported this package as deprecated by its maintainers (PrimeFaces) with
explicit guidance to migrate to @primeuix/themes. Web verification confirms
that the official PrimeVue 4 install documentation at primevue.org/vite/
now specifies `@primeuix/themes` directly, not the deprecated path:
pnpm add primevue @primeuix/themes
import Aura from '@primeuix/themes/aura';
@primeuix/themes is maintained by the same maintainers (mert.sincan,
cagatay.civici), has the same API surface (Aura preset, definePreset,
semantic tokens), and is the path PrimeVue 4's documentation now
prescribes. The substitution is not a deviation from PrimeVue v4
conventions — it IS the current PrimeVue v4 convention.
The RFC will be amended in B9 to align AD-2 and Appendix B with this
ecosystem state.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Both comments in apps/app/playwright/index.ts (header block lines 38-45
and inline at line 66) state that the Vuetify provider gets replaced by
PrimeVue in F3. This predates the RFC clarification that test-runtime
flip is F5, not F3 (per ARCH-TESTING.md §7).
F3 builds the PrimeVue runtime in main.ts but keeps the test runtime
on Vuetify. Component tests continue to mount with the Vuetify provider
until F5 deliberately swaps it. This commit aligns the comments with
that decision so no future contributor wonders whether the F3 sprint
should have touched this file.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Lefthook v2 runs `git lfs pre-push` internally for pre-push hooks (per
docs/usage/features/git-lfs.md; confirmed in internal/run/controller/
lfs.go where the internal handler invokes `git lfs pre-push <remote>
<url>` with a buffered `cachedStdin`). Our manual `git-lfs:` command
in lefthook.yml was a second invocation against the same remote; the
duplicate is directly visible in `LEFTHOOK_VERBOSE=1` output as
`[git-lfs] executing hook` (internal) followed by `[lefthook] run:
git lfs pre-push` (manual).
The previous fix attempt (piped: true, commit 1b06804) was based on a
wrong understanding of `piped`'s semantics — `piped` controls
fail-fast behavior, not stdin routing or sequencing. Default lefthook
behavior is already sequential per docs/configuration/parallel.md.
That "fix" was placebo; incident 2 (F2 push, zero LFS objects, commit
99eedb6) proved it.
Phase A investigation: documentary + source confirmation that lefthook
owns the LFS pre-push call. Phase B sandbox test against a filesystem
remote confirmed the duplicate execution in logs but did NOT reproduce
the production hang — likely because the duplicate manual call against
a local remote has no LFS server to interact with. A network-y remote
(Gitea over SSH/HTTPS) appears to be part of the trigger. Two
mechanisms remain plausible (H1: PTY-stdin without EOF in
`while read` loop per docs/configuration/use_stdin.md; H4: server-side
LFS interaction on the duplicate call). Both are eliminated by the
same fix: remove the manual command. LFS uploads continue to work via
lefthook's internal handler (verified in sandbox post-fix).
Regression coverage: scripts/test-lefthook-pre-push.sh asserts exactly
one internal LFS invocation, zero manual ones, and `Uploading LFS
objects: 100%` present, against a disposable sandbox.
See dev-docs/ADR-LEFTHOOK-LFS-INTEGRATION.md for full context, both
misconceptions to prevent regression, and the alternative-scenarios
playbook if Phase E ever regresses.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Closes B5 of F2 (RFC-WS-FRONTEND-PRIMEVUE). PRIMEVUE_COMPONENTS.md
joins the synced doc set so Claude Project Knowledge picks it up on
next upload of .claude-sync/.
Sync output:
- .claude-sync.conf: 34 → 35 entries (+1: PRIMEVUE_COMPONENTS.md)
- .claude-sync/*.md: 34 → 35 files (sync script output;
SYNC_MANIFEST.md auto-regenerated, not counted as net-new)
VUEXY_COMPONENTS.md kept in conf (deprecation stub still useful as a
forwarding marker during F4); removed in F6.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Mirrors CLAUDE.md changes in B3:
- Stack line notes PrimeVue + Tailwind v4 as target, Vuetify as legacy
- New "UI framework strategy (migration-aware)" section forwards to
PRIMEVUE_COMPONENTS.md with surface-level guidance
- Vuexy reference-path section retained but scoped to legacy surfaces
- Vue 3 section split: Tailwind + pt on migrated, Vuetify-first on
legacy
- Top blockquote signals this file evolves as F4 progresses
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
CLAUDE.md updated for the Vuetify→PrimeVue migration phase per
RFC-WS-FRONTEND-PRIMEVUE F2:
- Stack line: notes PrimeVue + Tailwind v4 as target, Vuetify still
present on un-migrated surfaces
- Replaced "Vuexy reference source" + "Vuexy-first strategy" sections
with a single "UI framework strategy (migration-aware)" section that
splits guidance into migrated / un-migrated / new surfaces and
forwards to PRIMEVUE_COMPONENTS.md
- Forms section now documents both target (@primevue/forms + Zod
resolver via FormField) and legacy (ref + VForm + :rules) patterns,
with the surface-level-consistency rule
- UI section reframed: PrimeVue + Tailwind on migrated surfaces,
Vuetify utilities on legacy surfaces, three-state pattern preserved
on both
- Order of work: framework note added for new pages during F4
Framework-agnostic sections (database, multi-tenancy, ULID,
controllers, models, security, testing) untouched.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Vuexy/Vuetify component reference is superseded by PRIMEVUE_COMPONENTS.md
per RFC-WS-FRONTEND-PRIMEVUE. Stub forwards readers to the new doc and
provides the explicit pre-F2 SHA (1c449ff620)
for retrieving the original 777-line content during F4a–F4c on
un-migrated surfaces.
File deleted entirely in F6 cleanup. Stub-not-delete decision per
2026-05-10 project chat (Bert): explicit forwarding marker beats
git-history archaeology while parallel-mode is in force.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Foundation document for F2 of RFC-WS-FRONTEND-PRIMEVUE. Encodes
Crewli-specific conventions for the Vuetify→PrimeVue migration:
- Component mapping by category (form / layout / data display /
feedback / navigation / overlays), each with a paragraph on
migration spirit; cross-references PrimeVue docs rather than
duplicating reference material
- Aura theme + Crewli teal primary token plan (full token list in
RFC Appendix B; F3 implements)
- Canonical forms pattern: @primevue/forms + Zod resolver +
<FormField> wrapper (full API spec lives in RFC Appendix A —
cross-referenced, not duplicated)
- DataTable conventions: lazy / virtual / column-template, with a
slot translation cheat sheet from VDataTable
- pt API + Tailwind v4 + Aura tokens decision matrix
- Migration phase guidance (surface-level consistency rule, no
back-porting, F6 cliff)
- VIcon stays Iconify-Tabler per RFC AD-5; PrimeIcons not installed
Length: 385 lines. F4 sub-packages will extend §3 as surfaces migrate.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Default parallel execution of sync-check and git-lfs commands within
the pre-push hook deadlocks: both read from stdin (git pipes the push
refspec to pre-push hooks), and two parallel readers never reach EOF.
Add piped: true to force sequential execution. sync-check runs first
(only inspects push_files via lefthook templating, doesn't actually
consume stdin), then git-lfs runs second with clean stdin access.
Observed during chore/test-infra-001 sprint: LFS upload completed
100% but pre-push hook hung indefinitely. Workaround was --no-verify;
this commit removes the need for that.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Marks all three sprint backlog entries Resolved with sprint commit
references and documented deviations:
- TEST-INFRA-001 (b8d18e6, 82af117, f6509d9, 2dfb1e8) — Playwright
foundation operational locally. CI deferred.
- TEST-CONTRACT-001 (2dfb1e8) — 409 conflict shape verified against
real Laravel. Single-context replay instead of two-browser
concurrent edit; UI rollback assertion deferred to F4.
- TEST-VISUAL-001 (f6509d9) — 5 composite baselines from canonical
prototype. Composite-over-isolated rationale: prototype DOM lacks
data-* attributes; isolated artist-name locators would rot. F4
adds isolated baselines using stable data-test-id.
Opens TEST-INFRA-002 for the deferred CI work: Gitea/GitHub Actions
decision, runner image, caching, screenshot-diff artifacts, label-
gated nightly e2e. No deadline; surfaces when first review cycle
feels drift.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
B5 of TEST-INFRA-001 (RFC-WS-FRONTEND-PRIMEVUE Amendment A-1).
- Add dev-docs/ARCH-TESTING.md (~13 KB):
§1 Five-tier pyramid (Unit / Component / Integration / Visual /
E2E) with environment, cost, and purpose per tier
§2 Decision tree — pick by what is being verified, not by speed
§3 Mock-vs-real-backend rules + the self-confirming-bias anti-
pattern that motivated TEST-CONTRACT-001
§4 Visual baseline workflow including the composite-over-isolated
strategy used in B3
§5 CI strategy stub — deferred to TEST-INFRA-002
§6 Conventions + 5 anti-patterns
§7 Vuetify-during-PrimeVue-migration: explicit doc that the
Vuetify plugin in playwright/index.ts is INTENTIONAL TEMPORARY
STATE replaced in F3 by PrimeVue. Forbids the "abstract the UI
framework provider" deferred-cost trap.
§8 Host setup — Node, pnpm, Chromium, Git LFS, MySQL 8, PHP, .env;
known risks (unpkg.com flakiness, shared crewli_test DB)
§9 Deferred work cross-references to BACKLOG entries
- Update CLAUDE.md ### Testing section to reference ARCH-TESTING.md
- Add ARCH-TESTING.md to .claude-sync.conf so the dev-docs sync
pipeline picks it up; sync script run.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
B4 of TEST-INFRA-001 (RFC-WS-FRONTEND-PRIMEVUE Amendment A-1).
- Add api/database/seeders/E2EBaselineSeeder.php — deterministic seed
for Playwright e2e: e2e@test.local user (org_admin) on a fresh org +
event + stage + StageDay + artist + engagement + performance
(version=0). Writes seeded IDs to api/storage/app/e2e-fixtures.json
so the Playwright fixture can construct API URLs without API
discovery calls.
- Add apps/app/tests/playwright-e2e/global-setup.ts — runs
`php artisan migrate:fresh --force --seed` against crewli_test (the
existing PHPUnit MySQL test DB) before the test suite starts.
Uses --env=testing to satisfy the dangerous-bash hook's migrate:fresh
guard.
- Add apps/app/tests/playwright-e2e/utils/fixtures.ts — typed reader
for e2e-fixtures.json. Cached after first read.
- Add apps/app/tests/playwright-e2e/utils/auth.ts — login helper that
POSTs /api/v1/auth/login and returns user/org IDs. Uses Bearer-via-
cookie flow (per api/.../SetAuthCookie.php), not stateful Sanctum.
- Add apps/app/tests/playwright-e2e/timetable/409-conflict.spec.ts —
the contract test: first move with version=0 returns 200, second
move with same stale version returns 409 with shape
`errors.conflict: 'version_mismatch'`. Catches the schema-drift
bug class that timetable-stabilization B5 surfaced.
- Update apps/app/playwright.config.ts — wire globalSetup, webServer
for `php artisan serve --port=8001`, baseURL `http://localhost:8001`
(NOT 127.0.0.1 — auth cookie's domain=localhost requires hostname
match).
- Update .gitignore — runtime e2e-fixtures.json never committed.
DoD-19 met locally: `pnpm test:e2e` passes against a real Laravel
test server. CI integration deferred to TEST-INFRA-002 (per A-1
amendment).
Constraint: e2e tests share the crewli_test DB with PHPUnit. Running
both concurrently would collide. Documented in ARCH-TESTING.md (B5).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
B3 of TEST-INFRA-001 (RFC-WS-FRONTEND-PRIMEVUE Amendment A-1).
- Add tests/playwright-ct/visual/static-server.mjs: 60-line Node http
server that serves the canonical prototype directory. No new
dependency added (vs. http-server / serve packages).
- Wire static server into playwright-ct.config.ts via webServer; tests
navigate to http://127.0.0.1:5179/crewli-timetable.html.
- Add tests/playwright-ct/visual/prototype-smoke.spec.ts to verify the
prototype loads in CT runner.
- Add tests/playwright-ct/visual/prototype.spec.ts with 5 @visual
composite baselines:
canvas-friday.png — all status colors, b2b indicators,
multi-lane stacking
canvas-saturday.png — conflict ring + capacity warnings
stage-row-multilane.png — first row in isolation
wachtrij-populated.png — sidebar list with parked + pending
popover.png — block-click popover layout
9 additional surfaces from RFC §A.3's enumerated list are documented
as test.skip() with reasons (cancelled status absent from prototype
data, isolated-block locators would lock to artist names, drag-mode
flaky under simulated pointer events, empty Wachtrij/empty day not
reachable from canonical seed). All deferred to F4 component-level
Vue baselines that will use stable data-test-id attributes.
- Baselines stored at tests/playwright-ct/__screenshots__/visual/
prototype.spec.ts/*.png; tracked via Git LFS (.gitattributes).
Composite-over-isolated rationale: the prototype's DOM exposes status
only via inline style.background, no data-* attributes. Isolated-block
baselines would require artist-name locators that silently rot if
prototype data changes. Composite captures yield the same visual
vocabulary in fewer, more stable images. dev-docs/ARCH-TESTING.md (B5)
documents this strategy and the F4 transition plan.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
B2 of TEST-INFRA-001 (RFC-WS-FRONTEND-PRIMEVUE Amendment A-1).
- Add tests/playwright-ct/utils/mountWithProviders.ts: ergonomic
wrapper around Playwright CT's mount() exposing buildMountArgs()
and readNotificationState(). Documents the Vue Test Utils ↔
Playwright CT API divergence (provider plugins must be wired in
beforeMount, not at call time) and the Vuetify-temp lifecycle
(replaced by PrimeVue in F3).
- Add tests/playwright-ct/components/SanityButtonHarness.vue: a
v-btn harness with a click counter; lives in a .vue file so Vite
bundles its CSS-side-effect imports for the browser context
(Playwright CT runs the test orchestrator in Node and components
in a Vite-bundled browser, unlike Vitest's single jsdom graph).
- Add tests/playwright-ct/components/sanity-vuetify.spec.ts: two
tests proving (a) v-btn renders and propagates clicks, (b) the
--v-theme-primary CSS variable resolves to a parseable RGB triplet.
- Update playwright/index.ts: import 'vuetify/styles' so the v-btn
renders with its actual visual appearance (not unstyled). Required
for B3's visual baselines.
3 component tests pass. 402 Vitest tests still pass unchanged.
Lint + typecheck clean on new files.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Trigger: timetable-stabilization sprint (PR #18, #19) surfaced three
diagnostic incidents that the RFC v1.0 sequencing did not anticipate.
Adds TEST-INFRA-001 as prerequisite sprint before F2 (Playwright +
visual regression infrastructure, baselines against prototype HTML).
Extends F5 with dual-tier visual regression scope. Adds R-11 to risk
register, DoD-16 through DoD-20 to Definition of Done.
No changes to F2-F6 internal architecture, Aura preset, FormField API,
Tailwind v4, or bundle size targets.
Effort impact: +5-7 working days. Total now 15-19 days.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Three trigger upgrades + one new entry, in priority order:
TEST-INFRA-001 — trigger upgraded from "before opening Sessie 5" to
"eerstvolgende sprint na merge van fix/timetable-stabilization", with
explicit dependency: ART-S4-UX-PARITY and all Sessie 5+ work gate on
TEST-INFRA-001 merge. Reden quote captures the three sprint-blok
incidents that proved jsdom-tests do not protect against schema /
filter / UX drift.
TEST-VISUAL-001 — scope expanded to use the prototype HTML at
`./resources/Crewli - Artist Timetable Management/` as the visual
baseline source (not hand-curated screenshots). Added explicit state
matrix per surface: PerformanceBlock 8 states + B2B + cascade-pulse;
PerformancePopover full detail; AddPerformanceDialog drag-mode +
button-mode; Wachtrij filtered/grouped axes. Trigger remains "tweede
toevoeging na TEST-CONTRACT-001" inside the TEST-INFRA-001 sprint.
TEST-CONTRACT-001 — unchanged. Trigger ("eerste e2e na TEST-INFRA-001
lands") was already correct.
ART-S4-UX-PARITY (NEW) — captures Bert's screenshot-report findings as
a seed list grouped A/B/C/D (component-shape / interaction / logic /
AddPerformanceDialog two-mode). Explicit pointer at the bottom to the
Phase A finalization report for the full 20-item itemisation with
severity ratings. Trigger gates Sessie 5 + all subsequent Artist-domain
frontend work behind ART-S4-UX-PARITY merge.
Spelling consistency: VEE-001 entry "formalized" → "formalised" to
match British-English already used elsewhere in the doc and now
mandated by the new CLAUDE.md "Diagnostic discipline" section.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
New section in CLAUDE.md after "Order of work for each new module".
Three consecutive incidents in the timetable sprint led to formalising
this principle:
- B1 (controller assumed buggy, seeder was wrong) — Phase A's
schema-verify gate against SCHEMA.md:1285 + RFC §10.2 inverted the
fix direction.
- B5 (enum-shape assumed drifted, decimals were wrong) — Phase A's
field-by-field response audit caught the actual decimal-as-string
drift before any "fix" against the wrong hypothesis was written.
- Timetable UX (test-passing layer diverged from prototype) — the
mechanical-vs-UX split surfaced via browser test, not via the
389-test suite which all agreed with the buggy state.
Pattern across all three: the initial hypothesis was wrong. The fix
prompts ALL gated Phase A as STOP-and-report; the schema/contract/
prototype audit was reviewed before any code was written. Codifying
this as an explicit project principle so future fix prompts inherit
the gate by default.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase A finding A5 traced this race in the browser logs:
GET .../performances?day={festival_id} → 200, 0 results ← wrong day
GET .../children → 200, 3 sub_events
GET .../performances?day={subevent_id} → 200, 13 results ← correct
The pre-fix `isFlatEvent` was:
computed(() => !subEvents.value || subEvents.value.length === 0)
While `subEvents` was still loading (undefined), `!undefined` is `true`,
so isFlatEvent erroneously returned `true` for festivals during the
loading window. dayOptions then took the flat-event branch and seeded
validSubEventIds with the FESTIVAL id. useActiveDay's corrective watcher
rewrote the URL to `?day={festival_id}` and fired a wasted query that
returned zero results (correct semantics — performances live at sub-event
level — but waste + visible URL flicker).
Fix:
computed(() => eventDetail.value?.event_type === 'event')
EventResource always serialises event_type (verified at
api/app/Http/Resources/Api/V1/EventResource.php:26). EventTabsNav
already consumes event_type / is_festival from the same shape
(apps/app/src/components/events/EventTabsNav.vue:175,266) so this is
the canonical signal, not a one-off addition.
New behavior trace:
- Both queries pending → eventDetail=undefined → isFlatEvent=false
→ festival branch returns (subEvents ?? []).map(...)
→ validSubEventIds=[] → activeDayId=null
→ usePerformances.enabled=false → NO fetch
- subEvents resolves first → festival branch populates dayOptions
→ fetch fires with correct sub-event id
- eventDetail resolves first to flat event → flat branch fires
→ fetch with eventDetail.id (correct)
- eventDetail resolves first to festival → still false until subEvents
→ no false-positive flat-event fetch
402 tests still pass; typecheck + lint + production build all green.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
apps/app/tests/unit/schemas/timetableContractShape.test.ts (NEW, 5 tests):
- base shape: one performance with stage assigned + full engagement
(Bert's browser-tested sample, field-for-field). Asserts decimal-as-
string contract on fee_amount/buma_percentage/vat_percentage AND
enum-label wrapper on booking_status AND nested computed object.
- parked shape: stage_id=null, stage=null (Wachtrij case)
- multi-perf shape: two performances sharing engagement_id
(RFC §D17 "Friday + Saturday under one combined deal")
- sanity: individual performanceSchema parses each fixture element
- regression guard: a payload with NUMBER fee_amount throws (locks
out the pre-B5 bug class)
Every fixture spells out explicit `null` for the schema's nullable-but-
required fields (timestamps, notes, deal_breakdown) so the
nullable() vs optional() distinction is exercised, not glossed over.
Schema surface change to support the test:
apps/app/src/schemas/timetable.ts now EXPORTS performanceArraySchema
(previously a private const inside useTimetable.ts).
apps/app/src/composables/api/useTimetable.ts imports the shared one
instead of redeclaring it locally — single source of truth for the
array shape consumers and tests share.
Test count: 397 → 402 (+5). Typecheck clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase A diagnosed the "Kon timetable niet laden" browser symptom as Zod
schema drift. The prompt's hypothesis (enum {value, label} mismatch) was
incorrect — the schema already uses the enumLabel() wrapper for every
enum field. The actual drift is decimal-cast columns: Laravel serialises
`decimal(N,M)` columns as strings to preserve precision, but the schema
expected numbers, so the very first response triggered a ZodError.
Affected fields, all on `artist_engagements`:
fee_amount decimal(10,2) → wire `"11503.58"`, schema was z.number()
buma_percentage decimal(5,2) → wire `"7.00"`, schema was z.number()
vat_percentage decimal(5,2) → wire `"21.00"`, schema was z.number()
deposit_percentage decimal(5,2) → wire `"…"`, schema was z.number()
Backend has no explicit `decimal:N` cast on these columns
(api/app/Models/ArtistEngagement.php:64-85 — the `casts()` method covers
the enums + booleans + dates + integers, but skips decimals).
Per the strategic decision (frontend adapts, backend stays):
- schemas/timetable.ts: four fields → z.string().nullable()
- types/timetable.ts: matching ArtistEngagement interface fields →
`string | null`
- PerformancePopover.vue:129: only consumer doing arithmetic on a
decimal field; coerce at the use site via Number(...).toFixed(2).
Single line.
- tests/component/PerformanceBlock.test.ts + tests/a11y/axe.test.ts:
spot-checked mocks; the two with hand-built engagement payloads
flipped fee_amount/buma_percentage/vat_percentage from numbers to
strings to match the new schema. No other mocks needed updating.
The {value, label} enum wrapper claim in the prompt was specifically
debunked in Phase A — every consumer (Wachtrij, PerformanceBlock,
WachtrijCard, PerformancePopover, AddPerformanceDialog, page entry)
already uses .value/.label access against an enumLabel-wrapped schema.
B6 will lock the wire-format contract with a real-API fixture
regression test.
All 397 tests still pass; typecheck clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
B4 — jsdom-runnable assertions for the structural pieces of B2/B3.
apps/app/tests/unit/lib/timetable/row-height.test.ts (4 tests):
- laneCount=0 → 52px (Math.max(1, 0) fallback path)
- laneCount=1 → 52px (single-lane stage row)
- laneCount=3 → 148px
- laneCount=10 → 484px (10 × 48 + 4)
apps/app/tests/component/StageHeaderCell.test.ts (4 tests):
- row-height-px prop applies as inline blockSize on the root
- prop omitted → no inline blockSize set (legacy `block-size: 100%`
CSS path takes over for any caller still relying on parent-driven sizing)
- 484px for laneCount=10 round-trips through the prop without truncation
- conflict badge renders only when conflictCount > 0 (existing behavior;
locked in as part of touching this surface)
Visual scroll/alignment proof (sticky-left freeze pane, sticky-top axis,
horizontal scroll cohesion across 14 stages, diagonal trackpad scroll,
pixel-perfect header↔row alignment) is deferred to TEST-VISUAL-001
explicitly: jsdom does not compute position:sticky offsets, scrollbar
visibility, layout overflow chains, or scroll containment ancestry. This
is a known limitation of jsdom-based component testing — not a test gap
in this branch. The sticky behavior, z-index ladder, and DOM structure
are all in place per E1-E4; their validation requires a real browser,
which is exactly what the Playwright CT migration on TEST-INFRA-001 +
TEST-VISUAL-001 unlocks.
No existing tests asserted the old broken layout (no references to the
deprecated `tt-page__rows`, `tt-page__stages`, or `<GridBg>` in tests/).
The unused GridBg component file remains on disk; deleting it is a
stylistic cleanup outside this stabilization scope.
Test count: 389 → 397.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Restructures the canvas so the spreadsheet-feel works correctly with
the seeder's 14 stages: horizontal scroll moves the rows AND the
TimeAxis together; vertical scroll moves the rows but keeps TimeAxis
pinned; both panes intersect at a fixed corner cell. Diagonal trackpad
scroll behaves naturally because there's only one scroll container.
DOM restructure (E2 — sticky resolves to its nearest scroll ancestor;
fixed by giving sticky elements the right scroll-container parent
instead of patching with absolute positioning):
.tt-page__canvas position: relative; overflow: auto
└ .tt-page__layout display: grid; grid-template-columns: 200px auto;
inline-size: max-content
├ .tt-page__corner sticky top:0 left:0 z=3
├ .tt-page__axis sticky top:0 z=2 (full 1872px wide, no clip)
└ for each stage:
├ .tt-page__header-cell sticky left:0 z=2
│ └ <StageHeaderCell :row-height-px="row.rowHeightPx">
└ .tt-page__row-cell normal z=1 (height = same value)
└ <StageRow>
Z-index ladder (E1) is documented in the page CSS:
corner=3, axis row=2, header rail=2, row content=1, blocks=auto.
Popover + AddPerformanceDialog stay above via Teleport-to-body.
Drops the broken pre-stabilization layout:
- `grid-template: "corner axis" 28px "stages rows" 1fr / 200px 1fr`
that put ALL stage headers in ONE grid cell (cause of "lanes too tall"
via headers stretching to 100% of the 570px cell)
- nested `overflow: auto` on `.tt-page__rows` (cause of horizontal-scroll
desync — only the rows pane scrolled, axis stayed put)
- `overflow: hidden` on `.tt-page__axis` (E4 — clipped axis ticks beyond
the 1fr cell width)
- `<GridBg :total-height="0" />` which was a no-op anyway; gridlines now
render directly on each `.tt-page__row-cell` background
`inline-size: max-content` on the layout grid forces it wider than the
canvas viewport, so `overflow: auto` on the canvas actually fires a
horizontal scrollbar. Without this, the `auto` second column shrinks to
viewport and nothing overflows.
The page now passes `:row-height-px` to StageHeaderCell (B2 seam, now
load-bearing). Both header and row cell get the same explicit blockSize
inline so the freeze panes align pixel-for-pixel under whatever
laneCount each stage resolves to.
Visual scroll/alignment proof is deferred to TEST-VISUAL-001 — jsdom
cannot verify position:sticky behavior, scrollbar visibility, or pixel
alignment of the freeze panes. This is a known limitation, not a test
gap. B4 covers the structural assertions jsdom CAN verify.
All 389 existing tests still pass; production build smoke clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Pure structural seam — no layout changes yet (B3 wires the page through).
apps/app/src/lib/timetable/row-height.ts (NEW):
computeStageRowHeight(laneCount, laneHeightPx, lanePadPx) — one-line pure
function with the existing math: max(1, laneCount) * (laneHeight + lanePad) + lanePad.
Math.max(1, laneCount) keeps an empty stage row visible at single-lane
height instead of collapsing.
apps/app/src/components/timetable/StageRow.vue:
Switches its inline rowHeightPx computation to call the helper. Behavior
identical (the math was the helper's body).
apps/app/src/components/timetable/StageHeaderCell.vue:
New optional `rowHeightPx?: number` prop. When provided (B3 will pass it
from the page via the same helper), the header root applies blockSize
inline so the sticky-left column aligns pixel-for-pixel with the row.
When omitted, the legacy `block-size: 100%` CSS still applies — every
existing call-site keeps working.
apps/app/src/lib/timetable/index.ts: re-export the new helper.
Tests still green (389 across 54 files); typecheck clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase A diagnosed an empty SPA timetable as a controller filter bug. B1.1's
schema-verify gate proved the opposite: the seeder violates Model A, the
controllers are correct.
Canonical model (Model A) per:
- dev-docs/SCHEMA.md:1285 artist_engagements.event_id → festival OR flat event
- dev-docs/SCHEMA.md:1329 performances.event_id → sub-event OR flat event ("show host")
- dev-docs/RFC-TIMETABLE-Artist-Timetable-Module.md:1247-1257 (§10.2 contract)
"performance.event_id must be flat event OR a sub-event of the
engagement.event_id festival"
- dev-docs/RFC-TIMETABLE-Artist-Timetable-Module.md:455-477 (§D17)
"Friday + Saturday under one combined deal = 1 engagement, 2 performances"
— only works if engagement is at festival level
Controller audit (B1.2): all five filters in
api/app/Http/Controllers/Api/V1/Artist/{PerformanceController,
ArtistEngagementController, StageController}.php already match Model A.
No controller changes needed.
Seeder change (B1.3) — single consistent fix:
ArtistTimetableDevSeeder::seedForFestival now creates one engagement per
(artist, festival) instead of per (artist, sub-event). When the same artist
recurs across iterations on different sub-events, the existing engagement
is reused and another performance is added (the D17 multi-perf path).
Performances continue to carry event_id = sub-event.
Same model fix in seedForSeries (engagement at parent series, performance
at week sub-event).
seedForFlatEvent already conformed (engagement.event_id = performance.event_id
= the flat event itself).
Existence-check semantics shift from `where event_id = $subEvent->id` to
`where event_id = $festival->id` (or $parent->id for series). Numerically
the test counts hold because the bucket-cycling makes scheduled artists
distinct within the festival window.
Tests (B1.4) — new TimetableSeederControllerIntegrationTest with 7 assertions:
- engagement.event_id is at festival level (DB invariant)
- performance.event_id is at sub-event level (DB invariant)
- GET /performances?day={subEvent} returns non-empty + correct event_ids
- GET /performances unfiltered returns all sub-event performances
- GET /performances?stage_id=null returns the seeded parked perf
- GET /engagements returns engagements with event_id = festival
- GET /stages returns 5 stages with event_id = festival
This locks the visible-symptom regression from Session 4: an empty SPA
timetable on a freshly-seeded festival cannot land again silently.
Existing ArtistTimetableDevSeederTest (4 tests) and the broader Artist
suite (121 tests) all stay green. composer analyse + Pint clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The timetable canvas page at /events/{event}/timetable was added in
RFC-TIMETABLE Session 4 but had no UI entry point. EventTabsNav now
exposes it as the "Programma" tab between Artiesten and Briefings on
flat events, and between Artiesten and Briefings on festivals (in the
re-ordered tab list, post-Artiesten / pre-Briefings).
Changes:
- baseTabs gains the Programma entry at position 6 (after Artiesten).
- The festival re-order computed switches from positional indexing
(baseTabs[5], [6], [7]) to name-based lookup via a findTab helper —
insertions to baseTabs no longer break the festival branch.
- Icon: tabler-calendar-time. Conservative Dutch label "Programma" —
doesn't collide with "Programmaonderdelen" (the festival sub-events
page) since festivals see both tabs side-by-side.
vitest.config.ts: extend the component-project AutoImport to include
'vue-router' so tests of components that auto-import useRoute/useRouter
mount cleanly. (EventTabsNav was the first such test.)
tests/component/EventTabsNav.test.ts (NEW, 4 assertions):
- Programma tab is rendered with the correct label
- it carries the tabler-calendar-time icon
- the route binding resolves to the events-id-timetable name with the
/events/.../timetable URL pattern
- the tab is also visible on a festival (re-ordered tab list path)
Mocks the useEvents composables so the component skips its skeleton/
error branches and renders tabs immediately.
Test count: 385 → 389.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Lint cleanup spotted during Phase C — `router.replace` returns Promise<void>
which the no-void rule rejects. The dropped void had no behavioural effect.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three new entries that codify the test-architecture roadmap surfaced
during the Session 4 follow-up:
TEST-INFRA-001 — Migrate timetable component+a11y tests to Playwright
Component Testing. **Trigger: before opening the Sessie 5 prompt.**
Sessie 5 builds Engagement Detail (6 tabs) + Portal pages (drag-to-
reorder, file uploads); adding more jsdom-based tests for those
surfaces compounds the migration cost.
TEST-CONTRACT-001 — End-to-end 409 contract test against running Laravel.
Trigger: first e2e flow added after TEST-INFRA-001 lands. Highest
contract-protection value per line of test code.
TEST-VISUAL-001 — Visual regression baselines for PerformanceBlock
states (RFC D21/D22/D25/D26). Trigger: second addition to the
TEST-INFRA-001 sprint.
ART-S4-TESTS marked ✅ Resolved with the audit trail of all 9 commits
that landed the test coverage closure (252 → 385 tests across both PRs).
.claude-sync/ regenerated by the post-commit hook (gitignored;
re-uploaded to Project Knowledge separately).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two integration tests that drive the entire RFC D17 lifecycle through
the mutation composable + TanStack cache:
1. happy-path lifecycle (5 stages):
- ADD → POST /performances + Idempotency-Key
- DRAG → POST /timetable/move (target_lane=1, version bumps),
server returns cascaded[] sibling — both surface in
the resolved Promise
- RESIZE → POST /timetable/move with new end_at + new version
- PARK → POST /timetable/move with target_stage_id=null
- DELETE → DELETE /performances/p1
final wire: 4 POSTs + 1 DELETE
2. drag rollback on 409:
- server returns version_mismatch
- mutation rejects with VersionMismatchError shape
- notification.show() invoked with the Dutch toast + 'error'
Why not the full page mount: events/[id]/timetable/index.vue requires
EventTabsNav, useEventDetail, useEventChildren, multiple VTabs/VBtn/
VDialog teleports — too brittle for jsdom CI. The end-to-end + visual
flavour of this flow lives on TEST-INFRA-001's Playwright migration
backlog (and TEST-CONTRACT-001 covers the 409 path against a real
backend).
Test count: 383 → 385.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three jsdom axe scans covering the user-facing surface of the canvas.
The scans surfaced two real a11y bugs which are fixed in this same
commit:
1. PerformancePopover — VProgressLinear (advancing aggregate) had no
accessible name. Added aria-label that announces "X van Y secties
afgerond (N%)".
2. AddPerformanceDialog — the icon-only close button (×) was missing
aria-label. Added 'Sluiten'.
Test scenarios:
- PerformanceBlock with focus
- PerformancePopover open
- AddPerformanceDialog open
Page-level axe rules (region, page-has-heading-one, landmark-one-main,
color-contrast) are disabled for fragment scans — they only make sense
on a full page, and color-contrast resolution is jsdom-blind. Both are
covered by Playwright CT in TEST-INFRA-001 / TEST-VISUAL-001.
Test count: 380 → 383.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
11 tests for useTimetableKeyboard (RFC v0.2 D20):
- Arrow Left → nudge(-SNAP_MIN, 0, 0)
- Arrow Right → nudge(+SNAP_MIN, 0, 0)
- Shift+Arrow → nudge(±60min)
- Arrow Up/Down → ±lane
- Shift+Arrow Up/Down → ±stage
- ] / [ → cycle stages preserving time + lane
- Enter → openPopover with the selected performance
- Delete → remove with the selected performance
- Space → drag mode + aria-live announce; Arrow keys accumulate; Enter
commits with the cumulative offset; aria-live announces 'bevestigd'
- Esc cancels keyboard drag, no mutation, aria-live announces 'geannuleerd'
- all keys are no-ops when no performance is selected
Tests the composable directly with a host component that owns a focusable
canvas root and exposes the spies + announce ref — much more reliable
than mounting the whole timetable page (heavy + asynchronous).
Test count: 369 → 380.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Minimal seam in src/composables/api/useTimetableMutations.ts: the move()
mutation's onError now calls useNotificationStore().show(...) on a 409
status. Generic axios errors stay quiet here — the global response
handler in lib/axios/factory.ts already toasts those. RFC D14 wanted
the version-mismatch toast specifically.
apps/app/tests/component/useTimetableMutations.test.ts (NEW, 5 tests):
- on success: returns server payload with bumped version + sends the
Idempotency-Key supplied by the caller
- 409: rejects with VersionMismatchError + notification.show()
invoked once with the Dutch translation + 'error' level
- cascade: success with cascaded[] populated puts those peers into
the result.cascaded array
- Idempotency-Key uniqueness: two distinct logical move() calls send
distinct keys
- Idempotency-Key reuse: caller-controlled retry within the same
logical action sends the SAME key on the wire (so the backend's
60s idempotency middleware dedupes)
The two existing unit-project tests now register a Pinia instance
(createPinia + setActivePinia) so useNotificationStore() resolves.
Existing assertions unchanged.
Test count: 364 → 369.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
4 component tests via mountWithVuexy:
- happy path: valid form values → POST /performances called with the
correct body shape (engagement_id, event_id mapped from dayId,
stage_id, start_at, end_at)
- end_at < start_at → submit blocked, schema-level error visible on
the end_at field
- empty engagement_id → submit blocked, error visible on the engagement_id
field
- cancel button → emits update:modelValue=false
Test seam: AddPerformanceDialog.vue gains `defineExpose({ form, errors,
submit })` so jsdom tests can drive validation deterministically without
piping through Flatpickr / VAutocomplete plumbing. Three lines, exposes
internal refs only — no behavioural change.
VDialog stubbed in the test (it teleports to body, which puts content
outside the wrapper); App* wrappers stubbed (we test the schema +
submit pipeline, not Flatpickr ergonomics).
Test count: 360 → 364.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
17 component-level tests via mountWithVuexy:
Visual states (10):
- status palette × 3 (option, confirmed, cancelled) — asserts both the
CSS class AND that the matching --tt-status-{X}-bg custom property
resolves on :root (proves the token sheet really loaded)
- capacity icon present when crew + guests > stage.capacity
- capacity icon absent when sum ≤ capacity
- capacity icon absent when stage.capacity is null (no warning possible)
- B2B left dot present when b2bLeft prop true
- B2B right dot present when b2bRight prop true
- no dots when neither prop true
- conflict ring class when warnings includes 'overlap'
- cascade-pulse class when pulse=true
- aria-label includes artist + stage + status + HH:mm time window
- tabindex="0" for keyboard focus
Interactions (5, in second describe):
- click → emits select with performance + DOMRect
- pointerdown → emits pointerdown with (event, performance)
- Delete keypress → emits delete
- Enter keypress → emits select
Test count: 333 → 350.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Per Phase A finding A6 — the previous three-watcher Pinia-store design had
no validation. Landing on /events/{e}/timetable?day=DOES_NOT_EXIST quietly
set store.activeDayId to that bogus value and showed an empty page.
Cross-org sub-event IDs were silently accepted (backend OrganisationScope
returned an empty perf list, so the UI looked broken without telling the
user).
New design (Session 4 follow-up Step 5):
- src/composables/timetable/useActiveDay.ts (NEW)
- The URL `?day` is the source of truth; Pinia does NOT hold this value.
- `activeDayId` is a computed: queryDay if it appears in `validIds`,
else the first valid id, else null when the list is empty.
- One corrective watcher (immediate:true, flush:'post') quietly rewrites
the URL when `?day` is missing or invalid; runs after Vue settles and
after validIds has been recomputed from a fresh fetch.
- `setActiveDay(id)` is the user-driven entry point — calls replace().
- Cross-org IDs are blocked transparently: OrganisationScope keeps them
out of validIds, so they fail the .includes() check and fall back.
- src/stores/useTimetableStore.ts
- Removed `activeDayId` state and `setActiveDay()` action; the store
docstring now documents that day-state lives at the URL.
- src/pages/events/[id]/timetable/index.vue
- Replaced the three watchers + onMounted bootstrap with one
`useActiveDay({ queryDay, validIds, replace })` call. The day-change
side-effect watcher (clear drag, deselect performance) stays.
- VTabs binds dayIdRef + setActiveDay directly.
- tests/unit/pages/timetableDaySync.test.ts (NEW, 9 tests)
- Valid ?day=X → activeDayId=X, no URL rewrite.
- Missing / invalid / cross-org ?day → fallback + URL replaced once.
- Empty validIds → activeDayId=null, URL untouched.
- setActiveDay(id) → calls replace.
- setActiveDay(null) → no-op.
- External URL change (browser back) → activeDayId follows.
- validIds populated AFTER mount → fallback fires correctly.
- tests/unit/stores/useTimetableStore.test.ts: assert that activeDayId
and setActiveDay are GONE from the store surface.
Test count: 324 → 333.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Per Phase A finding A5 — Zod schemas in @/schemas/timetable.ts were
types-only; nothing parsed actual server responses. Backend → frontend
contract drift would only surface as TypeError deep in components.
useTimetable.ts queries now parse:
- useStages → stageArraySchema.parse()
- usePerformances → performanceArraySchema.parse()
- useWachtrij → performanceArraySchema.parse()
- useEngagement → artistEngagementSchema.parse()
useTimetableMutations.ts mutations now parse:
- move success → moveTimetableSuccessSchema.parse()
- move 409 errors → moveTimetableConflictSchema.parse() (the .errors
sub-object — see backend canon at TimetableMoveController:64)
- create / updateNotes → performanceSchema.parse()
- createStage / updateStage → stageSchema.parse()
The move() success parse runs OUTSIDE the try/catch so a Zod failure on
a 200 response surfaces as a true error rather than being misclassified
as a 409. Per Phase A finding A8 the conflict shape already matches
backend field-for-field; no schema correction needed, but the parse()
locks future drift in.
Regression test (tests/unit/composables/api/zodParseFailure.test.ts):
- move() success with missing fields → rejects with ZodError
- move() 409 with malformed errors payload → rejects with ZodError
- createStage() with missing fields → rejects with ZodError
Existing test fixture for createStage was missing created_at/updated_at;
fixed in same commit (real backend responses always include them).
Test count: 321 → 324.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Foundation for the upcoming component / integration / a11y tests.
vitest.config.ts now declares two projects:
- "unit" — pure-logic tests under tests/unit/, src/**/__tests__/,
and tests/*.spec.ts (the legacy sanity test).
happy-dom, no Vuetify, fast path.
- "component" — tests under tests/component/, tests/integration/,
tests/a11y/. jsdom, Vuetify inlined via SSR noExternal,
CSS imports processed (so :root token sheet loads), and
no global vue-router mock so the real router can run.
Both share the same alias map and AutoImport bag.
tests/utils/mountWithVuexy.ts (new):
- Real Vuetify with the Crewli theme tokens
- createTestingPinia (actions execute by default; stubActions opt-in)
- vue-router with memory history at the configured initialPath + ?query
- Fresh QueryClient per call (zero cross-test cache leak)
- Notification mock injected via Pinia plugin so any useNotificationStore()
resolves to { show: vi.fn(), hide: vi.fn() } — matches the actual
NotificationStore API surface (per Phase A finding A4)
- Imports `@/styles/tokens/_timetable.css` at module load so JSDOM resolves
var(--tt-…) when components call getComputedStyle()
tests/setup.component.ts (new):
- vitest-axe matcher registration
- JSDOM polyfills: scrollIntoView, ResizeObserver, visualViewport, body
bounding rect — Vuetify menus / overlays would crash without them
- Deterministic crypto polyfill (mirrors tests/setup.ts so
generateIdempotencyKey() is stable, but without the router mock)
tests/component/_smoke.test.ts (new):
- Mounts a trivial component → asserts wrapper, queryClient, pinia,
router, notificationMock all populated
- Calls getComputedStyle(documentElement).getPropertyValue('--tt-status-confirmed-bg')
→ asserts '#e8f8f0' (proves the CSS token sheet really loaded)
devDependencies added: jsdom, axe-core, vitest-axe, @pinia/testing.
Total: 319 → 321 tests; 42 → 43 files. Both projects green.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Per Phase A finding A2 — `_timetable.scss` was functionally pure CSS:
only :root custom properties + @keyframes + one .tt-cascade-pulse class.
The only SCSS-specific syntax was `// line comments`. Zero $vars, @use,
@mixin, @function, nesting, or color functions.
Why move to .css: Vitest+jsdom can `import '@/styles/tokens/_timetable.css'`
directly so getComputedStyle() resolves var(--tt-…) in component tests
(needed for the upcoming PerformanceBlock visual-state assertions). SCSS
imports require Vite's SCSS plugin, which the vitest.config.ts intentionally
skips for unit-test speed.
Changes:
- `_timetable.scss` → `_timetable.css` (line comments converted to /* */
block comments; everything else byte-identical)
- `assets/styles/styles.scss`: switch from `@use "@/styles/tokens/timetable"`
to `@import "@/styles/tokens/_timetable.css"`
- Production `npm run build` passes (16s, no asset warnings)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Strict-regex sweep of apps/app/src/ confirms zero VeeValidate usage:
no `from 'vee-validate'` imports, no <Field|Form|ErrorMessage>,
no defineRule(), no useForm(). The 15 prior fuzzy matches were
false positives where /useForm/ matched useFormDraft/useFormSteps/
useFormSchemas/useFormFailures.
Changes:
- Remove `vee-validate` and `@vee-validate/zod` from apps/app/package.json
- Regenerate pnpm-lock.yaml (no other deps shifted)
- CLAUDE.md "Forms": replace VeeValidate prescription with the actual
ref + @core/utils/validators + Zod-payload-schema pattern that the
codebase already uses everywhere
- VUEXY_COMPONENTS.md: correct the stale "Registration uses VeeValidate"
claim (the page actually uses useFormDraft + validators); update the
"Form validation" reference row
- BACKLOG.md: close VEE-001 with the audit trail
All 319 existing tests still pass; vue-tsc clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Drift from Session 4 step 11 — unplugin-vue-components and unplugin-vue-router
regenerated their .d.ts files for the new timetable surface. Was missed in the
original commit because the test runner doesn't trigger regen.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- PerformancePopover.vue — teleported floating panel; closes on Esc; shows
status chip, advancing %, computed Buma/VAT/total cost; deal-summary +
delete + open-detail buttons. Position math (340px wide, 12px margin,
flip side if no room) ports prototype's pickPos verbatim.
- AddPerformanceDialog.vue — Vuetify VDialog + raw ref form pattern (matches
CreateShiftDialog and the rest of the codebase). Uses createPerformancePayloadSchema
for client-side validation; falls back to surface-level errors map per field.
- StageEditor.vue — single-stage CRUD modal with name + capacity + 10-swatch
palette picker. Window.confirm cascade-park warning on delete.
- LineupMatrix.vue — stages × sub-events checkbox matrix; only dirty stages
fire replaceStageDays (atomic per stage).
- Wachtrij.vue — sidebar with search + 9 toggleable status chips with counts;
reads/writes useTimetableStore.statusFilter and searchQuery.
- WachtrijCard.vue — initials avatar + status dot + dot label + cancelled
strike-through. role=button, tabindex=0.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
usePointerDrag — PointerEvents primitive with capture, escape-cancel,
keyboard-cancel, and onBeforeUnmount cleanup. Replaces the legacy
mousedown stack the prototype used.
useDragOrClick — threshold-based drag/click disambiguation (4px Manhattan,
matches prototype audit §4.1). Emits onClick when the pointer never crossed
the threshold; otherwise enters drag mode and emits onDragStart / onDragMove /
onDragEnd. Installs the one-shot capture-phase click suppressor on drag-end
so the synthetic click never opens the popover.
RFC v0.2 D7 — implemented once instead of three times like the prototype.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
PerformanceBlock is the heart of the canvas:
- Status palette via CSS tokens (D21) — one class per booking_status enum value
- Cancelled hatch overlay + line-through (D5)
- Trashed-artist dashed border + ⌂ overlay icon (D27)
- Conflict ring + glow when warnings.includes('overlap') (D5)
- Capacity icon driven by evaluateCapacity() with warn/critical levels (D25)
- B2B left/right dots (D26 — 3-min threshold)
- Cascade-pulse class fired by parent on cascaded[] non-empty (D18)
- aria-label structure per D20: artist, stage, time window, status, advancing
- tabindex 0 + Enter/Space → select; Delete → emit delete
StageRow positions blocks by lane_resolved (D19) — server is authoritative.
StageHeaderCell uses Vuexy VMenu pattern for the per-stage actions.
EmptyDayState routes the user to LineupMatrix when no stages are active.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Extract generateIdempotencyKey() from useFormDraft into reusable lib/
- New types/timetable.ts mirrors PerformanceResource, ArtistEngagementResource,
StageResource, GenreResource and the four enums verbatim
- New schemas/timetable.ts adds zod parsers for runtime validation of API
responses + form payloads (createPerformance, createStage, moveTimetable)
RFC v0.2 §10 contract surface for the upcoming timetable canvas.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two new BACKLOG entries surfaced during Session 3:
- **TECH-OBSERVER-TEST-CONVERGENCE** — track removal of the
artist_advance.bootstrap_on_org_create config flag once the five
FormSchema-counting tests are updated to expect the auto-bootstrapped
schema. Goal: productiegedrag = testgedrag, geen branching.
- **ART-ADVANCE-SECTION-FK** — replace the name-based bridge between
advance_sections (engagement-scoped) and form_schema_sections
(org-scoped) with a real FK. Today's name-match works for default-
seeded schemas but breaks on UI rename and offers no integrity
guarantee. Includes migration outline (form_schema_section_id
nullable FK, ArtistEngagement::created provisioning hook,
best-effort backfill).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
OrganisationObserver was gated on app()->runningUnitTests() — replaced
with config('artist_advance.bootstrap_on_org_create') (default true,
phpunit.xml overrides to false). Behaviour identical, but the seam is
explicit and removable. Tracked for full convergence by new BACKLOG
entry TECH-OBSERVER-TEST-CONVERGENCE — productiegedrag = testgedrag,
geen branching, na test-cleanup.
idempotency_key for the engagement-scoped draft simplified from
'aa-' + sha1(engagement_id)[0:27] to 'aa:' + engagement_id (29 chars,
fits varchar(30)). Same uniqueness guarantee, recognisable shape.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
§17.3 footnote already accurately describes ArtistResolver::fromPortalToken
(checked at commit cc48011). Wired event_id end-to-end on the cleaner
path: FormSubmissionService::createDraft now accepts event_id via the
\$context bag, and the EngagementPortalController passes it from
\$resolved->eventId. Replaces the prior post-save fallback. Per WS-4
denormalisation requirement.
ART-OBSERVER-ADVANCE-AGGREGATE moved from open to closed — landed in
Session 3 as the AdvanceSectionObserver (commit 1716e09).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three backend endpoints under public throttle:30,1:
GET /p/artist/{token} — engagement summary + sections
GET /p/artist/{token}/sections/{section} — form schema + draft values
POST /p/artist/{token}/sections/{section} — section submit
Token resolution via ArtistResolver::fromPortalToken (Step 2). The
master Artist becomes the FormSubmission subject; engagement.event_id
populates form_submissions.event_id per WS-4 denormalisation. Token
mismatches map to 404 (InvalidPortalTokenException), soft-deleted
master artists to 410 Gone (ArtistDeletedException).
Section submit reuses the existing FormBindingApplicator pipeline
(RFC-WS-6 v1.3.1) by dispatching FormSubmissionSectionSubmitted —
no parallel apply path. Drafts are idempotent on
'artist_advance:{engagement_id}', so repeated POSTs find the same
submission. AdvanceSection (engagement-scoped) ↔ FormSchemaSection
bridge: case-sensitive name match against the org's artist_advance
schema; the default seeder names them in lockstep.
Frontend in Session 5 — backend complete here.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Seeds 5 default sections per RFC v0.2 D15 (General Info, Contacts,
Production, Technical Rider, Hospitality) on a per-organisation
artist_advance FormSchema with section_level_submit=true. Each
section ships with 3-4 illustrative form_fields; organisations
customise via the FormBuilder UI later.
Wired into org-creation via the new OrganisationObserver so new
tenants receive the schema automatically. Existing orgs get
coverage via the new artist:seed-advance-default artisan command
(idempotent — orgs that already own a schema are skipped).
Note: introduces a new production-grade default-seeder convention.
Prior FormBuilder defaults were dev-only via FormBuilderDevSeeder
called from DevSeeder::run(). This is the first non-dev path.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Resolves the artist subject + event_id + engagement for the
artist_advance portal flow. Per RFC v0.2 D15 + ARCH-FORM-BUILDER
§17.3 footnote: master Artist is the subject (preserves
form_submissions.subject_type='artist'), engagement provides
event_id (per WS-4 denormalisation), and engagement itself rides
along so callers can resolve advance_section context without a
second query.
Token comparison uses SHA-256 hex digest matching Session 1's
storage shape (commit eb6d396). Two domain exceptions distinguish
404 (no matching token → InvalidPortalTokenException) from 410
(master artist soft-deleted post-engagement → ArtistDeletedException
with engagementId attached).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes ART-OBSERVER-ADVANCE-AGGREGATE. Recomputes
artist_engagements.advancing_completed_count + advancing_total_count
on every section lifecycle event (created / updated-status-only /
deleted). Atomic via DB::transaction + lockForUpdate on both the
parent engagement and the sibling section rows; concurrent section-
status changes serialise correctly. Counter updates use
disableLogging() — counter sync is housekeeping, not audit. The
section's own updated event continues to log via LogsActivity on
AdvanceSection.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Single-count drift: the new park-path explicit activity entry in
LaneCascadeService accesses $parked->engagement?->organisation_id
(same shape as the existing schedule-path access, which the baseline
already accepts). Baseline grew 1740 → 1741 errors; same-shape, no
novel rule.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Surfaced during Session 2 review: events.start_date/end_date (date type)
forces day-boundary semantics in WithinEventBounds. Adding start_time/
end_time would let the Session 4 timetable viewport honour real event
hours and boundary checks reject post-event-close performances.
Cross-cutting schema change — out of scope for Artist Timetable sprint
per Charter §2. Tracked for opportunistic landing alongside a future
events-module sprint OR concrete UX-gap discovery during Session 4.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
LaneCascadeService::move() now calls disableLogging() before every
save inside the transaction (locked performance + cascade-bumped
peers + park-path). The two explicit activity('performance')
->event('moved'|'parked') entries with cascade_count + cascaded_ids
properties are the only audit records per move, matching RFC §8's
"single parent entry summarising the cascade" requirement.
Park path additionally writes an explicit 'performance.parked'
entry per RFC §8 vocabulary instead of falling back to a generic
'updated' auto-log entry.
Two new tests verify:
- cascade move with N peers produces exactly 1 activity entry on
the moved subject and 0 on each cascade-bumped peer
- park writes exactly 1 'parked' entry
PerformanceObserver::saving (version bump) is unaffected:
disableLogging() suppresses only the activity log trait, not
Eloquent model events.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
109 new Larastan findings, all same-shape as patterns already absorbed
in the baseline:
argument.type 18 (baseline had 56)
property.notFound 12 (baseline had 501)
method.notFound 8 (baseline had 31)
missingType.iterableValue 2 (baseline had 98)
Per CLAUDE.md "Larastan static analysis at level 6 with accept-all
baseline. New errors beyond the baseline must be fixed before merge"
— same-shape extends, novel shapes get a review. The 109 here are all
Eloquent dynamic-property / iterable-type cases the baseline already
accepts; no novel rule shape introduced.
Baseline grew 7873 → 8293 lines (1631 → 1740 errors absorbed).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two new tech-debt entries surfaced by Session 2:
AUTH-PERMISSIONS-MIGRATION — Crewli is role-based today; RFC-TIMETABLE
§9 references permission strings. Phase A (2026-05-08) chose Option B
(role-based, with permission strings as docblock references). The
eventual cross-cutting migration is tracked here. Trigger:
customer/charter requirement, not internal preference.
ART-DEMOTE-NOTIFICATION — Session 2's daily option-expiry command
writes activity log only; e-mail to the project leader waits for the
post-Accreditation notification framework.
Also append a Session-2 paragraph to the existing
RFC-TIMETABLE-V0.2-DOC-CLEANUP entry describing the §9 permission-string
mapping decision.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
`artist:demote-expired-options` artisan command finds every
ArtistEngagement still in Option whose option_expires_at has passed,
transitions it back to Draft via the existing state-machine
(transitionStatus), and writes an `option_expired` activity entry
with the original expiry timestamp captured in properties so the
audit log distinguishes system-driven expiries from manual demotions.
Idempotency: the state-machine bails when the engagement is no longer
in Option, so a second run within the same minute is a no-op for any
given row. The auto-logged `updated` row + the explicit
`status_changed` + the `option_expired` entries are emitted only by
the run that actually performs the transition.
Scheduled in routes/console.php daily at 03:00 Europe/Amsterdam,
matching the existing nightly low-traffic window.
Notification (email project leader on demotion) is deferred to the
notification framework that lands post-Accreditation; tracked under
BACKLOG entry ART-DEMOTE-NOTIFICATION.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
LogOptions on Artist, ArtistEngagement, Stage, Performance, Genre now
list the specific attributes the audit log captures (per §8 last
paragraph) instead of logFillable. Each model gets a distinct
log_name (artist / artist_engagement / stage / performance / genre)
so the activity-log filter can scope queries by domain.
tapActivity() on every model adds organisation_id (and event_id where
relevant) to the activity entry's properties. The audit-log filter in
the SPA can then query
`->where('properties->event_id', $event->id)` without joining through
multiple subject types.
Performance gets dontLogIfAttributesChangedOnly(['updated_at',
'version']) so the bookkeeping touch from PerformanceObserver doesn't
generate noise when nothing user-meaningful changed.
Custom activity events emitted by services for the cases where the
auto-log can't infer intent:
performance.moved — LaneCascadeService::move writes a single
parent entry with cascade_count and
cascaded_ids[] after the cascade-bump
commits. Per-row updates still flow
through the model trait so the audit log
shows both the summary and the diffs.
stage.day_added /
stage.day_removed — StageDayService::replaceDays writes one
entry per added/removed event_id, performed
on the parent Stage so the log groups by
stage rather than by pivot row.
stage.reordered — StageService::reorder writes one entry on
the parent Event with the full new
stage_ids[] order.
artist_engagement.
status_changed /
cancelled — ArtistEngagementService::transitionStatus
emits one of these depending on the target
status; pairs with the auto-logged `updated`
row.
The remaining artist_engagement.option_expired event lands in Step 10
when the DemoteExpiredOptions command writes a system-causer entry.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Six thin controllers under app/Http/Controllers/Api/V1/Artist/. Zero
business logic: every mutation routes through a service from
app/Services/Artist/. Authorization via Gate::authorize matching
PersonController convention (request authorize() returns true; gates
fire in the controller).
ArtistController — org-scoped CRUD + restore. Catches
DuplicateArtistException → 409 with
duplicate_artist_id so the dialog can
offer "use existing".
GenreController — org-scoped CRUD; catches GenreInUseException
→ 409 with referencing_artists_count.
ArtistEngagementController — event-scoped CRUD; catches
InvalidStatusTransitionException → 422
with a Dutch-readable message.
StageController — event-scoped CRUD + reorder + replaceDays;
catches StageDaysOrphanedPerformancesException
→ 409 with the orphaned performance ids
and the removed event ids per RFC §10.5.
destroy returns the parked performance
count (cascade-park).
PerformanceController — event-scoped CRUD with index filters
`?day={subevent}` and `?stage_id=null`
(wachtrij). update is non-placement only.
TimetableMoveController — single __invoke for POST /timetable/move.
Catches VersionMismatchException → 409
with current_version + server_data per
RFC D14.
Routes wired into api/routes/api.php nested under the existing
organisations/{organisation}/events/{event} prefix group, matching
PersonController and ShiftController structure. The move endpoint
gets the new `idempotency.60s` middleware alias for R1. `stages/order`
and `stages/{stage}/days` registered before the apiResource so the
literal path wins over the wildcard.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
RFC v0.2 R1 — Idempotency-Key replay window for POST
/api/v1/events/{event}/timetable/move. Narrow scope by design: the
12-hour ARCH §10 default would let a cached cascade-bump response
overwrite a fresh edit; 60 seconds covers honest network retry but
expires before a meaningful conflict can emerge.
Backed by the Laravel Cache facade (Redis in non-test env). Cache key
namespace `idempotency:60s:*` distinct from FormSubmission's
DB-column idempotency. Replays carry an `Idempotency-Replayed: true`
header so observability can distinguish them.
Registered as the route-middleware alias `idempotency.60s` in
bootstrap/app.php; will be applied on the move route in Step 8.
Missing or empty Idempotency-Key returns 400 with
`{"error":"idempotency_key_required"}`.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Six resources under app/Http/Resources/Api/V1/Artist/ matching
FormSubmissionResource conventions (final class, @mixin model,
optional()->toIso8601String, whenLoaded relationships).
GenreResource — id, name, color, sort_order, is_active
ArtistResource — master + lifetime/upcoming engagement counts
computed lazily from the engagements relation
ArtistContactResource — paired with ArtistResource.contacts
ArtistEngagementResource — full deal block with the RFC D26 Buma/VAT
formulas computed live in `computed.*`:
buma_amount = fee × buma_pct/100
IFF Organisation handles BUMA
vat_grondslag = fee + (buma when Organisation)
vat_amount = vat_grondslag × vat_pct/100
when vat_applicable
total_cost = fee + buma + vat + Σ breakdown
Frontend (Session 5) ports the same formula.
StageResource — adds stage_days as a flat array of event_ids
(not nested Event resources, to keep payload
light)
PerformanceResource — `lane` (raw, persisted), `lane_resolved`
(computed per D19), `warnings` (overlap +
B2B at minimum; capacity-warn refined later)
LaneResolver under app/Services/Artist/ is the pure-logic helper that
PerformanceResource calls. Greedy lowest-non-conflicting lane
assignment over the (stage_id, event_id) cohort sorted by start_at
then by raw lane (so cascade-bumped rows stay where they were
visually). Frontend port lands in Session 4.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Created under app/Http/Requests/Api/V1/Artist/, mirroring the
existing FormRequest pattern (final class, authorize() returns true,
controller-level Gate::authorize). One request per CRUD shape plus the
two domain-specific endpoints:
artists create / update
genres create / update (with org-scoped unique)
stages create / update (with event-scoped unique)
stages/order ReorderStagesRequest — permutation check
engagements create / update — per RFC §10.3, with
ContractRequiresFee + OptionExpiresInFuture
conditional rules wired
performances create / update — per §10.2; cross-FK
engagement.event_id ↔ event_id chain
enforced via withValidator closure;
update is non-placement only (placement
edits go through /timetable/move)
timetable/move per §10.4; resolves target_event_id from
target_stage_id + target_start_at via
stage_days, then reuses StageActiveOnEvent
+ WithinEventBounds for downstream rules
stages/{stage}/days §10.5 matrix replace; each event_id must
equal stage.event_id (flat) or be sub-event
(festival)
Custom error messages in Dutch where user-facing. Cross-FK rules that
span request inputs (engagement vs event-id chain, day matrix sub-event
membership) live in withValidator after-closures so the rule cache is
stable per request.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
StageActiveOnEvent — checks the candidate stage_id is linked to the
given event_id via stage_days. Covers performance create/update
(perf.event_id ↔ stage) and the timetable move endpoint
(target_stage_id ↔ resolved target event).
WithinEventBounds — checks a candidate datetime is inside the event's
[start_at, end_at] window. Used for performance start/end dates and
move-target dates against the relevant sub-event for festivals.
OptionExpiresInFuture — conditional rule fired only when
booking_status === 'option'. Asserts option_expires_at is set and in
the future. Implementation of RFC §10.1 transition gate at the
request layer (the service layer enforces the same invariant).
ContractRequiresFee — conditional rule fired only when
booking_status === 'contracted'. Asserts fee_amount is set and > 0.
Same dual-layer enforcement as OptionExpiresInFuture.
All four pass silently when the validated field is null or the
context is irrelevant — the FormRequest still owns the surrounding
required/nullable/exists rules.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
GenreService, ArtistService, ArtistEngagementService (state machine),
StageService, StageDayService, PerformanceService, LaneCascadeService
under app/Services/Artist/. Plain final classes with constructor
injection — matches FormSubmissionService convention.
ArtistEngagementService implements the RFC §10.1 booking_status state
machine: terminal Cancelled/Rejected/Declined, Option requires future
option_expires_at, Contracted requires fee_amount > 0. transitionStatus
is the focused entry point; update() routes through it whenever the
payload mutates booking_status. cancel() composes transitionStatus +
soft delete in one transaction so the existing
ArtistEngagementObserver cascade fires.
LaneCascadeService is the D18 transactional move algorithm. Locks the
dragged Performance row FOR UPDATE, validates client version against
the persisted version (D14), then either parks (stage_id=null, no
cascade) or places onto (stage, event, lane) with single-level
cascade-bump of any time-overlapping rows on the target lane. Returns
a MoveResult value object carrying the moved + cascaded performances
so the controller maps them to API resources without a second query.
StageDayService implements the §10.5 atomic matrix replace. Detects
non-cancelled performances on event_ids about to be removed; throws
StageDaysOrphanedPerformancesException unless force_orphan=true. The
orphans are not deleted — they persist with the same stage_id so they
re-appear when the day re-activates (D5/D27 retention).
ArtistService.create raises DuplicateArtistException carrying the
existing master so the controller can offer a "use existing" choice
instead of forcing the booker to abandon their dialog. ArtistEngagement
defaults buma_handled_by based on artist.agent_company.handles_buma
per RFC D26.
GenreService.delete is hard-blocked (GenreInUseException) when artists
still reference the genre via default_genre_id; the frontend rebinds
those artists first.
StageService.delete cascade-parks performances (stage_id → null, lane
preserved) and returns the parked count for the activity-log entry
the controller writes in Step 9.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ArtistPolicy, ArtistEngagementPolicy, StagePolicy, PerformancePolicy,
GenrePolicy. Role-based authorization mirroring PersonPolicy/ShiftPolicy
pattern: super_admin bypass, org-membership check via wherePivotIn,
event_manager fallback for event-level operations.
Each policy carries a class-level docblock mapping the RFC §9
permission strings (events.view_program, events.manage_program,
organisations.manage_artists, organisations.manage_settings) to the
roles authorised, deferring permission-based authorisation to
AUTH-PERMISSIONS-MIGRATION.
ArtistPolicy.delete additionally guards on no-active-engagements
(D27): blocks soft-delete while any engagement is not Cancelled,
Rejected, or Declined.
PerformancePolicy.move and StagePolicy.reorder reuse canManageProgram
so the move endpoint and stage-reorder share the manage_program
permission semantics.
Auto-discovered by Laravel 11 (policies live at App\Policies\* matching
top-level App\Models\* — no explicit Gate::policy registration needed).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Add the two RFC-TIMETABLE §9 roles. Authorization stays role-based per
Phase A Option B; RFC §9 permission strings map to roles in policy
class docblocks, not seeded as Spatie permissions. The eventual
cross-cutting migration to fine-grained permissions is tracked under
AUTH-PERMISSIONS-MIGRATION.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Schema reality (varchar(64), accommodating SHA-256 hex digest) diverges
from RFC v0.2 §5.3 ("ULID unique nullable"). Session 1 implementation is
correct; RFC needs amendment in next legitimate cycle. Tracked under
RFC-TIMETABLE-V0.2-PORTAL-TOKEN-SCHEMA-AMEND. Distinct from
RFC-TIMETABLE-V0.2-DOC-CLEANUP (which covers stale cross-references).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
advance_submissions.advance_section_id FK changed from cascadeOnDelete
to nullOnDelete; column made nullable. Aligns implementation with
RFC v0.2 §5.4 audit-immutability ("submissions remain for retention
compliance") — when ArtistEngagementObserver::deleted hard-deletes a
section, its submissions persist as orphans rather than disappearing.
Migration edited in place (branch unpushed, dev-only). Observer
docblock + test assertion updated to match. Removed pre-existing
follow-up comment that documented the deviation.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
New Phase C test files:
- tests/Unit/Models/Artist/ArtistDomainModelsTest.php — relationships,
casts, soft-delete trait presence, slug uniqueness within/across
organisations, isParked() helper, AdvanceSection's primary scope,
PURPOSE_SUBJECT_FQCN['artist'] resolves to instantiable class.
- tests/Feature/Artist/ArtistEngagementObserverTest.php — auto-fill
organisation_id from artist, cross-tenant guard throws, soft-delete
cascades to performances + hard-deletes advance_sections.
- tests/Feature/Artist/PerformanceObserverTest.php — version starts
at 0, increments by 1 per UPDATE, no bump on no-op save.
- tests/Feature/Artist/ArtistDomainScopeLeakageTest.php — 5 scoped
models (Artist/Genre/Engagement direct + Stage/Performance FK-chain)
isolate cross-org queries.
- tests/Feature/Artist/ArtistTimetableDevSeederTest.php — fixture-count
smoke (4 stages, 12 stage_days, 6 artists, 12 engagements,
13 performances incl. 1 parked).
Cross-cutting fixes that Phase C surfaced:
- AppServiceProvider: morph-map block 2 extended with the 8 new
artist-domain models (artist_engagement, artist_contact, genre,
stage, stage_day, performance, advance_section, advance_submission).
Block 1 'artist' alias was already wired via PurposeRegistry.
- 5 form-builder backfill tests bumped --step rollback counts by +10
to account for the 10 new May 8 migrations sitting at HEAD between
the test's calibration point and current head.
- phpstan-baseline.neon regenerated (1631 entries) — all errors are
same patterns existing baselined code already exhibits
(Factory generic typing, Model property docblock gaps). Tracked
systematically under TECH-LARASTAN-* in BACKLOG.
Tests: 1646 passing (was 1624 pre-Session-1 → +22 net, no losses).
Larastan: 0 errors over baseline.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
RFC-TIMETABLE v0.2 §5.3 moved portal_token from artists to
artist_engagements (one master artist may have multiple per-event
portal links). PortalTokenController and PortalTokenMiddleware
queried the now-removed artists.portal_token column.
Update both lookups to query artist_engagements.portal_token, joining
to artists for the master name. Response shape unchanged: data.id =
engagement id, data.name = artist name, data.booking_status = engagement
status. Middleware sets portal_context='artist' (unchanged); the
attached portal_person object now carries the engagement row.
PortalTokenSecurityTest seeds artist_engagement rows via a private
helper that writes both an Artist (master) and an artist_engagements
row with the hashed token; test assertions adjusted to check the new
shape (no more milestone fields exposed since they don't exist on
the engagement).
Out of scope refactor disclaimer: this is a forced schema-migration
follow-up, not a Session 2-style controller refactor — the controller
queries the new table with minimal change.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
PortalTokenController stores hash('sha256', \$plainToken) — a 64-char
hex digest. RFC v0.2 §5.3's "ULID unique nullable" annotation is loose;
in practice the column holds a hash, not a ULID. char(26) silently
truncates under MySQL strict mode (1406 Data too long) — surfaced
when PortalTokenSecurityTest exercised the auth path against the new
schema. Widen to varchar(64) to fit the hash.
Schema dump regenerated against crewli_test.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ARCH-09 (Artist Eloquent model + migration) closed under
"Opgeloste items (mei 2026)" with summary of what landed in
RFC-TIMETABLE v0.2 Session 1. Removed from Phase 3 status table
and from "Nieuwe backlog items".
Two new tech-debt entries:
- ART-OBSERVER-ADVANCE-AGGREGATE: AdvanceSection lifecycle observer
to recompute artist_engagements.advancing_*_count, deferred to
Session 3 when section-level submit lands.
- RFC-TIMETABLE-V0.2-DOC-CLEANUP: capture stale ARCH-PLANNED-MODULES.md
cross-references in the Approved RFC v0.2 §1 + §15 for next amendment.
Approved RFCs are not patched ad-hoc.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
§3.2.5: clarify that advance_sections are engagement-scoped (not
artist-scoped). One master artist with two engagements advances each
trajectory independently. Drop the prose section enumeration that
predated the AdvanceSectionType enum and conflated section names
with section types — section type is the enum, name is a free string,
default seeds land in Session 3 with ArtistAdvanceDefault.
§17.3: footnote on the artist_advance row documenting engagement
context resolution — ArtistResolver::fromPortalToken looks up
artist_engagements.portal_token, returns the master Artist as subject,
populates form_submissions.event_id from the engagement.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the pre-RFC-v0.2 design (event-scoped artists, milestone bool
flags, artist_riders, itinerary_items) with the master+engagement
split per RFC-TIMETABLE v0.2 §5.3:
- genres (org-scoped vocab, D24)
- artists (master, org-scoped, slug-unique)
- companies.handles_buma column note
- artist_contacts (master-scoped)
- stages, stage_days (event/sub-event pivot)
- artist_engagements (per-event booking — D9, D10)
- performances (engagement-scoped, nullable stage_id, D13/D14)
- advance_sections (engagement-scoped — was artist_id)
- advance_submissions (audit-immutable per RFC §5.4)
- 7 enums under App\Enums\Artist\ documented in their own subsection
artist_riders and itinerary_items removed — RFC v0.2 §5.3 does not
create them; rider data lives in advance-section submissions, and
itineraries are deferred to a future RFC.
TOC anchor unchanged (slug `#357-artists--advancing` still resolves).
ARCH-PLANNED-MODULES.md was assumed to exist by the RFC's pre-amble
and the original session prompt, but does not — §3.5.7 was already in
SCHEMA.md, so the work is an in-place rewrite. Closes ARCH-09.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The string-literal workaround was added before the Artist model existed
(ARCH-09 prerequisite). With the model now landed (RFC-TIMETABLE v0.2
Session 1), resolve to Artist::class directly so morph-map registration
matches the rest of the registry. MorphMapAlignmentTest still green.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Eight factories with named states (Genre, Artist, ArtistContact, Stage,
ArtistEngagement, Performance, AdvanceSection, AdvanceSubmission).
ArtistTimetableDevSeeder hooked into DevSeeder::seedEchtFeesten after
the form-builder showcase. Produces:
- 4 stages (Mainstage, Havana, Stairway, Socialite) with prototype-style
hex colours
- 4 stages × 3 sub-events = 12 stage_days rows
- 4 genres (Hardstyle, Techno, Indie, Live band)
- 6 master artists, each with one tour-manager ArtistContact
- 12 engagements with status mix (1 Draft, 2 Requested, 3 Option,
2 Confirmed, 3 Contracted, 1 Cancelled). Two artists have two
engagements each (different sub-events) — exercises D17 multi-
engagement-per-artist.
- 13 performances, including one parked (stage_id=null = wachtrij)
and one B2B pair within 3 minutes on Mainstage Saturday to seed
the Session 4 frontend B2B detector.
Also fix LogOptions method name across 8 models: dontSubmitEmptyLogs()
→ dontLogEmptyChanges() (Spatie's actual API; surfaced when DevSeeder
ran).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ArtistEngagementObserver:
- creating: auto-fills organisation_id from parent Artist (RFC v0.2 D10
denormalisation), asserts artist.organisation_id == event.organisation_id;
cross-tenant linkage throws CrossTenantEngagementException (extends
DomainException, included in this commit).
- saving: no-op marker reserved for Session 2 state-machine validation.
- deleted: cascades soft-delete to Performance children, hard-deletes
AdvanceSection children. AdvanceSubmission rows are immutable per
RFC §5.4 and remain attached.
PerformanceObserver:
- saving: increments version by 1 on UPDATE only (D14 optimistic lock).
MoveTimetablePerformanceRequest in Session 2 uses this for concurrent-
edit detection.
Both observers registered in AppServiceProvider::boot.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Anticipatory migrations from 2026-04-08 encoded the old §3.5.7 design
(artists.event_id, advance_sections.artist_id). RFC v0.2 §5.3 replaces
both tables with the engagement model. No model/factory/test/seeder
references exist. Removing before Step 1 ensures the new migrations
match RFC §5.3 verbatim.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Capture inventory, data model, component architecture, interaction
patterns, pure logic algorithms (with verbatim excerpts), design tokens,
and 20 RFC v0.2 observations from the standalone React prototype at
resources/Crewli - Artist Timetable Management/.
Read-only audit; no prototype files modified.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Mark TECH-CHANNEL-AUTH-ORG-ADMIN as resolved with PR reference,
date, and one-paragraph summary of what was delivered.
Three edits:
1. Open entry block removed from "Technische schuld" section.
2. Closure bullet appended under "Opgeloste items (mei 2026)" — full
summary of the three-path auth (submitter / super_admin / org_admin),
pattern source (FormSubmissionActionFailurePolicy::canAccess port),
the audit-surfaced super_admin bypass bonus, test deltas, and
sibling FRONTEND-ECHO-IDENTITY-MATCH-SUBSCRIPTION pointer.
3. Stale forward-reference inside FRONTEND-ECHO-IDENTITY-MATCH-SUBSCRIPTION
updated: "submitter-only voor nu" → "submitter / super_admin /
org_admin van submission's organisatie — TECH-CHANNEL-AUTH-ORG-ADMIN
closed mei 2026". Closes the same no-compromises gap as the FORM-05
stub-status touch-up (PR #12).
Sibling BACKLOG entry FRONTEND-ECHO-IDENTITY-MATCH-SUBSCRIPTION stays
open — that's the frontend portal IdentityMatchBanner work that pairs
with this channel auth extension.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per BACKLOG TECH-CHANNEL-AUTH-ORG-ADMIN.
Four new tests + one deleted; existing three preserved.
NEW:
- test_super_admin_can_subscribe (positive, app-wide bypass via Spatie
HasRoles assignRole('super_admin'))
- test_organisation_admin_of_submission_org_can_subscribe (positive,
pivot-table org_admin → submission's organisation)
- test_organisation_admin_of_different_org_cannot_subscribe (CRITICAL
cross-tenant guard — admin of org B cannot subscribe to a submission
in org A)
- test_regular_organisation_member_cannot_subscribe (org_member role
on the pivot is NOT enough; only org_admin passes)
DELETED:
- test_org_admin_is_currently_denied_per_backlog_entry (the "should
flip" denied-by-default test from PR #11; superseded by the four
positive/negative tests above)
PRESERVED:
- test_submitter_is_authorised
- test_other_authenticated_user_is_denied (User with no organisation
membership → falls through every auth branch)
- test_subscription_is_denied_when_submission_does_not_exist
Test-fixture refinement: makeSubmission() now accepts an explicit
$submitter so positive role-based tests can use a separate User as
submitter, ensuring the submitter short-circuit doesn't accidentally
authorise role-based test subjects.
Test results: 7 passed in this file; 1624 in full suite (was 1621).
0 Larastan errors.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per BACKLOG TECH-CHANNEL-AUTH-ORG-ADMIN.
WS-6 v1.3-delta D2 (PR #1123a5696) introduced submission.{id} private
channel with submitter-only authorization, deferring org-admin auth
to a follow-up after the Spatie Permission helper convention was
audited. This commit closes that follow-up.
Authorization now permits (cheap-first short-circuit):
1. Submitter (submitted_by_user_id === user.id) — unchanged
2. super_admin (Spatie HasRoles app-wide bypass) — audit-surfaced bonus,
matches every analogous policy in the codebase
3. Organisation admins of the submission's organisation — new
Pattern: direct port of FormSubmissionActionFailurePolicy::canAccess.
Spatie teams is disabled in config/permission.php, so org-scoping
lives in the user_organisation pivot table's `role` column with
wherePivot('role', 'org_admin') — codebase canonical (used in 17+
policy sites). withoutGlobalScopes() preserved on both FormSubmission
and Organisation lookups so channel auth is a structural gate, not a
tenant-scoped query.
Inline TODO removed; the BACKLOG entry transitions to resolved in a
follow-up commit on this branch.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Three edits closing concessies surfaced in chat review of the closure
docs-PR:
1. FORM-05 'Resterend werk' sub-paragraph: surgical replacement of
resolveStatus references (method removed in D2, PR #1123a5696).
Updated to describe post-D2 reality: gate + invariant +
handle()-internal status derivation. Ticket stays open (the
detectMatchesByValues extension is unbuilt).
2. FRONTEND-ECHO-IDENTITY-MATCH-SUBSCRIPTION (NEW): tracks the frontend
follow-up where the portal IdentityMatchBanner subscribes to the
submission.{id} channel for live banner updates. Previously
documented in PR #11 body and RFC §Q1 v1.3 add 2 commentary but
without an actionable BACKLOG ticket.
3. HARD-DEADLINE-QUERY-TIMEOUT (NEW): tracks the upgrade from soft
post-call microtime deadline to a hard deadline that can interrupt
hanging MySQL queries (connection-level timeouts, MAX_EXECUTION_TIME
hints, or pcntl_alarm). Previously documented as 'soft deadline
limitation' inline in code comments without an actionable BACKLOG
ticket.
No spec changes; no code changes. Closes the chat-identified gaps so
WS-6 v1.3-delta closure has zero un-anchored mental TODOs.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
§1 Status: add Implementation status line citing D1 (PR #10c6f4d1b)
and D2 (PR #1123a5696), both 2026-05-08.
§10 Document history: append v1.3-delta closure entry summarising what
D1 and D2 each delivered + what remains as separate operational task
(GlitchTip alert rule configuration in the web UI) and frontend
follow-up (Echo subscription).
No spec changes — purely lifecycle marker update.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per RFC-WS-6 §Q3 v1.3 + ARCH-BINDINGS §11.
Nieuwe runbook-sectie §7 (na §6 Audit trail) die de triage-flow
documenteert wanneer GlitchTip een FormBindingApplicatorException
event opbrengt:
- §7.1 failure_response_code classificatie (schema_config_error /
temporary_error / data_integrity_error / unknown_error) drijft het
initiële triage-pad
- §7.2 form_schema.has_public_token tag onderscheidt klant-zichtbare
failures (alert-waardig) van organizer-driven failures (admin-UI only)
- §7.3 retry/dismiss decision-matrix met form-failures:retry artisan
command + DismissalReasonType enum cases
- §7.4 severe-failure escalatie criteria (>10/uur op één schema = P1)
- §7.5 cross-references naar RFC, ARCH-BINDINGS, en erasure-runbook
Companion van de operationele GlitchTip alert-rule (apart geconfigureerd
in de GlitchTip web UI op monitoring.hausdesign.nl).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
~30 new tests + 6 modified covering D2 deliverables.
NEW test files:
- FormSubmissionSubmittedListenerOrderTest: rewritten — flips
identity-match assertion from sync to ShouldQueue + adds AST-level
structural guard that every queued listener has the
apply_status=COMPLETED gate as an early statement
(form-builder.queued-listener.skipped_apply_failed log line + ApplyStatus::COMPLETED check).
- TriggerPersonIdentityMatchOnFormSubmitTest: rewritten — drops
failsafe-pad assertions; adds gate-skip tests (null/PENDING/PARTIAL/FAILED);
invariant-violation throw test; broadcast-dispatch test.
- ApplyBindingsOnFormSubmitTest: extended — initial
identity_match_status='pending' write, apply_completed_at on both
paths, classifier-derived failure_response_code per exception subclass,
unknown_error fallback, deadline wrapper invocation captured by
test double, outer-transaction failure record.
- SyncTagPickerSelectionsOnSubmitGateTest (NEW): canonical skip-log
assertion for null/PENDING/PARTIAL/FAILED apply_status; no-skip-log
assertion for COMPLETED. Uses Log::spy because FormTagSyncService
is final and can't be Mockery-mocked.
- FormBindingApplicatorDeadlineTest (NEW): withDeadline returns clone;
no-deadline path; generous-deadline path; timeout exception thrown
with correct submissionId + reasonCode (temporary_error inherited
via FormBindingInfraException). Uses incident_report purpose for
anonymous-allowed branch to avoid PersonProvisioner constraints.
- RetryServiceFailureClassifierTest (NEW): per-subclass
failure_response_code mapping in recordFailure; apply_completed_at
symmetry-fix coverage.
- SubmissionChannelAuthTest (NEW): submitter authorised, other user
denied, missing submission denied, org admin currently denied
(locks v1 contract per BACKLOG TECH-CHANNEL-AUTH-ORG-ADMIN).
- FormSubmissionResourceIdentityMatchTest: extended — DataProvider
iterates over all six non-person purposes asserting
identity_match=null per RFC §Q2 v1.3 contract.
MODIFIED to fit v1.3 layout:
- IdentityMatchOnSubmitTest: rewritten — directly invokes the listener
with apply_status=COMPLETED pre-set, mirroring ApplyBindings'
happy-path output (the test fixtures lack an identity-key binding
so going through full event dispatch fails at PersonProvisioner).
Drops the failsafe-pad assertion in test_public_submission_marked_pending;
replaces with v1.3 contract: subject_type=null leaves
identity_match_status untouched.
- TagPickerSyncListenerTest: same fix — sets apply_status=COMPLETED
on the submission and invokes the listener directly.
Full suite: 1621 passing (4281 assertions). Larastan: 0 errors.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
WS-6 v1.3-delta D2 ships the broadcast channel auth callback in
routes/channels.php with submitter-only scope. Org-admin access is
deferred because the codebase has no vetted Spatie Permission helper
for organisation-scoped role checks; guessing the API would risk
incorrect authorisation without test coverage.
Tracking entry under "Technische schuld", referenced from the inline
TODO in routes/channels.php and the v1.3-delta D2 PR description.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per RFC-WS-6 §Q1 v1.3 addition 4.
Configurable deadline for FormBindingApplicator::apply(). Default 5
seconds catches the long tail of slow applies before they hang the
public flow. Tunable per environment via FORM_BUILDER_APPLY_DEADLINE_SECONDS.
Consumed by ApplyBindingsOnFormSubmit::handle's withDeadline() call
(landed in Phase B).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per ARCH-BINDINGS §7.1 v1.2 retry-service asymmetry note + RFC-WS-6 §Q3 v1.3 addition 2.
recordFailure() now mirrors ApplyBindingsOnFormSubmit's outer-transaction
failure path:
1. failure_response_code via FormBindingExceptionClassifier::classify($e).
Same classification logic as the listener — single behaviour-change
point per the v1.3-delta D1 design.
2. apply_completed_at = now() — closes the asymmetry where the listener
wrote this column on both happy and failure paths but the retry
service only wrote it on the success path.
recordSuccess() unchanged — already writes apply_completed_at via the
shared transaction block in retry().
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per ARCH-BINDINGS §5.6 v1.2.
The queued tag-sync listener now skips unless apply_status === COMPLETED.
PARTIAL and FAILED both fall through to the early-return — rebuilding
user_organisation_tags against a Person whose tag-binding may have been
the binding that failed would propagate partial state into derived data.
Logs at info level when skipped (form-builder.queued-listener.skipped_apply_failed)
for triage visibility. The fresh() reload is required because the inner-txn
commit happens between dispatch and worker pickup.
ApplyBindingsOnFormSectionSubmitted (the other queued listener under
app/Listeners/FormBuilder/) listens to FormSubmissionSectionSubmitted, a
different event — the §5.6 gate is specifically about
FormSubmissionSubmitted's post-apply-status state, so the section-level
listener is intentionally left without this gate.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per RFC-WS-6 §Q1 v1.3 addition 2.
- routes/channels.php (NEW): authorization callback for the
submission.{id} private channel. v1 authz scope is submitter-only
(matches submitted_by_user_id); org-admin access is deferred per
BACKLOG TECH-CHANNEL-AUTH-ORG-ADMIN. Frontend Echo subscription
lands as a separate frontend follow-up.
- bootstrap/app.php: registers routes/channels.php via withRouting()
channels: parameter. This is NEW broadcasting wiring — Laravel's
broadcasting auth middleware was not previously connected to the
framework. Without this registration the channels file is dead code.
- AppServiceProvider:👢 comment block updated to v1.3 listener
layout (1 sync ApplyBindings + N queued, all gated on
apply_status=COMPLETED per ARCH-BINDINGS §5.6). Comment on
TriggerPersonIdentityMatch flipped from "(sync)" to "(queued
post-v1.3)".
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per RFC-WS-6 §Q1 v1.3 (queueing) + §Q2 (invariant + IdentityMatchInvariantViolation)
+ §Q1 v1.3 addition 2 (broadcast).
- Implements ShouldQueue (was sync). Gate as first statement: skip if
apply_status !== COMPLETED (handles PARTIAL and FAILED identically per
ARCH-BINDINGS §5.6). Logs at info level when skipped for triage
visibility.
- Failsafe-pad removed in favour of strict invariant: subject_type='person'
+ apply_status=COMPLETED implies subject_id IS NOT NULL. Violation throws
IdentityMatchInvariantViolation, routed via Laravel queue worker to
GlitchTip + form_submission_action_failures.
- Status derivation preserved (string semantics 'matched'/'pending'/'none')
— PersonIdentityService::detectMatches returns a Collection; status
computed via user_id check + isNotEmpty(). matchCount derived from
$matches->count() for the broadcast payload only (not persisted).
- Person-not-found between dispatch and worker pickup terminates as
'none' rather than throwing — rare race-window where the person was
deleted; banner gets a sensible final state.
- Dispatches FormSubmissionIdentityMatchResolved on the submission.{id}
private channel after writing the final identity_match_status.
Frontend Echo subscription is a separate follow-up (out of WS-6 scope).
The 4 existing failsafe-pad tests need rewriting in Phase I.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per RFC-WS-6 §Q1 v1.3 addition 1, 4 + §Q3 v1.3 addition 2 + ARCH-BINDINGS §5.3.
- FormBindingApplicator::withDeadline(int) returns a clone configured to
throw FormBindingApplicatorTimeoutException if apply() exceeds the
deadline. Soft post-call microtime check; cannot interrupt mid-query
but catches the long tail. apply() refactored to single-return so the
deadline check sits at one site instead of duplicated.
- ApplyBindingsOnFormSubmit::handle:
- Initial identity_match_status='pending' write inside inner
transaction (when subject is or becomes a person) so HTTP response
carries the right state for the IdentityMatchBanner first-paint
copy. Final state comes from the queued TriggerPersonIdentityMatch
(D2 Phase C).
- Wraps apply() with config('form_builder.apply_deadline_seconds', 5).
- Catch block uses FormBindingExceptionClassifier::classify to write
failure_response_code in the outer transaction alongside
apply_status=FAILED. submission_id from the exception (when in the
binding-applicator hierarchy) is also captured in context JSON.
Tests added in Phase I.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
32 new tests covering D1 deliverables:
- Migration shape (3): failure_response_code column presence,
type/length/nullability, index name. MySQL information_schema
introspection.
- Exception hierarchy (11): abstract base, RuntimeException ancestor,
per-subclass constructor + reasonCode (named-args asserting
submissionId is preserved structurally), Timeout extends Infra and
inherits temporary_error, all subclasses extend base, previous-throwable
chaining works, IdentityMatchInvariantViolation is NOT in the
binding-applicator hierarchy and IS a DomainException.
- FormBindingExceptionClassifier matrix (6): each subclass maps to its
reason code; Timeout dispatches to inherited 'temporary_error';
arbitrary RuntimeException -> 'unknown_error'; IdentityMatchInvariantViolation
-> 'unknown_error' (intentional fallback per docstring).
- FormFieldBindingMergeStrategy::validForTargetType (4 tests covering
the full 4 strategies x 3 target types matrix).
- FormSubmissionIdentityMatchResolved (4): ShouldBroadcast contract,
private channel naming ('private-submission.{id}'), broadcast-as
string, payload assignment.
- FormSubmission failure_response_code cast (4): persists as plain
string, NULL by default, factory state composes with apply_status,
round-trips for all four canonical codes.
Baseline regenerated to absorb new tautological-assertion entries (48
lines) — these are class-hierarchy regression guards that Larastan
correctly flags as statically known. The pattern is established in the
codebase per existing baseline entries for similar tests.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Same root cause as 832375b — the new failure_response_code migration
sits at the top of the WS-5/WS-6 stack, so every test that pins --step
to walk back through that stack needs +1.
- FormFieldOptionsBackfillTest: 6 -> 7 (10 occurrences)
- ConditionalLogicBackfillTest: 10 -> 11 (4 occurrences)
- FormFieldConfigBackfillAndDropTest: 16 -> 17 (1 occurrence)
- FormFieldValidationRuleBackfillTest: 19 -> 20 (7 occurrences)
Total: 22 backfill tests now green again.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per RFC-WS-6 §Q3 v1.3 addition 2.
- Added 'failure_response_code' to FormSubmission $fillable + 'string' cast.
Plain string (not enum) — the exception subclass on
form_submission_action_failures is the canonical classification source;
this column is a denormalised mirror for response-shape rendering.
- Factory fluent state method withFailureResponseCode() with documentation
of the four valid values.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per RFC-WS-6 §Q3 v1.3 addition 2.
Centralises the Throwable -> failure_response_code mapping so the
listener (ApplyBindingsOnFormSubmit::handle catch block) and the
retry-service (FormFailureRetryService::recordFailure) produce
identical classifications. Single behaviour-change point.
Resolution order: FormBindingApplicatorException subclass dispatch via
reasonCode(); fallback 'unknown_error' for anything outside the hierarchy.
Wiring into the listener and the retry service lands in D2.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per RFC-WS-6 §Q1 v1.3 addition 2.
Broadcast event class only — not yet dispatched. D2 wires the dispatch
call into TriggerPersonIdentityMatchOnFormSubmit::handle (after the
final identity_match_status write), and the channel-authorization
callback into routes/channels.php.
Frontend Echo subscription is a separate frontend follow-up (out of
WS-6 v1.3-delta scope).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per RFC-WS-6 §V1 + ARCH-BINDINGS §4.2.
Implements the strategy x target-type validity matrix. Append is the
only non-trivial case: valid only for COLLECTION targets. The
AppendStrategyRequiresCollectionTarget publish-guard uses this method
(D2 wiring confirms call sites; this commit provides the building block).
Existing methods (nullWinnerBehaviour, isValidForScalarTargets) untouched.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per RFC-WS-6 §Q3 v1.3 addition 2 (binding hierarchy) + §Q2 (invariant exception).
- Refactored FormBindingApplicatorException from concrete final to abstract
base. Constructor (submissionId, message, previous?) preserves submissionId
as a public readonly property so D2's outer-transaction handler can write
it structurally to form_submission_action_failures.context JSON without
regex-parsing the message. Replaced public-readonly reasonCode property
with abstract reasonCode(): string method.
- Added 3 reason-coded subclasses:
- FormBindingSchemaConfigException -> 'schema_config_error' (422)
- FormBindingInfraException -> 'temporary_error' (503, NOT final because
Timeout extends it)
- FormBindingDataIntegrityException -> 'data_integrity_error' (422)
- Added FormBindingApplicatorTimeoutException extending FormBindingInfraException
(timeout = temporary infra issue from user perspective; reasonCode inherited).
- Added IdentityMatchInvariantViolation as a sibling DomainException — NOT
in the FormBindingApplicatorException hierarchy because it's thrown
outside the binding-applicator pipeline.
- Migrated 3 existing throw sites in FormBindingApplicator::apply():
- 'no_transaction' -> FormBindingInfraException (developer-error wants
infra-triage workflow: GlitchTip alert + retry-after)
- 'no_schema' -> FormBindingSchemaConfigException
- 'unknown_purpose' -> FormBindingSchemaConfigException
- Updated FormBindingApplicatorIntegrationTest::test_no_transaction_guard_present
to assert against the new throw shape (FormBindingInfraException + new
message string) while preserving the test's intent (guard exists in source).
Wiring (deadline wrapper, classifier integration in listener catch +
retry-service recordFailure) lands in D2.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The forward + rollback migration tests pin --step to a fixed count to
walk the WS-5/WS-6 stack back to known pre-states. The new
2026_05_08_000001_add_failure_response_code_to_form_submissions
migration sits at the top of that stack, so both rollback step counts
need +1 to reach the same destinations.
- pre-WS-5a rollback: --step 21 -> 22 (used twice)
- pre-WS-5b rollback (from fully-forward): --step 19 -> 20 (used once)
Comments updated to enumerate the v1.3-delta D1 migration in the WS-6
group.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per RFC-WS-6 §Q3 v1.3 addition 2 + ARCH-BINDINGS §7.1 v1.2.
Denormalised mirror of the FormBindingApplicatorException subclass
classification, written by ApplyBindingsOnFormSubmit's outer-transaction
catch block (D2) when apply_status='failed'. Drives response-shape copy.
NULL when apply_status is not 'failed'.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Three code-vs-docs drifts surfaced by the 2026-05-08 v1.3-delta audit.
None changes architecture; all three close the gap between code on main
(845b6e6) and the v1.3 amendment text.
- RFC §3 (Q1): apply_status enumerations updated to four cases (added
PARTIAL alongside PENDING/COMPLETED/FAILED). PARTIAL is the
BindingPassResult outcome when the pass committed with mixed
per-binding outcomes; not a separate runtime path. Long-term direction
remains BACKLOG PARTIAL-BINDING-SUCCESS.
- ARCH-BINDINGS §5.6: new "PARTIAL handling" subsection clarifying the
gate treats PARTIAL identically to FAILED until partial-success work
lands. The gate code itself was already correct (strict equality on
COMPLETED); this closes the explanatory gap.
- ARCH-BINDINGS §7.1: status-columns table extended with apply_completed_at
row. Intro line updated. Retry-service asymmetry noted as D2 follow-up
(FormFailureRetryService::recordFailure currently does not write
apply_completed_at; D2 fixes this).
RFC v1.3 -> v1.3.1; ARCH-BINDINGS v1.1 -> v1.2.
Refs: dev-docs/RFC-WS-6.md, dev-docs/ARCH-BINDINGS.md, dev-docs/BACKLOG.md (PARTIAL-BINDING-SUCCESS, unchanged)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
PR-4 commit 3 — closure-bookkeeping nu de implementation-PRs en de
twee runbooks gemerged zijn.
- RFC-WS-7-OBSERVABILITY.md: nieuwe §9 Implementation status (mei 2026)
vat samen welke acceptance criteria via PR-1..PR-4 zijn voldaan en
welke (1, 2, 7, 9, 10) op Bert's deploy-checklist resteren. Pointer
naar ARCH-OBSERVABILITY.md als levende reference; de RFC blijft
historisch document.
- SECURITY_AUDIT.md: nieuwe sectie 'WS-7 Observability — finale audit
(mei 2026)' tussen A13-10 en Positive Findings. Bevat (1) acceptance
criteria checklist met status per criterium, (2) processing register
entry voor GlitchTip (controller-not-processor, retention 90 dagen,
TLS+full-disk-encryption+2FA), (3) zeven security controls die WS-7
introduceert (PII scrubbing, CSP whitelist, sourcemap upload-only,
listener registration discipline, runtime portal-context-split,
multi-tenant tag invariant, impersonation.active binary signal),
(4) pointer naar runbooks/observability-erasure.md voor Art. 17.
- BACKLOG.md: status-overzicht-tabel boven de OBS-entries. Toegevoegd
als entry: OBS-2 (early-pipeline log context, ✅ Resolved), OBS-3
(sentry-context middleware coverage, ✅ Resolved — opgevouwen in
AuthScopeContextListener), OBS-5 (Crewli render handlers report()
invariant, ✅ Resolved via 48f2a00 + ExceptionReportingTest), en
OBS-9 (Active — staging environment GlitchTip CSP whitelist follow-up
bij staging-introductie). Bestaande OBS-1, 4, 6, 7 ongewijzigd
(Active); OBS-8 staat al op Resolved sinds dee1401.
- .claude-sync.conf: drie nieuwe doc-paths toegevoegd
(ARCH-OBSERVABILITY.md, runbooks/observability-triage.md,
runbooks/observability-erasure.md). Post-commit sync-claude-docs
hook regenereert SYNC_MANIFEST.md met deze entries.
Closes WS-7 documentation acceptance criteria 8 (ARCH) en 14
(SECURITY_AUDIT). Resterende criteria (1, 2, 7, 9, 10) zijn
deploy-checklist door Bert.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the WS-6 skeleton with a full post-implementation reference
for the observability stack. Eleven sections covering scope, component
overview, tag taxonomy (replacing RFC §3.6 as source-of-truth), tag
binding architecture, scrubbing semantics, runtime context split, CSP
whitelist, sourcemap upload, GDPR + privacy, maintenance + extension
guidance, plus cross-references.
Form Builder exception classification from the old skeleton §3 is
preserved in §5.4 — concrete answer for which Crewli exception
classes do or do not go to GlitchTip.
Lengte: 730 regels markdown. Closes WS-7 acceptance criterion 8.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
PR-3 follow-up. Live smoke surfaced that the @sentry/vue SDK was
running correctly and emitting events, but Crewli's strict
connect-src directive blocked every POST at the browser layer. No
fallback — events evaporated silently with a CSP-violation log in
DevTools console only.
Updated locations (audited the CSP surface; only two locations actually
need the whitelist):
- apps/app/index.html — dev meta CSP, adds http://localhost:8200 to
connect-src so local dev hits the docker-compose GlitchTip stack.
- deploy/nginx/csp-spa.conf — prod organizer SPA CSP, adds
https://monitoring.hausdesign.nl to BOTH the report-only and enforce
add_header lines so a future flip between modes can't silently break
observability.
NOT updated (deviation from prompt):
- api/config/security.php — the API CSP is `default-src 'none';
frame-ancestors 'none'` for JSON responses. Browsers don't enforce
connect-src on JSON contexts (no document, no fetch origin). Adding
connect-src would be semantically a no-op and confuse the deny-by-
default policy.
Regression guard: tests/Feature/Security/CspConnectsToObservabilityTest.
Reads both the dev meta tag and the prod nginx conf directly (the SPA's
CSP is not Laravel-served, so $this->get() can't reach it). Apply-with-
revert verified: stashing both fixes makes both cases fail with a clear
"Refused to connect because it violates the following CSP directive"
hint; popping the stash restores green.
SECURITY_AUDIT.md A13-9 updated with a WS-7 follow-up note documenting
the GlitchTip whitelist as an explicit security control: outgoing
observability traffic restricted to a single known host.
Test count 1549 to 1551. Larastan + Pint clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
WS-7 PR-3 commit 4. RFC §6 acceptance criteria 4, 5, 6 now satisfied
by the frontend SDK PR; entries marked ✅ with brief implementation
references.
Updated criterion 4 to reference Crewli's actual token-based portal
paths (/portal/advance/:token, /register/:public_token) instead of the
RFC's speculative /p/* — the contextBinding guard detects via
route.meta.public + route.meta.context which is the canonical Crewli
signal already used by other guards.
Added a "Voortgang (mei 2026)" subsection at the end of §6 mapping
each PR to the acceptance criteria it closed, plus what remains for
PR-4 (live smoke, ARCH-OBSERVABILITY.md, alerting config, retention
config, SECURITY_AUDIT.md update).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
WS-7 PR-3 commit 3, RFC §3.5.
- deploy.sh: export VITE_SENTRY_RELEASE=crewli-app@<short-sha> before
the Vite build so the release identifier is inlined into the bundle
via import.meta.env.
- New step 4a after the build: when SENTRY_AUTH_TOKEN and
VITE_SENTRY_DSN_FRONTEND are present, upload sourcemaps via
`npx @sentry/cli@latest sourcemaps upload` to project crewli-app
with --url-prefix=~/assets/ matching Vite's default asset path.
Soft-fails with a warning so deploy can still succeed if GlitchTip
is unreachable.
- Always run `find apps/app/dist -name '*.map' -delete` after upload
(or after skipped upload). No public-mapped sources reach nginx —
RFC §3.5 invariant.
- .gitignore: defensive `apps/app/dist/**/*.map` exclusion (dist/ is
already broadly ignored; this is belt-and-suspenders against
accidental commits of build output).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
WS-7 PR-3 commit 2.
- scrubber.spec.ts (18 cases): mirrors backend PiiScrubbingTest semantics.
Body/header/query scrubbing, form_values wholesale replacement, all
SENSITIVE_BODY_KEYS at top + nested levels, max_depth guard, cookies +
storage + user.cookies sanitisation.
- contextBinding.spec.ts (11 cases): exercises the Vue Router beforeEach
guard against a real router with mocked Sentry scope (capturing every
setTag/setUser call into a per-test buffer). Cases:
- portal-token zone — actor_scope=portal, no user_id
- platform route + super_admin — actor_scope=platform
- platform route without super_admin — does NOT tag platform
- organizer route with active org — actor_scope=organisation +
organisation_id
- organizer route without active org — actor_scope=user, no org tag
- unauthenticated public — actor_scope=anonymous
- actor_type role hierarchy
- RFC §3.8 ULID-only user identity (no email leakage)
- route_name + app=app baseline tags
- cross-zone leak guard: navigating from organizer to portal-token
calls scope.clear() and does not bind user
Frontend test count 223 to 252. Typecheck clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
WS-7 PR-3 commit 1. Frontend mirror of the backend SDK install
(commits bdb89a2..adab3be), wired against the existing apps/app SPA.
- pnpm add @sentry/vue@10.52.0 (pinned).
- src/observability/sentry.ts: initSentry() — empty DSN no-op (RFC §3.3),
errors-only (tracesSampleRate=0, profilesSampleRate=0; RFC §2 amend.B),
sendDefaultPii=false, Console integration off, beforeSend wired to the
scrubber, initial scope tag app=app for GlitchTip filtering.
- src/observability/scrubber.ts: TypeScript port of backend
SentryEventScrubber. RFC §3.7 frontend block — body / header / query
scrubbing, form_values wholesale replacement, cookies wholesale,
defensive strip of contexts.storage and user.cookies, max-depth guard.
- src/observability/contextBinding.ts: Vue Router beforeEach guard that
binds RFC §3.6 auth-scope tags per navigation. Three zones via
route.meta.public + route.path matching:
- portal token zone (meta.public + meta.context=portal) → actor_scope=
portal, no user_id (RFC §3.6 explicit)
- /platform/* with super_admin → actor_scope=platform, no org tag
- default authenticated → actor_scope=organisation when an active
organisation is selected (useOrganisationStore.activeOrganisationId),
otherwise actor_scope=user
- unauthenticated public pages → actor_scope=anonymous
Reads useAuthStore (user, appRoles, isSuperAdmin) and
useOrganisationStore (activeOrganisationId) — corrected vs. RFC's
speculative auth-store API.
- src/observability/index.ts: barrel.
- src/main.ts: initSentry runs before registerPlugins so Sentry's Vue
errorHandler hooks before any plugin or component initialises;
installContextBinding runs after registerPlugins so pinia is up.
- env.d.ts: VITE_SENTRY_DSN_FRONTEND + VITE_SENTRY_RELEASE typed.
- .env.example: new file (didn't exist before) documenting all SPA env
vars including the new Sentry pair.
- vite.config.ts: build.sourcemap=true (RFC §3.5 — generated, uploaded
to GlitchTip by deploy.sh, then stripped before nginx serves dist/).
Typecheck: green. Build: green, *.map files emitted alongside *.js
chunks as expected.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Drie regression-tests die de klasse fouten uit PR-2 nazorg empirisch
voorkomen:
1. test_authenticated_listener_registered_exactly_once
2. test_token_authenticated_listener_registered_exactly_once
3. test_job_processing_tag_listener_registered_exactly_once
— vangen OBS-8 patroon (auto-discovery + explicit listen samen) plus
accidentally-removed registrations door toekomstige refactors. Walk
Event::getRawListeners() en faalt met count != 1 met een duidelijke
message ("auto-discovery re-enabled? OR explicit Event::listen
missing?"). Empirisch geverifieerd: zowel duplicate als missing
registratie wordt gevangen.
4. test_impersonation_active_tag_invariant_on_captured_events
— RFC §3.6 binary signal invariant op een echte HTTP request flow.
Vangt regressie waar de baseline-tag-binding verdwijnt.
BACKLOG.md OBS-8 entry toegevoegd en gemarkeerd als Resolved met
verwijzing naar de drie commits van deze sessie + architecturaal
pattern (explicit > implicit voor observability-kritische bindings).
Test count 1545 to 1549. Larastan + Pint clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
RFC §3.6 vereist impersonation.active als always-present binary signal
op authenticated events. Originele PR-2 architectural-fixes verplaatste
impersonation-tagging naar HandleImpersonation middleware, die alleen
draait bij actieve impersonation. Resultaat: non-impersonation events
hadden GEEN tag, niet 'false' tag — wat filtering op "alle impersonation
events" in GlitchTip stilletjes onmogelijk maakte.
Fix: AuthScopeContextListener::bindForUser() zet baseline 'false';
HandleImpersonation overschrijft naar 'true' + impersonator_user_id
wanneer actief. Default-in-listener, override-in-middleware pattern.
HandleImpersonation deed de override-set al correct sinds commit
9414d09; alleen de baseline ontbrak.
Bert's live verification toonde de gap: super_admin event zonder
impersonation actief, GlitchTip event zonder impersonation.active tag.
Tests:
- test_impersonation_active_default_false_for_non_impersonation_authenticated_event
(was test_authenticated_event_does_not_set_impersonation_tags;
hernoemd + assertion gewijzigd)
- test_impersonation_active_default_false_across_every_actor_scope_branch
walks elke actor_scope branch (user/organisation/platform) en bewijst
baseline geldt uniform — vangt toekomstige refactors die per branch
vroegtijdig returnen.
Test count 1544 to 1545. Larastan + Pint clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Auto-discovery + explicit Event::listen() runt observability listeners
twee keer per event (verified via php artisan event:list duplicate
entries). Vandaag idempotent vanwege scope-tag overwrite semantics, maar
architecturaal onacceptabel — toekomstige additive listeners zouden
onmiddellijk breken zonder waarschuwing.
Optie A (Bert bevestigd, RFC-WS-7 OBS-8): expliciete registraties
behouden in AppServiceProvider::boot(), auto-discovery globaal uit via
->withEvents(discover: false) in bootstrap/app.php. Reden: explicit >
implicit voor observability-kritische bindings — grep-baar, IDE-
navigeerbaar, direct zichtbaar bij code review.
TagJobAttemptOnSentry registratie ook van class-string naar array-
callable vorm gebracht zodat event:list de gebonden methode toont
(consistent met AuthScopeContextListener-registraties).
Test count ongewijzigd op 1544. Larastan + Pint clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Live HTTP smoke test on the post-architectural-fixes branch surfaced
that captured Sentry events carried only route-scope tags (app,
route_name, http.method) — auth-scope tags (user_id, actor_type,
actor_scope) were absent on every request.
Root cause: Sanctum's Guard fires Laravel\Sanctum\Events\TokenAuthenticated
(vendor/laravel/sanctum/src/Guard.php:77) on bearer-token resolution,
NOT Illuminate\Auth\Events\Authenticated. The Authenticated event only
fires from SessionGuard
(vendor/laravel/framework/src/Illuminate/Auth/SessionGuard.php:833),
which Crewli does not use — CookieBearerToken middleware injects the
httpOnly cookie as Authorization: Bearer, then auth:sanctum invokes
Sanctum's Guard. So the listener never ran on Crewli's HTTP path.
Offline tests in AuthScopeContextListenerTest passed because they
dispatch event(new Authenticated(...)) directly, bypassing the Guard
layer. Sanctum::actingAs() in tests has the same blind spot — it
short-circuits the Guard via guard('sanctum')->setUser() and fires
neither event.
Fix:
- New handleTokenAuthenticated(TokenAuthenticated $event) method on
AuthScopeContextListener extracts the user via $event->token->tokenable
and delegates to a private bindForUser() shared with handle().
- AppServiceProvider registers the listener for both Authenticated
(covers SessionGuard / login flow / future authenticators) and
TokenAuthenticated (covers Crewli's bearer-token Sanctum flow).
Regression coverage: AuthScopeBindingHttpFlowTest exercises the real
Sanctum Guard via $user->createToken() + Authorization: Bearer header.
Three cases:
- super_admin on a user-scope route: actor_scope=user, all auth tags
present.
- super_admin on an admin.* route: actor_scope=platform, no
organisation_id (correct platform-mode behaviour).
- org_admin on a route with {organisation} param: actor_scope=
organisation, organisation_id valid ULID.
Test count 1541 to 1544. Larastan clean. Pint clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
RFC §3.6 — context tagging tabel volledig vervangen na de PR-2 follow-up
architecturale fixes. Belangrijkste wijzigingen:
- Tag-binding gesplitst in route-scope (BindSentryRouteContext middleware)
en auth-scope (AuthScopeContextListener op Authenticated event).
- Nieuwe actor_scope tag (organisation/platform/user/anonymous).
- Multi-tenant invariant verfijnd: organisation_id is altijd correct
gerelateerd aan actor_scope in plaats van "altijd aanwezig". Platform-
routes zonder org-context worden niet meer gefabriceerd; default
authenticated user-scope omitt organisation_id (Crewli's User<->Organisation
is many-to-many, geen reliable single-org hint).
- impersonation.* tags expliciet gedocumenteerd als afkomstig uit
HandleImpersonation middleware (post-swap), niet uit auth-listener.
- ActorType waarden bijgewerkt na verwijdering van VOLUNTEER case.
RFC §3.14 — status-note toegevoegd dat D-06 indexes al via Spatie's
nullableMorphs default-migratie zijn aangemaakt, met regression-guard
verwijzing.
§6 acceptance criterium 12 markeert D-06 als al voldaan.
BACKLOG.md krijgt vier nieuwe OBS-entries:
- OBS-1: VOLUNTEER actor_type promotion wanneer rol komt
- OBS-4: PHPUnit metadata deprecation cleanup pre-PHPUnit-12
- OBS-6: sentry-laravel install gap awareness + bootstrap test
- OBS-7: custom render handlers report() invariant + coverage
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
PR-2 verified that Spatie's activitylog default migration creates the
composite indexes RFC-WS-7 §3.14 / addendum D-06 require — via
nullableMorphs('subject') and nullableMorphs('causer'), which emit
indexes named `subject` on (subject_type, subject_id) and `causer` on
(causer_type, causer_id).
This test queries information_schema.STATISTICS and fails if either
composite is missing, regardless of the index name. It guards against
silent regression when:
- A future Spatie major release changes nullableMorphs semantics.
- A developer rewrites the activity_log migration without preserving
the morph indexes.
- A schema-dump regeneration drops them.
Test count 1539 to 1541. Larastan clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
PR-2 live smoke test surfaced that super_admin platform-route
exceptions arrived without organisation_id, and the original RFC §3.6
invariant (always-present organisation_id on authenticated events)
would force misleading attribution if it tried to fill that gap.
Refined invariant: every authenticated event carries actor_scope
(organisation/platform/user/anonymous), AND when actor_scope is
organisation, organisation_id MUST be a valid ULID. Platform-mode
correctly omits organisation_id rather than fabricate one.
Resolution chain in AuthScopeContextListener:
1. {organisation} or {event} URI parameter -> actor_scope=organisation
2. portal_event request attribute -> actor_scope=organisation
3. super_admin on admin.* named route -> actor_scope=platform
(Crewli's platform-admin routes use the admin. name prefix)
4. Default authenticated -> actor_scope=user, no org tag
(User<->Organisation is many-to-many; no reliable single-org hint)
Eight new test cases in AuthScopeContextListenerTest cover each branch
and the conditional invariant, including ULID validity via
Symfony\Component\Uid\Ulid::isValid.
Test count 1531 to 1539. Larastan clean. Pint clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Sentry-context binding split into two responsibilities:
- Route-scope (app, http.method, route_name) stays in middleware on
the api group as BindSentryRouteContext — works on every request,
no auth required.
- Auth-scope (user_id, actor_type) moves to AuthScopeContextListener
on Illuminate\Auth\Events\Authenticated — works on every
authentication mechanism (Sanctum, portal-tokens, future
authenticators) without per-route middleware-attachment. Listener
also augments Log::withContext with user_id (closes OBS-2).
Architecturally fault-preventing rather than fault-detecting: new
authenticated route groups need no separate sentry.context aliasing,
so silent observability gaps are no longer possible (closes OBS-3).
Impersonation tagging is co-located with HandleImpersonation: after
the user-swap, the middleware re-tags Sentry scope with the target
user_id/actor_type and adds impersonation.active /
impersonation.impersonator_user_id / impersonation.session_id. The
Authenticated event fires for the admin (Sanctum's natural flow),
the listener tags the admin, then HandleImpersonation overwrites
post-swap.
Files renamed:
- BindSentryContext -> BindSentryRouteContext (route-scope only)
- BindSentryContextTest -> BindSentryRouteContextTest (4 cases)
Files added:
- AuthScopeContextListener
- AuthScopeContextListenerTest (6 cases)
bootstrap/app.php drops the sentry.context alias and prepends
BindSentryRouteContext to the api group. routes/api.php drops every
sentry.context middleware string from auth:sanctum groups.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
VOLUNTEER was reserved-but-unused. Resolver mapped non-admin
authenticated users to ORG_MEMBER because Crewli has no dedicated
volunteer Spatie role; volunteer-ness is behaviour (shift assignments),
not identity.
Dead enum cases are YAGNI violations under zero-compromise: a future
developer could use the case without realising no resolution path
leads to it, producing a silent no-op. Re-introduce alongside a real
volunteer role split when that lands (BACKLOG OBS-1).
ActorType keeps ORGANIZER_ADMIN, SUPER_ADMIN, PORTAL_TOKEN, ORG_MEMBER,
UNAUTHENTICATED. Tests at 1537, Larastan clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The scrubber is fully stateless. Container-resolution per event was
overhead without value, the closure indirection polluted the config
layer with executable logic, and stack traces showed an anonymous
closure frame instead of the class name.
- SentryEventScrubber::scrub() and its private helpers all become
static methods. No instance fields, so the change is mechanical.
- config/sentry.php before_send switches from a closure that calls
app() to PHP array-callable notation [Class, method]. Symfony
OptionsResolver accepts array-callables for static methods.
- PiiScrubbingTest swaps (new SentryEventScrubber)->scrub(...) for
SentryEventScrubber::scrub(...). Semantics unchanged.
Tests 1537 unchanged. Larastan and Pint clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
PR-2 follow-up. The PR-2 backend SDK install passed unit tests because
they exercised the scrubber and the BindSentryContext scope writer in
isolation, but live exceptions from controllers never reached
GlitchTip — they were correctly logged to laravel.log but the report()
call had no Sentry-aware reporter to invoke.
Root cause: sentry-laravel 4.x does NOT auto-register an exception
reporter. The host application is required to wire Integration::handles
inside withExceptions in bootstrap/app.php (per the package README and
Sentry docs). Without it, report and Laravels automatic
report-before-render flow only hit the default log channel.
Fix: add Integration::handles at the top of withExceptions so
sentry-laravel registers a reportable callback that calls
captureUnhandledException for every reported throwable. Filtering
remains downstream:
- ignore_exceptions in config/sentry.php drops Validation,
Authentication, Authorization (RFC §3.10).
- SentryEventScrubber::scrub returns null for sub-500 HttpException
via the before_send hook (RFC §3.7).
Regression coverage: tests/Feature/Observability/ExceptionReportingTest
installs a real Sentry client with a recording before_send and exercises
the full request to capture pipeline through the auth and sentry.context
middleware. Five cases: RuntimeException IS captured (with §3.6 tags
attached), ValidationException is not, NotFoundHttpException 404 is
not, AuthorizationException 403 is not, request-context tags ride along
on the captured event.
Test count: 1532 to 1537. Larastan clean. Pint clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
WS-7 PR-2 commit 3. RFC §3.13.
- app/Http/Middleware/BindRequestLogContext.php: tags every Laravel log
line written during the request with request_id, organisation_id,
user_id, and route name. Sets X-Request-Id on the response so the
SPA can correlate to backend log lines via one click.
- Client-supplied X-Request-Id is honoured only if it parses as a ULID
via Str::isUlid. Junk input (empty, non-ULID) is rejected and a
fresh ULID is generated server-side.
- Registered as a global api-group middleware via the prepend list so
it runs before authentication. Unauthenticated 4xx responses still
carry the X-Request-Id header.
- Test count: 1523 to 1532. Larastan clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
WS-7 PR-2 commit 2.
- app/Http/Middleware/BindSentryContext.php: sets RFC §3.6 tags on the
active Sentry scope (app, http.method, route_name, actor_type,
user_id, organisation_id, event_id, impersonation). Multi-tenant
invariant: throws RuntimeException in local/testing when an auth
request to a tenant-scoped route lacks organisation_id; logs a
warning in production so the user flow still completes.
- app/Listeners/Observability/TagJobAttemptOnSentry.php: tags
queue.attempt on the scope from the JobProcessing event. Default
stack-trace grouping preserved per §3.11.
- ActorType: VOLUNTEER case reserved for a future role split. Current
resolver maps non-admin authenticated users to ORG_MEMBER.
- bootstrap/app.php: registers sentry.context alias. Applied inside
auth:sanctum groups in routes/api.php so it runs after auth.
- AppServiceProvider::boot registers the queue listener.
Test count: 1507 to 1523. Larastan clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Operational docs for the GlitchTip stack landed in the previous two
commits.
- dev-docs/GLITCHTIP.md: new runbook covering local dev, project
provisioning + DSN-to-vault flow, production deploy on
monitoring.hausdesign.nl (DNS, DirectAdmin Let's Encrypt, Apache
reverse proxy with WS upgrade), backup install + restore drill,
smoke tests, troubleshooting.
- dev-docs/SETUP.md: services table now includes GlitchTip; new
docker/glitchtip/.env subsection points at the runbook.
- dev-docs/RFC-WS-7-OBSERVABILITY.md §3.1: amended to record that the
same compose file drives local dev (Mailpit at bm_mailpit:1025), so
prod and dev cannot drift.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Daily pg_dump → gzip → retention pipe for the GlitchTip database.
Configurable via env vars (defaults: ./backups/glitchtip, 30-day
retention, glitchtip-postgres container). Streams directly through
gzip so no plaintext dump touches disk; output 0600.
Cron example in the script header. RFC-WS-7-OBSERVABILITY §5
acceptance criterion 11.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
WS-7 PR-1 — bring up self-hosted GlitchTip alongside the existing
dev stack. One compose file is portable to the production monitoring
host (RFC-WS-7 §3.1).
- docker-compose.glitchtip.yml: web/worker/postgres/redis pinned, web
bound to 127.0.0.1:8200, internal network for postgres + valkey.
- docker/glitchtip/.env.example: documented dev defaults + production
checklist; .env itself ignored.
- Makefile: services / services-stop merge both compose files; new
services-glitchtip-status tail target.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two charter amendments from the original WS-7 brief:
- Sentry -> GlitchTip (self-hosted, protocol-compatible). Same
Sentry SDKs on backend (sentry-laravel) and frontend
(@sentry/vue), pointed at a self-hosted GlitchTip DSN. Avoids
Sentry SaaS pricing and keeps event data on infrastructure
Bert controls.
- Performance monitoring out of scope (errors-only). WS-7
delivers exception capture + alerting + scrubbing + RBAC
only. APM/tracing/spans deferred to a later workstream if
ever needed; pre-launch with no users, the cost/benefit
doesn't justify it now.
RFC-as-first-commit pattern (per WS-6) so the scope-alignment
document is in main before any infra/code changes land.
WS-3 PR-C delivered the per-file DELETE/REWRITE/KEEP_AND_PURGE
matrix on all 9 files referenced in the entry, plus the
out-of-scope post-edit-eslint.sh hook fix. Doc-rot removed:
~80 KB of obsolete bootstrap and prompt-template content.
Single SPA, single cookie, single deploy host. WS-3 complete.
Five files removed, all describing project states that no longer
apply post-WS-TOOLING-001:
- .cursor/instructions.md (8.4 KB): Phase 1-4 roadmap with all
checkboxes empty; Phase 1 has been done for ~6 months. Broken
'make portal' target. Content overlaps with CLAUDE.md.
- .cursor/ARCHITECTURE.md (18.9 KB): pre-WS-3 framing (dual SPA,
dual cookies, dual SANCTUM_STATEFUL_DOMAINS) AND pre-Form-Builder
schema (volunteer_profiles, public_forms with JSON fields). All
six sections superseded by SCHEMA.md, AUTH_ARCHITECTURE.md,
design-document.md, API.md, 102_multi_tenancy.mdc.
- dev-docs/MASTER_PROMPT_CC.md (13 KB): 'paste this above every task'
workflow superseded by auto-loaded CLAUDE.md and structured
Claude Chat-authored prompts. Stale dual-SPA + pre-Form-Builder
assumptions throughout.
- dev-docs/MASTER_PROMPT_CURSOR.md (7.5 KB): same workflow obsoletion;
Cursor is now IDE-only (Claude Code does all implementation).
.cursor/rules/ system handles IDE-level guidance.
- dev-docs/dev-guide.md (32 KB): bootstrap-from-scratch document
containing embedded snapshots of pre-Form-Builder CLAUDE.md,
pre-Form-Builder SCHEMA.md, pre-Form-Builder API.md as
copy-paste templates. Section 5 prompts pre-WS-TOOLING-001 era.
Section 6 (agents) overlaps with CLAUDE_CODE_TOOLING.md.
Total: ~80 KB doc-rot removed.
Cross-reference check found four files outside the deleted set
referencing the deleted paths; all updated in the same commit:
- README.md: Documentation table rebuilt around CLAUDE.md +
dev-docs/* (also dropped stale resources/design/ row pointing
at a directory that no longer exists, and corrected docs/*
paths to dev-docs/*)
- dev-docs/CLAUDE_DESKTOP_SETUP.md: dropped MASTER_PROMPT_CC,
MASTER_PROMPT_CURSOR, dev-guide entries from the
bewust-verwijderd exclusion list; updated Gerelateerd pointer
from dev-guide.md -> SETUP.md
- dev-docs/ARCH-CONSOLIDATION-2026-04.md: updated future-distribution
pointer from dev-guide.md -> SETUP.md (sprint briefing is
historical so the change is purely doc-hygiene)
- dev-docs/VIBE_CODING_CHECKLIST.md: removed Dev guide row from
the bestandspaden table
Remaining references in dev-docs/BACKLOG.md (lines 862-869) live
inside the TECH-DOCS-APPS-PORTAL-PURGE entry that closes in the
next commit.
Canonical replacements: dev-docs/SETUP.md (rewritten this PR),
CLAUDE.md, CLAUDE_CODE_TOOLING.md, and the ARCH-*.md series.
Three pre-WS-3 references purged: regex (apps/(app|portal)),
grep (apps/(app|portal)), and the now-obsolete "apps/portal/ is
planned but not present" defensive comment. The $spa variable
becomes redundant with only one SPA — collapsed to direct
apps/app/ references.
Net: simpler script, no behavioural change for actual files in
apps/app/ (still runs pnpm eslint --fix). Files outside apps/app/
were already a no-op.
Single-line fix in the hooks reference table. The post-edit-eslint
hook used to scope to apps/app/ or apps/portal/; post-WS-3 there's
only apps/app/.
Code change in the hook script itself lands in the next commit.
Surgical updates reflecting post-WS-3 single-SPA reality. The
OrganisationScope rules, three-level authorization, and invitation
flow are unchanged — they're still the canonical guidance.
Changes:
- globs: drop apps/portal/**/*.{vue,ts}
- Portal Architecture: "two access modes in apps/portal/" ->
"two access modes under /portal/* routes within apps/app/"
- Token flow URL example: portal.crewli.app -> crewli.app/portal/
with note about 301 redirect from legacy host
- Login flow URL: portal.crewli.app -> crewli.app
- CORS allowed_origins: drop FRONTEND_PORTAL_URL line
- Production example: collapse dual-host to single-host with
pointer to AUTH_ARCHITECTURE.md §11 for the legacy env key
Drops 17 KB of embedded code templates that had drifted from actual
implementations in apps/app/src/ (auth store template still used
localStorage; portal router guards still showed dual-mode logic that
was consolidated to /portal/* routes within apps/app in PR-B1/B2a).
Slim rewrite: principles + file structure + pointers to actual
reference code in apps/app/src/. Globs narrowed to apps/app/**/*
since apps/portal/ no longer exists. Vuexy component selection
deferred to dev-docs/VUEXY_COMPONENTS.md as canonical registry.
Net: ~17 KB -> ~3 KB, less drift surface, points at living code
instead of duplicating it.
Three load-bearing files still described the pre-WS-3 dual-SPA
reality. Surgical edits to reflect the single-SPA architecture
shipped in WS-3 PR-B (B1: portal moves; B2a: auth+routing
consolidation; B2b: server-side cookie consolidation).
CLAUDE.md:
- Quality-gates ts-reset bullet (line 27): "both SPAs" → "the SPA"
- Quality-gates Vitest bullet (lines 30-32): rewrite from "apps/portal
has 113+ tests; apps/app currently has no Vitest setup (TECH-APP-VITEST)"
to current truth: apps/app has Vitest with 213 tests as of PR-B2a.
TECH-APP-VITEST is implicitly closed.
- Repository layout (line 44): drop apps/portal/ bullet; rephrase
apps/app/ as the single workspace
- "Apps and portal architecture" → "App architecture": rewrite for
single-workspace + two access modes. Login-based covers
organizers + volunteers + crew + super_admin (context-routed
in-app via useAuthStore.availableContexts); token-based covers
artists, suppliers, press
- CORS subsection: collapse two-origin config to single origin
(localhost:5174 dev, https://crewli.app prod). Preserve the
existing crewli.nl marketing-only note
WS-TOOLING-001 sections (Larastan, Rector, Telescope tooling
configuration) verified untouched via `git diff CLAUDE.md`.
README.md (line 25): collapse the Applications table from two rows
(Organizer + Portal) to one (SPA). Adjust trailing sentence accordingly.
Makefile:
- .PHONY list: drop `portal`
- help echo: drop "make portal" line
- portal target: removed (the underlying `cd apps/portal && pnpm dev`
would fail since the directory was removed in PR-B1)
Out of scope (deferred to TECH-DOCS-APPS-PORTAL-PURGE backlog item):
.cursor/ instructions, MASTER_PROMPT_*, dev-docs/SETUP, dev-docs/dev-guide,
dev-docs/CLAUDE_CODE_TOOLING. WS-3-SESSION-1C-AUDIT.md skipped (frozen
historical doc).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
dev-docs/AUTH_ARCHITECTURE.md (v1.0 → v2.0):
- Title section updated to single-SPA / single-cookie reality
- Client Applications table collapsed to one row
- Cookie Specification table collapsed to one row (crewli_app_token)
- Token Lifecycle / Validation section: Origin-based resolution
language removed; middleware described as origin-agnostic
- Cross-app isolation paragraph removed (no second app)
- Configuration Reference table marks FRONTEND_PORTAL_URL as legacy,
pointing at TECH-FRONTEND-URL-CONSOLIDATE
- New §11 "History" preserves the pre-WS-3 dual-cookie context for
future readers, mentions PR-B2a + PR-B2b roles in the unwind
dev-docs/BACKLOG.md — three new entries:
- TECH-FRONTEND-URL-CONSOLIDATE: refactor email controllers to drop
per-app URL map (EmailChangeController, PasswordResetController,
PersonController) — low priority, code-cleanliness only
- TECH-DOCS-APPS-PORTAL-PURGE: sweep apps/portal references from
briefing/tooling docs (.cursor/, MASTER_PROMPT_*, SETUP, dev-guide,
CLAUDE_CODE_TOOLING) — single chore(docs) PR, low priority
- OPS — DNS retirement of portal.crewli.app — operational task,
deferred until traffic monitoring confirms zero usage
dev-docs/SECURITY_AUDIT.md:
- A13-1 narrative actualised: pre-WS-3 dual-cookie context kept as
history, status flipped to RESOLVED (the localStorage→httpOnly
migration shipped earlier in the consolidation arc)
- A13-3: status flipped to RESOLVED by WS-3 PR-B2b; description
rewritten to reflect the new postLoginRedirect.ts validator and
the 16 spec coverage
- Priority remediation table item 8 strikes through A13-3
Backend test suite: 1487 passed (unchanged from Commit 2 baseline).
Frontend: 223 passed (unchanged from Commit 1 baseline).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
deploy.sh referenced apps/portal which was deleted in WS-3 PR-B1; the
script has been broken in main since that merge (npm run build
-w apps/portal would fail). Collapse to a single-app build.
Changes:
- deploy.sh: replace dual-build block (build app + portal, verify both
dist/) with single-app build (build app, verify dist/index.html)
- deploy/nginx/csp-portal.conf: deleted (content was identical to
csp-spa.conf — verified before removal)
- deploy/README.md: replace "Portal (portal.crewli.app)" server-block
section with "Legacy portal redirect" — a 301 server block
template that redirects portal.crewli.app → crewli.app preserving
the request URI. Notes that DNS retirement is a separate ops task
Out of scope: actually retiring the portal.crewli.app DNS record
(operational, tracked separately).
bash -n deploy.sh: clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The dual-cookie machinery (crewli_app_token + crewli_portal_token,
Origin-based resolution) was load-bearing only when the second SPA
existed. apps/portal/ was deleted in WS-3 PR-B1; the resolver code
has been carrying dead branches since then. Collapse to one cookie.
Cookie name retained as crewli_app_token — no session breakage on
deploy. crewli_portal_token is fully purged from the server-side.
CookieBearerToken middleware:
- COOKIE_NAMES array → single COOKIE_NAME constant
- resolveCookieName method (Origin/Referer parsing, host+port
matching against frontend_app_url/frontend_portal_url) → removed
- Body collapses to: skip if Authorization header present; else
read crewli_app_token cookie and inject Bearer header
SetAuthCookie trait:
- COOKIE_MAP / resolveCookieName / originMatches → removed
- makeAuthCookie / forgetAuthCookie now take only $token; the
cookie name is the trait's private constant
Five callers updated to drop the resolveCookieName($request) line
and the cookie-name argument: LoginController (3 sites),
MfaVerifyController (1 site), AuthRefreshController (1 site),
LogoutController (1 site), InvitationController (1 site — caller
list in the prompt missed this one but the same pattern applies).
frontend_portal_url config key retained (per Phase A directive Q1):
EmailChangeController, PasswordResetController, PersonController are
non-auth consumers that build per-app URL maps for outbound emails.
The map structure is now functionally redundant (production resolves
all FRONTEND_* env vars to the same host) but stays structurally
intact. Refactor tracked as TECH-FRONTEND-URL-CONSOLIDATE in the
upcoming docs commit.
HttpOnlyCookieAuthTest:
- Removed 4 dual-cookie tests (login_sets_portal_cookie_for_portal_origin,
app_cookie_does_not_authenticate_portal_requests,
portal_cookie_does_not_authenticate_app_requests,
correct_cookie_authenticates_with_matching_origin)
- Renamed login_sets_app_cookie_for_unknown_origin →
login_sets_app_cookie_regardless_of_origin; expanded to four
Origin variants (none, app, unknown, foreign) — pins the new
origin-agnostic contract
- Removed Origin headers from request calls in remaining tests
(now meaningless)
Backend test count: 1491 → 1487 (-4 deleted, dual-cookie tests
encoding the obsolete contract). Pint clean. Larastan clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the WS-3 PR-B2a minimum precaution (`startsWith('/') &&
!startsWith('//')`) with a layered validator that rejects every input
that is not a strict relative path.
isSafeRelativePath rejects:
- Empty / null / undefined input
- Non-`/`-prefixed paths (including leading whitespace)
- Protocol-relative URLs (`//evil.com`)
- Backslash anywhere (browsers normalise `\` → `/` in some contexts;
`/\evil.com` parses as `//evil.com`)
- ASCII control characters `\x00`–`\x1F` and `\x7F` (NUL, tab, LF, CR,
DEL, etc. — header-injection vectors)
- Anything the URL constructor parses to a different origin than the
synthetic invalid origin used as the resolution base
The URL-constructor check is the authoritative guard; the prefix and
character checks are fast pre-filters that short-circuit common
attack shapes without paying the URL allocation.
Test coverage expands from 6 → 16 cases. New cases pin the
backslash, control-character, leading-whitespace, and positive-
character-set contracts. The URL-encoded-slash-in-query case
documents that we don't false-positive on `%2F` in query strings.
Closes A13-3 (open-redirect on post-login).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Documents the deterministic guard-rail layer: hooks reference,
crewli-reviewer subagent usage, three slash commands, how to test
hooks, how to disable temporarily, and the binding design principle
(settings/hooks deterministic, CLAUDE.md advisory — never duplicate).
/sprint-status — branch, last package, uncommitted work, next BACKLOG item.
/review-multitenancy <Model> — model+migration+policy+tests checklist.
/sync-docs — runs the dev-docs sync pipeline and reminds to upload .claude-sync/.
Each command's frontmatter declares a least-privilege allowed-tools list.
Isolated-context code review against the zero-compromise principles.
Read/Grep/Glob/Bash only — no Edit, so the reviewer cannot patch
code. Outputs MUST FIX / SHOULD FIX / CONSIDER, every finding cited
as path:line.
inject-sprint-context.sh fires on SessionStart with matcher=compact
and emits branch, last 10 commits, and the top of BACKLOG.md so
Claude resumes with sprint context after auto-compaction. Output
capped at ~600 tokens.
post-edit-pint.sh runs vendor/bin/pint --dirty from api/ after any
.php edit. post-edit-eslint.sh runs pnpm eslint --fix inside the
matching SPA dir for .vue/.ts/.tsx/.js files under apps/app/ or
apps/portal/. Both exit 0 unconditionally — formatting failures must
not block the agent.
protect-files.sh blocks Edit/Write to secrets, lock files, default
Laravel migrations, the deleted apps/admin/ tree, .claude/ itself,
and dev-docs/SCHEMA.md.
block-dangerous-bash.sh blocks destructive git operations, blanket
dependency updates, and database wipes that aren't scoped to the
testing environment.
Both signal block via exit 2 with a reason on stderr; both stay well
under 500ms per invocation.
Registers PreToolUse, PostToolUse, and SessionStart hooks for the
deterministic guard-rail layer. settings.local.json stays gitignored
for per-user overrides.
Workstream-sized item geborgt voor uniforme typed + runtime-validated
contracts op de API-grens (backend PHP Enums, frontend Zod schemas,
codegen TS types). Scope, sequentie (post-PR-C/WS-7, pre-RFC-FORM-BUILDER-UI),
en open beslissingen vastgelegd. Verwijst naar dev-docs/ARCH-API-VALIDATION.md
skeleton voor architectuur-detail.
Voorkomt dat S3b technische schuld stapelt — landt vóór RFC-FORM-BUILDER-UI
zodat nieuwe composables vanaf dag één het gevalideerde patroon consumeren.
The card consumed the API directly via useVerifyMfa() (TanStack Query
mutation). Per Decision F's intent (store owns business logic, the
component consumes typed results), the card now calls
useAuthStore.verifyMfa() and pattern-matches on the MfaVerifyResult
discriminated union.
Changes:
- MfaChallengeCard: drop useVerifyMfa import; call authStore.verifyMfa
with camelCase args (sessionToken, trustDevice, deviceFingerprint,
deviceName); local isVerifying ref replaces verifyMutation.isPending.
On result.kind === 'authenticated' emit `verified` (no payload —
the store has already refreshed user state); on 'failed' surface
result.reason with a generic fallback.
- emit signature: `verified: [data: unknown]` → `verified: []`.
- login.vue: onMfaVerified no longer calls authStore.refreshUser —
authStore.verifyMfa() refreshes internally. Page just routes to
resolvePostLoginTarget().
Adds 4 vitest specs in components/auth/__tests__/MfaChallengeCard.spec.ts
covering: success path emits `verified` with camelCase args, failure
path shows reason and suppresses emit, trustDevice toggle honours
fingerprint + device name, fallback message when reason is empty.
Test count 209 → 213. Lint + typecheck clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The auth-store merge made portal_events available on the unified
/auth/me response (held in useAuthStore.portalEvents). usePortalStore
now sources userEvents from the auth store, eliminating the duplicate
fetch that the legacy slim usePortalAuthStore had compensated for.
Changes:
- types/auth.ts: add portal_events?: PortalEvent[] to MeResponse
- useAuthStore: add portalEvents ref, populated in setUser from
me.portal_events, cleared in clearState
- usePortalStore: replace loadUserEventsFromApiAndStorage (which
fetched /auth/me) with syncEventsFromAuthStore (which reads
authStore.portalEvents). A reactive watch keeps userEvents in sync
whenever the auth store updates (login, refresh, logout). The
sessionStorage merge stays as offline cache + post-registration
bridge.
- types/portal.ts: drop the now-unused AuthMeUser type — MeResponse
is the canonical shape post-merge.
Boundaries: usePortalStore (stores-portal) statically imports
useAuthStore (stores) — already allowed by the matrix
(stores-portal allow includes stores).
Adds 4 vitest specs covering: userEvents reflects auth.portalEvents,
no apiClient.get('/auth/me') call from the portal store,
sessionStorage fallback when auth has not hydrated, reactive update
on auth.portalEvents change.
Test count 205 → 209. Lint + typecheck clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
usePortalStore now persists state in sessionStorage instead of
localStorage. Tab-close clears the session implicitly; explicit logout
+ 401 paths invoke reset() which iterates the `crewli:portal:` prefix
and removes every key (forward-compatible for future portal-namespaced
state).
Storage keys are renamed under the canonical prefix:
- crewli_portal_user_events_v1 → crewli:portal:events
- crewli_portal_active_event_id_v1 → crewli:portal:activeEventId
The single new prefix-clear function (clearStoragePrefix) replaces the
hand-listed key removals, so future portal-namespaced state additions
need no reset() change.
useAuthStore.handleUnauthorized() (the 401 interceptor target) is now
async and invokes clearAll() — the canonical session-cleanup hub —
restoring the portal-storage cleanup that the deleted
usePortalAuthStore.handleUnauthorized previously owned. The merge in
Phase E left this gap; this commit closes it.
Adds 7 vitest specs in stores/portal/__tests__/usePortalStore.spec.ts
covering: sessionStorage persistence, reset() prefix-iteration,
non-prefixed-key isolation, reactive state reset, useAuthStore.clearAll
+ handleUnauthorized integration.
Test count 198 → 205. Lint + typecheck clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
login.vue is rewritten to consume useAuthStore.login()'s discriminated
union — no more direct apiClient calls or branching on raw API response
shapes. The page maps result.kind to UI/routing decisions only:
- mfa-required → swap to MfaChallengeCard with the typed payload
- authenticated → resolvePostLoginTarget() (?to= relative, else
auth.resolveLandingRoute())
- must-set-password → forward-compatible placeholder route
- failed → field-level errors + rate_limit message branch
resolveLandingRoute() now returns a string path instead of
RouteLocationRaw — the typed router accepts string-paths cleanly,
removes the cast at every call site, and lets useAuthStore.spec.ts +
guards.spec.ts assert the resolved path directly.
A13-3 minimum precaution lives in a new utility:
src/utils/postLoginRedirect.ts. The relative-only check
(`startsWith('/') && !startsWith('//')`) rejects absolute, protocol-
relative, javascript:, and data: schemes. Full domain validation lands
in WS-3 PR-B2b.
6 vitest specs in utils/__tests__/postLoginRedirect.spec.ts cover the
six rejection / passthrough scenarios.
Test count 192 → 198. Lint + typecheck clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds components/shared/ContextSwitcher.vue — a Vuetify menu-button
that renders only when useAuthStore.showContextSwitcher is true (i.e.
the user has both portal and organizer contexts available). Click
calls useAuthStore.setLastContext + resolveLandingRoute and pushes
the new route.
Wired into both layouts:
- PortalLayout.vue: navbar right section, before UserAvatarMenu
- DefaultLayoutWithVerticalNav.vue (organizer navbar host): before
NavbarThemeSwitcher (OrganizerLayout.vue itself is a 10-line
wrapper around DefaultLayoutWithVerticalNav, so the component
wires into the actual navbar host).
Boundaries matrix update: components-shared now allows `stores` so
canonical shared chrome (ContextSwitcher, future global indicators)
can read useAuthStore directly without re-homing to
components/layout/. stores-portal stays disallowed for components-
shared by design — portal-specific state has no place in shared
chrome.
Adds 3 vitest specs covering: visibility gated by
showContextSwitcher, click invokes setLastContext + router.push.
Test count 189 → 192. Frontend lint + typecheck clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
usePortalAuthStore is deleted — its 114 lines were a slim wrapper over
the same /auth/me endpoint useAuthStore already consumes. The merged
store gains the full set of additions Bert specified for B2a:
State:
- availableContexts / defaultContext (from /auth/me contexts block)
- lastContext (localStorage-persisted)
- portalToken (in-memory only, for the bearer-axios flavour)
Getters: isPortalUser, isOrganizerUser, isPlatformAdmin (alias of
isSuperAdmin), showContextSwitcher, hasRole(), hasAnyRole().
Actions: login(), verifyMfa() — both return typed discriminated
unions so login.vue (Phase H) consumes results without branching on
raw API response shapes. setLastContext, setPortalToken,
resolveLandingRoute, clearAll. clearAll dynamically imports
usePortalStore.reset() to clear portal sessionStorage on session-end —
this is the canonical session-cleanup hub now that the merge has
happened.
5 source files migrated from usePortalAuthStore → useAuthStore. The
PortalLayout.spec.ts mock follows. The boundaries matrix gains a
single new edge (`stores → stores-portal`) replacing the deleted
stores-portal/usePortalAuthStore which previously owned that
cross-zone call.
Adds 16 vitest specs in src/stores/__tests__/useAuthStore.spec.ts
covering setUser context hydration, hasRole/hasAnyRole, lastContext
localStorage persistence, resolveLandingRoute precedence
(portal/organizer/super_admin/multi-role/forceContext/forbidden
fallback), portalToken state, and clearAll cleanup.
Test count 162 → 178 (16 new). Frontend lint + typecheck clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The single axios.ts file becomes a directory with:
- factory.ts — createApiClient + the registerDefaultInterceptors /
registerPortalTokenInterceptors seam (preserves the
TECH-AXIOS-STORE-COUPLING decoupling — no store imports inside)
- default.ts — cookie-authenticated client (organizer + cookie-auth
portal flows; existing 45 call sites resolve unchanged)
- portal-token.ts — Bearer-auth client for the artist-advance /
supplier-intake flows (forward-compatible groundwork; no active
consumers today)
- index.ts — re-exports apiClient + portalApiClient + the register* /
createApiClient surface; the existing `import { apiClient } from
'@/lib/axios'` continues to work directory-resolved.
The bindings plugin (plugins/3.axios-bindings.ts) now wires both
clients with a shared deps base + flavour-specific overrides. The
`getPortalToken` callback returns null until Phase E surfaces
`portalToken` on useAuthStore — no current consumers exercise the
Bearer path, so the null-return is intentional placeholder.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Additive enrichment to MeResource — existing fields untouched, MeTest stays green.
New fields:
- contexts.available: list<'portal'|'organizer'> derived from Person + Organisation memberships
- contexts.default: precedence super_admin > organizer > portal > fallback portal
- platform.is_super_admin: bool promoted from app_roles
- organisations[].roles: 1-element array form alongside the legacy scalar role,
forward-compatible for the multi-role pivot work tracked in TECH-PIVOT-ROLES-MULTI
UserFactory gains volunteer(), orgAdmin(), volunteerAndOrganizer(), superAdmin()
state methods — codified role categories for reuse across future workstreams.
Adds forbidden.vue placeholder (PublicLayout) for the context-failure landing in
the upcoming guard rewrite.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Completes WS-3 PR-B1 charter §4.2: portal is fully consumed by
apps/app under /portal/** (authenticated portal routes) and
/register/** (public token-based form-fill). All portal source has
moved or been merged in earlier commits in this PR.
Adaptations from the original prompt's Phase F:
- pnpm-workspace.yaml does not exist at the repo root (the monorepo
isn't a pnpm workspace; each app has its own package.json /
node_modules / scripts). No edit needed.
- Root package.json has no `dev:portal` / `build:portal` scripts.
No cleanup needed.
- Skipped `pnpm -w build` — apps/app builds via its own scripts.
Deletes 384 portal files (build configs, layouts, plugins, vendored
@layouts, public/, dev/prod Dockerfiles, nginx.conf, env.d.ts,
themeConfig, tsconfig, package.json, lockfile, etc.). All authentic
portal logic is preserved in apps/app/src — verified by:
- Vitest 23 / 162 passing
- vue-tsc --noEmit clean
- eslint clean (zero new errors / warnings)
NOT verified at this point: `pnpm build`. The build fails on a
pre-existing missing `flatpickr` stylesheet import in
src/@core/components/app-form-elements/AppDateTimePicker.vue —
present on main pre-PR, unrelated to this work, and tracked
separately. Reproduced on plain `main` without any of these changes.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds the WS-3 §4.2 sub-zone classification to the apps/app
boundaries matrix:
- components-{shared,portal,organizer} alongside the legacy
components type. components/{auth,settings} are folded into
components-shared as the legacy cross-context home for MFA dialogs
+ PasswordRequirements (used by both organizer reset-password and
portal wachtwoord-instellen / profiel).
- composables-forms (src/composables/forms/**) — pure form-runtime
helpers reusable from organizer Form Builder later.
- stores-portal (src/stores/portal/**) — keeps the portal auth +
portal store walled off from the organizer auth surface.
- pages-{register,portal,platform,organizer} alongside the legacy
pages type — register pages cannot reach into stores or
components-portal/-organizer; portal pages cannot reach
components-organizer; organizer + platform pages cannot reach
stores-portal or components-portal.
Cross-context edges are forbidden (organizer ↛ portal,
shared ↛ portal/organizer). Two pragmatic exceptions are documented
inline:
- components-shared accepts the legacy auth/ + settings/ paths
until PR-B2 cleanup re-homes them under shared/{auth,settings}/.
- pages-register may read stores-portal because success.vue
optionally enriches with the portal user when authenticated.
PR-B2 may move success.vue into pages-portal so this drops.
Lint: 0 errors / 0 new warnings (only the pre-existing
boundaries v5→v6 deprecation warnings, which apply to all 19 rules
now). Tests: 23 / 162 pass. Typecheck clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Routing wiring (Phase D of WS-3 PR-B1):
- apps/app/src/plugins/1.router/guards.ts: add a single early-return
carve-out before the org-selection redirect — `if (to.meta.context
=== 'portal') return`. Per ARCH-CONSOLIDATION-2026-04 §4.3,
meta.context is the canonical contract; PR-B2 evolves the guards
from this key to full context-aware logic (post-login landing,
context-switcher, role checks).
- apps/app/env.d.ts: extend RouteMeta with the new layout names
('OrganizerLayout' | 'PortalLayout' | 'PublicLayout'), context,
requiresAuth, requiresToken, navMode, navTitle.
- apps/app/typed-router.d.ts: regenerated by unplugin-vue-router to
pick up portal/* and register/* route names.
- Page meta finalisation: portal pages have layout: 'PortalLayout',
context: 'portal', preserving original requiresAuth + nav fields;
register pages have layout: 'PublicLayout' + public: true (the
apps/app guard convention for public routes, since meta.public is
what the existing guard recognises).
Form-types restructure (boundaries cleanup):
- apps/app/src/composables/forms/types/formBuilder.ts → src/types/forms/
- apps/app/src/composables/forms/utils/{formValidation,validators}.ts
→ src/utils/forms/
- All `@/composables/forms/{types,utils}/*` imports rewritten across
pages, components, composables, tests.
- This avoids a `types → composables` boundaries violation at
src/types/formSchema.ts which re-exports primitives from the
inlined form-schema. types/formSchema.ts now imports from
@/types/forms/formBuilder which is in the same boundaries zone.
Lint cleanup for moved portal sources (apps/portal had no
.eslintrc.cjs; the migrated code now has to pass apps/app's stricter
config):
- axios.isAxiosError → named import { isAxiosError }
(ClaimenTab, RoosterTab, profiel.vue)
- void schemaQuery.refetch() → schemaQuery.refetch()
(register/[public_token].vue)
- if-then-else collapsed to single boolean return (formatFieldValue)
- :delay-on-touch-only="true" → delay-on-touch-only shorthand
(FieldSectionPriority)
- ml-2 class → ms-2 (FieldAvailabilityPicker)
- multi-statement-per-line splits in profiel.vue + spec files
- unused emailConfigured ref removed (profiel.vue)
- one-component-per-file disabled with TODO TECH-WS3-PORTAL-LINT-CLEANUP
ref (FieldOptionsLocale.spec.ts — multi-Wrapper test pattern)
- restored `import Draggable from 'vuedraggable'` after lint:fix
removed it (template-only usage; the import IS needed)
- camelcase param renamed in FieldOptionsLocale harness factory
- typecheck nudge: spec state.data typed via PublicFormSectionOption[] /
PublicFormTimeSlot[] aliases instead of Record<string, unknown>
- PortalLayout.vue: explicit `import { useRoute, useRouter }` so the
vitest mock can intercept (the trimmed AutoImport set doesn't pull
vue-router's auto-imports)
Vitest: 23 / 162 passing. Lint: 0 errors / 0 new warnings (only the
pre-existing boundaries v5→v6 deprecation warnings remain). Typecheck:
clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Migrates the navbar (event/platform two-mode toggle), mobile drawer
with avatar header + logout, RouterView Suspense wrapper, and footer
from apps/portal/src/layouts/portal.vue into the PortalLayout.vue
skeleton from PR-A. The skeleton's structure (VApp / VAppBar / VMain
/ VFooter) is preserved as the outer shell.
Notable adaptations:
- useAuthStore → usePortalAuthStore (renamed in C.3)
- usePortalStore import path → @/stores/portal/usePortalStore
- mobile nav links now point at /portal/evenementen and /portal/profiel
(the new sub-zone paths) instead of /evenementen and /profiel
- explicit `import { useRoute, useRouter }` from vue-router so the
vitest mock can intercept (auto-import not configured for these in
the trimmed test config)
Updated PortalLayout.spec.ts to mock the two pinia stores plus
useSkins, vue-router, UserAvatarMenu, and AppLoadingIndicator. Tests
now assert the auth-conditional rendering: header + drawer hidden
when unauthenticated, main + footer always present.
Also pulls in the @form-schema → @/composables/forms/* import
rewrites in the C.4-moved composables that the previous commit's
rename-only diff left unstaged.
Vitest: 23 files / 162 tests, no errors.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Composables (apps/portal/src/composables → apps/app/src/composables/):
- useFormDraft, publicFormInjection → composables/ (root, used by
shared/public-form components)
- api/usePublicForm, api/usePublicFormSections,
api/usePublicFormTimeSlots → composables/api/ (no collisions)
- api/usePortalShifts, api/usePortalProfile, api/useVolunteerRegistration
→ composables/api/portal/ (subfolder per WS-3 PR-B1 charter to
leave room for organizer-side namesakes without clashes)
- api/useMfa → DELETED (apps/app version is a strict superset
with extra invalidateQueries calls and the admin-reset mutation)
Types (apps/portal/src/types → apps/app/src/types/):
- api, portal-shift, portal, registration → moved
- mfa → DELETED (byte-identical to apps/app/src/types/mfa.ts)
Schemas:
- apps/portal/src/schemas/registrationSchema.ts → apps/app/src/schemas/
Utils:
- deviceFingerprint, paginationMeta → DELETED (byte-identical
duplicates already in apps/app/src/utils/)
Lib:
- apps/portal/src/lib/{axios,query-client}.ts → DELETED. apps/app's
callback-bound axios (post-PR-A) and query-client are the
canonical versions. Portal pages currently importing
`@/lib/axios#apiClient` resolve to apps/app's apiClient with no
behavioral change for cookie-based requests.
Tests: 4 composable specs (useFormDraft x2, usePublicFormSections,
usePublicFormTimeSlots) moved into __tests__/ subdirs alongside
their composables.
@form-schema imports inside the moved files rewritten to
@/composables/forms/*.
Vitest now: 23 files / 162 tests passing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- apps/portal/src/stores/useAuthStore.ts →
apps/app/src/stores/portal/usePortalAuthStore.ts. The export and
defineStore id are renamed (useAuthStore → usePortalAuthStore,
'auth' → 'portalAuth') so it can coexist with the organizer's
apps/app/src/stores/useAuthStore. Lazy import inside
resetPortalStoresSync() updated to the new path.
- apps/portal/src/stores/usePortalStore.ts →
apps/app/src/stores/portal/usePortalStore.ts (no name change —
apps/app does not have a usePortalStore).
All call sites in moved pages/components now import from
@/stores/portal/{usePortalStore,usePortalAuthStore} and call
usePortalAuthStore() instead of useAuthStore().
PR-B2 will merge this back into a single context-aware auth store.
Also includes the C.1 page meta-block updates (layout: 'PortalLayout'
| 'PublicLayout', context: 'portal') that were left unstaged after
the page-rename commit picked up only the path change.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- public-form/** (18 files + 7 component tests) → shared/public-form/**
This is the runtime form-renderer; goes into shared/ because it will
be reused by the organizer-app Form Builder preview (S3b).
- event/{Claimen,Informatie,Overzicht,Rooster}Tab.vue → portal/event/**
- portal/{StatusCard,EventCard,UserAvatarMenu}.vue → portal/** (no
path change — both apps had a portal/ subfolder).
- AppLoadingIndicator.vue, auth/{PasswordRequirements,MfaChallengeCard}.vue,
settings/Mfa{Disable,Email,Totp}SetupDialog.vue: portal copies
deleted as duplicates of pre-existing apps/app components (diffs
were trivial formatting only).
Inside the moved files: rewrote @form-schema/* → @/composables/forms/*
and @/components/{public-form,event/[Tab]} → new sub-zone paths.
Updated apps/app/tsconfig.json to drop the @form-schema path alias
and the packages/form-schema include path. Updated formSchema.ts to
import from @/composables/forms/types/formBuilder. Carried the
crypto polyfill from apps/portal/tests/setup.ts into
apps/app/tests/setup.ts (needed by useFormDraft tests landing in C.4).
NOTE: Some moved tests still fail because they reference portal
composables (usePublicFormSections, useFormDraft) that move in C.4.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Per WS-3 PR-B1 charter §4.2: portal pages relocate into the
single-SPA layout under apps/app/src/pages/portal/** (authenticated
portal context) and apps/app/src/pages/register/** (public
token-based form-fill / confirmation).
Updated meta blocks:
- Portal pages: layout: 'PortalLayout', context: 'portal'
(preserving original requiresAuth + nav fields)
- Register pages: layout: 'PublicLayout' (drop requiresAuth)
Skipped (apps/portal duplicates of pages already in apps/app):
index.vue, login.vue, wachtwoord-{vergeten,resetten}.vue,
verify-email-change.vue. Deleted: [...path].vue (apps/app already
has [...error].vue catch-all).
NOTE: Component/store/composable imports inside these files still
point at apps/portal-relative paths and will be rewritten in the
next commits. Build will not be green again until commit 6
(composables/lib).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Inlines the form-schema source folder (no package.json, alias-only)
into apps/app/src/composables/forms. Drops the @form-schema alias
from apps/app/vite.config.ts (replaced by @/composables/forms via
the existing @ alias). apps/portal vite + vitest configs keep
@form-schema as a temporary alias pointing at the new location so
portal tests/build keep working until apps/portal is removed at the
end of this PR. Two pure-logic form-schema tests moved alongside.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Creates portal/register/shared/forms sub-folders ahead of the moves
in subsequent commits. Empty .gitkeep markers will be replaced by
real content as the moves land.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Records the naming-vs-coverage gap surfaced during WS-6 closure
verification: ARCH-FORM-BUILDER §31 references five integration
contract tests by name that don't exist under those filenames in
api/tests/Feature/FormBuilder/Integration/. Coverage may be intact
under different filenames; only the §31 naming index is stale.
Low priority — defer to whoever next touches FormBuilder
integration tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
WS-6 (FormBindingApplicator pipeline) is fully landed in main —
sessions 1, 2, and 3 all merged. Verification on 2026-05-04
confirmed every RFC-WS-6.md §7 deliverable plus the v1.1/v1.2
addenda. Backend test suite green at 1486 tests, above the RFC
§8 target of 1445-1465.
Adds a closure-marker note documenting what's verified in main
and adds a single status line under §6.2 of the consolidation
plan pointing at it.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-04 23:52:37 +02:00
1023 changed files with 65348 additions and 62517 deletions
description: Reviews code against Crewli's zero-compromise principles. Use after any backend or frontend implementation is complete, before committing. Returns a structured MUST FIX / SHOULD FIX / CONSIDER report.
tools: Read, Grep, Glob, Bash
model: claude-opus-4-7
---
You are a staff engineer reviewing code for Crewli — a multi-tenant Laravel 12 + Vue 3 SaaS platform with zero tolerance for technical debt.
Read /CLAUDE.md and the relevant /dev-docs/SCHEMA.md sections before reviewing. Do not patch code — produce a structured report only.
description: Run the dev-docs sync pipeline and remind to upload .claude-sync/
allowed-tools: Bash(npm:*), Read
---
Run `npm run sync:docs`.
After it completes, read `.claude-sync/SYNC_MANIFEST.md` and print the `Git SHA` and `Generated` lines.
End with this exact warning block:
> ⚠️ Manual step required: upload `.claude-sync/` (including `SYNC_MANIFEST.md`) to Project Knowledge in Claude Chat. The drift-check protocol depends on this. Without upload the sync is stale and Claude Chat will work from outdated dev-docs.
if ! echo"$cmd"| grep -Eq -- '--env=testing\b';then
block "migrate:fresh wipes the database""Add --env=testing to scope this to the test database, or run a non-destructive 'migrate' / 'migrate:rollback'"
block "blanket dependency update bumps everything without review""Use targeted 'composer require <pkg>' or 'pnpm add <pkg>' to bump one package at a time"
block "tooling self-modification — Bert reviews .claude/ changes by hand""Open the file in an editor outside Claude Code, or ask Bert to authorize the change explicitly"
**Golden Rule:** Laravel is exclusively a JSON REST API. No Blade views, no Mix, no Inertia. Every response is `application/json`. Vue handles ALL UI via two SPAs.
---
## Applications
### Organizer App (`apps/app/`)
**Purpose**: Main application for event management per organisation. Also serves as the platform admin interface for `super_admin` users via `/platform/*` routes.
**Users**: Organisation Admins, Event Managers, Staff Coordinators, Artist Managers, Volunteer Coordinators, Super Admins (platform management via `/platform/*`).
**Features**:
- Event lifecycle management (Draft through Closed)
| Volunteer / Crew | Login (`auth:sanctum`) | Long-term relationship, festival passport, shift history |
| Artist / Tour Manager | Token (`portal.token` middleware) | Per-event, advance portal via signed URL |
| Supplier / Partner | Token (`portal.token` middleware) | Per-event, production request via token |
| Press / Media | Token (`portal.token` middleware) | Per-event accreditation, no recurring relationship |
**Router guard logic**: If `route.query.token` -> token mode. If `authStore.isAuthenticated` -> login mode. Otherwise -> redirect to `/login`.
**Vuexy Version**: `typescript-version/starter-kit` (stripped: no sidebar, no customizer, no dark mode toggle; uses Vuetify components + Vuexy SCSS)
---
## Multi-Tenant Data Model
Shared database schema with organisation scoping on all tables. No row-level security at DB level; scoping enforced via Laravel Policies and Eloquent Global Scopes.
**Scoping Rule**: EVERY query on event data MUST have an `organisation_id` scope via `OrganisationScope` Global Scope.
**Tenancy Hierarchy**:
```
Platform (Super Admin)
└─ Organisation (client A)
└─ Event (event 1)
└─ Festival Section (Bar, Hospitality, Technical, ...)
├─ Time Slots (DAY1-EARLY-CREW, DAY1-EARLY-VOLUNTEER, ...)
└─ Shifts (Bar x DAY1-EARLY-VOLUNTEER, 5 slots)
```
---
## Three-Level Role & Permission Model
Managed via Spatie `laravel-permission` with team-based permissions.
**JSON Columns**: ONLY for opaque config (blocks, fields, settings, items). NEVER for dates, status values, foreign keys, booleans, or anything filtered/sorted/aggregated.
### 1. Foundation
| Table | Key Columns | Notes |
|-------|-------------|-------|
| `users` | id (ULID), name, email, password, timezone, locale, avatar, deleted_at | Platform-wide, unique per email. |
Production (registered domain **crewli.app**): API `https://api.crewli.app` (`APP_URL`); SPAs `https://crewli.app`, `https://portal.crewli.app` via the same env keys. Frontends use `VITE_API_URL=https://api.crewli.app/api/v1`. `SANCTUM_STATEFUL_DOMAINS` = comma-separated SPA hostnames only (e.g. `crewli.app,portal.crewli.app`). **`crewli.nl`** is reserved for a future marketing site only — not used for this application stack.
Crewli is a multi-tenant SaaS platform for professional event and festival management. It supports the full operational cycle: artist booking and advancing, staff planning and volunteer management, accreditation, briefings, and real-time show-day operations (Mission Control). Built for a professional volunteer organisation, with SaaS expansion potential.
import type { AxiosInstance, InternalAxiosRequestConfig } from 'axios'
import { useAuthStore } from '@/stores/useAuthStore'
For Vuexy component selection, consult `dev-docs/VUEXY_COMPONENTS.md` — the registry of @core wrappers and patterns. Always check that registry before writing a custom component.
- Components never import axios directly — always via composables
- Composables call axios via the singleton in `apps/app/src/lib/axios.ts`
- Mutations invalidate query keys after success
- No prop drilling — use Pinia stores when state crosses two component boundaries
### TypeScript Types
### UI
- Three states for every list view: **loading** (VSkeletonLoader), **error** (VAlert with retry button), **empty** (helpful message explaining what action to take)
- Custom CSS forbidden when a Vuetify utility class exists
- Tables on mobile (<768px) collapse to VList or card view — never horizontal scroll without a visual indicator
Production example (subdomains on **crewli.app**): `FRONTEND_APP_URL=https://crewli.app`, `FRONTEND_PORTAL_URL=https://portal.crewli.app`, and `SANCTUM_STATEFUL_DOMAINS=crewli.app,portal.crewli.app`.
Production example (registered domain **crewli.app**): `FRONTEND_APP_URL=https://crewli.app` and `SANCTUM_STATEFUL_DOMAINS=crewli.app`. The legacy `FRONTEND_PORTAL_URL` env key is retained for outbound-email controllers (per AUTH_ARCHITECTURE.md §11), but resolves to the same host post-WS-3.
- Global scope `OrganisationScope` on event-related models
- Policies for authorization — never inline role checks
## Vuexy reference path (mandatory)
- When looking at Vuexy examples, demos, or patterns, ALWAYS reference:
## UI framework strategy (migration-aware)
- Always read `dev-docs/PRIMEVUE_COMPONENTS.md` first — it's the authoritative reference for component selection, theming, forms, and DataTable conventions
- On migrated / new surfaces: PrimeVue + Tailwind v4; forms via `@primevue/forms` + Zod resolver through `<FormField>` (RFC-WS-FRONTEND-PRIMEVUE Appendix A)
- On un-migrated surfaces (during F4): match surrounding Vuetify code; pre-F2 Vuexy reference recoverable via `git show 1c449ff6204cae6371da08c34ea8934d6b2ffcb8:dev-docs/VUEXY_COMPONENTS.md`
- Never mix PrimeVue and Vuetify inside one form or one surface ("no back-porting")
## Vuexy reference path (legacy surfaces only)
- When extending an un-migrated surface, the Vuexy template reference is at:
- Login-based (`auth:sanctum`): organizers, volunteers, crew, super_admin. Includes **Platform Admin** section (`/platform/*`) for super_admin users (organisation management, user management, impersonation, activity log). Context-aware routing inside the SPA distinguishes organizer vs. volunteer experience based on `useAuthStore.availableContexts` (see `dev-docs/AUTH_ARCHITECTURE.md`).
- Token-based (`portal.token` middleware): artists, suppliers, press — persons without `user_id`. Stateless per-request token via `Authorization: Bearer` header or `?token=` query parameter.
### CORS
Configure two frontend origins in both Laravel (`config/cors.php` via env) and the Vite dev server proxy:
Single frontend origin in both Laravel (`config/cors.php` via env) and the Vite dev server proxy:
-app: `localhost:5174`
- portal: `localhost:5175`
-dev: `localhost:5174`
- prod: `https://crewli.app`
**Production (`crewli.app`):** API `https://api.crewli.app`, SPAs `https://crewli.app`, `https://portal.crewli.app` — see `api/.env.example` for `FRONTEND_*` and `SANCTUM_STATEFUL_DOMAINS`. **`crewli.nl`** is only for a future marketing site; this application stack uses **`crewli.app`** (not `.nl` for API, SPAs, or transactional mail).
See `api/.env.example` for `FRONTEND_*` and `SANCTUM_STATEFUL_DOMAINS`. **`crewli.nl`** is only for a future marketing site; this application stack uses **`crewli.app`** (not `.nl` for API, SPA, or transactional mail).
## Backend rules (strict)
@@ -160,56 +160,53 @@ PR as the migration.
- Use factories for all test data
- After each module: `php artisan test --filter=ModuleName`
Frontend test architecture (5 tiers: Unit / Component / Integration /
Visual / E2E) is documented in `dev-docs/ARCH-TESTING.md`. Choose the
right tier per the decision tree there before adding new tests.
## Frontend rules (strict)
### Vuexy reference source (mandatory)
### UI framework strategy (migration-aware)
When referencing Vuexy demo pages, components, or patterns, ALWAYS use the TypeScript Vue version at:
The SPA is migrating Vuetify/Vuexy → PrimeVue + Tailwind v4 per
PrimeVue is the framework. Follow [`PRIMEVUE_COMPONENTS.md`](./dev-docs/PRIMEVUE_COMPONENTS.md):
### Vuexy-first strategy
1.**Can a Tailwind utility do this?** (layout, spacing, typography) → use it.
2.**Does PrimeVue provide a component?** → use it (see §3 component mapping).
3.**Forms** → `@primevue/forms` + Zod resolver via `<FormField>` (§5; full API in [RFC Appendix A](./dev-docs/RFC-WS-FRONTEND-PRIMEVUE.md#appendix-a--formfield-api-specification)).
4.**DataTables** → `<DataTable>` with `:lazy="true"` for server-side (§6).
5.**None of the above?** → cross-reference https://primevue.org/ for the closest match. Add a note in `PRIMEVUE_COMPONENTS.md` §3 if it's a recurring need.
Before writing ANY frontend component, consult `/dev-docs/VUEXY_COMPONENTS.md` and follow this decision tree:
- Notifications: `v-snackbar` — never custom toast components
- Page layout: `v-row` + `v-col` with Vuetify breakpoint props — never CSS grid or custom flexbox
Vuetify + Vuexy `@core/` components remain in use until the surface's F4
sub-package lands. When extending these surfaces during the transition:
**Before ANY frontend task:** read `/dev-docs/VUEXY_COMPONENTS.md` to verify
you are using available components rather than building custom ones.
- Match the surrounding code (`<VBtn>`, `<VTextField>`, `<v-data-table-server>`, etc.)
- Reference the pre-F2 Vuexy registry via git: `git show 1c449ff6204cae6371da08c34ea8934d6b2ffcb8:dev-docs/VUEXY_COMPONENTS.md`
- Vuexy template reference (when needed): `resources/vuexy-admin-v10.11.1/vue-version/typescript-version/full-version/` — TypeScript Vue version is the only valid path
Do **not** introduce PrimeVue components inside an un-migrated surface
("no back-porting" — see `PRIMEVUE_COMPONENTS.md` §9).
#### On new surfaces (created during or after F4)
Start in PrimeVue. The migration phase is not a license to add new
Vuetify code.
### Vue components
@@ -236,10 +233,36 @@ you are using available components rather than building custom ones.
### Forms
- VeeValidate for form state + Zod for schema validation — always together
- Zod schemas must mirror the backend Form Request rules (field names, required/optional, types)
The canonical form pattern depends on the migration phase of the surface:
**Target state (migrated surfaces, new surfaces):** `@primevue/forms` +
Zod resolver via the `<FormField>` wrapper. Full API specification in
@@ -250,12 +273,12 @@ you are using available components rather than building custom ones.
### UI
- Always use Vuexy/Vuetify for layout, forms, tables, dialogs
- Do not write custom CSS when a Vuetify utility class exists
- Component framework selection: see "UI framework strategy" above and [`PRIMEVUE_COMPONENTS.md`](./dev-docs/PRIMEVUE_COMPONENTS.md). PrimeVue + Tailwind v4 on migrated/new surfaces; Vuetify on un-migrated surfaces during F4
- Do not write custom CSS when a framework utility (Tailwind on migrated surfaces, Vuetify utilities on legacy surfaces) exists
- Responsive: mobile-first, usable from 375px width
- **Three states per page:** every data-driven view must handle loading (skeleton/spinner), error (`v-alert` with retry button), and empty (helpful message with action button)
- Use Vuetify responsive props (`cols`, `sm`, `md`, `lg`) — no fixed pixel widths
- Custom CSS via `<style scoped>` only as last resort when no Vuetify utility exists
- **Three states per page:** every data-driven view must handle loading (skeleton), error (`Message` / `v-alert` with retry button), and empty (helpful message with action button) — both frameworks support this pattern
- Responsive layout: Tailwind grid (`grid grid-cols-12 gap-4` + `col-span-N md:col-span-M`) on migrated surfaces; Vuetify `v-row` + `v-col` with breakpoint props on legacy surfaces — no fixed pixel widths
- Custom CSS via `<style scoped>` only as last resort when no framework utility / `pt` API / Aura token can do the job
## Forbidden patterns
@@ -303,6 +326,38 @@ allowed only with a `TODO TECH-*` reference to a backlog item.
13. Vue page component in `src/pages/[module]/`
14. Add route in Vue Router
> **Framework note for steps 13–14 during F4 migration:** new pages
> follow the PrimeVue + Tailwind conventions in [`PRIMEVUE_COMPONENTS.md`](./dev-docs/PRIMEVUE_COMPONENTS.md).
> If the new module is grafted onto a not-yet-migrated surface (rare),
> match the surrounding Vuetify style and let the surface's F4
> sub-package migrate it later.
## Diagnostic discipline: audit before assume
When debugging or fixing any bug, the first action is to verify the
canonical model against the artifact in question — not to write a fix
based on the symptom or hypothesis.
This applies to:
- Schema drift (verify the resource shape against the Zod schema, line
by line)
- Filter logic (verify the data model in SCHEMA.md before assuming a
controller is wrong)
- UX divergence (verify the prototype line-by-line, not via spot check)
- Test failures (verify the assertion matches the documented contract,
not the current implementation)
Phase A of every fix prompt is STOP-and-report. No code is written
before the audit is reviewed.
This principle was formalised after three consecutive incidents where
initial hypotheses were wrong and the audit gate caught them: B1
(controller assumed buggy, seeder was wrong), B5 (enum-shape assumed
drifted, decimals were wrong), and timetable UX (test-passing layer
diverged from prototype, mechanical-vs-UX split surfaced via browser
test).
## User Documentation (VitePress)
End-user documentation lives in /docs/ as a VitePress site.
@@ -348,7 +403,8 @@ Rules:
3. One commit per logical unit of work (one feature, one bugfix, one refactor)
4. Never bundle unrelated changes in a single commit
5. Never commit with failing tests
6. Do NOT push automatically — only commit locally. The developer will push manually.
6. Push, open PRs, and merge are AUTHORISED for Claude (granted by Bert
2026-06-03), gated on a green merge gate — see "Push & Merge Authority".
Commit message format:
```
@@ -369,3 +425,26 @@ Examples:
- `feat: person tags system with org-level skills and sync endpoint`
- `fix: auth race condition on page refresh`
- `docs: update SCHEMA.md with person_identity_matches table`
### Push & Merge Authority
Claude may push feature branches, open Gitea PRs, and merge them **without
a separate human approval click**, provided ALL of the following hold:
1. **Green merge gate** — crewli-reviewer `REVIEW VERDICT: PASS` (no MUST
FIX), all applicable tests passing, lint + typecheck clean, and for
backend changes Larastan clean + the multi-tenancy 403 test present.
The gate — not a human click — is the safety mechanism. If any signal
is red, Claude does NOT merge; it returns the work to the implementer.
2. **Never push directly to `main`.** Integrate only via a `--no-ff` merge
of a reviewed feature branch (a merged Gitea PR with `merge_style: merge`,
or a local `--no-ff` merge then push).
3. **Never force-push** to `main` or any shared branch.
4. **Post-merge:** verify the merge landed on `main`, then delete the
feature branch locally and remotely (per the rule above).
5. Still **present the merge gate** for visibility before self-merging, but
Claude may proceed to merge once the gate is green rather than blocking
on an explicit `merge` reply.
This supersedes the earlier "developer pushes manually" rule and the
build-module skill's HUMAN GATE 2 (CLAUDE.md takes precedence over skills).
| **SPA** | `apps/app/` | 5174 | Single-SPA product covering**organizers, volunteers, crew, super admins** (context-routed in-app), plus token-based access for artists, suppliers, press. Includes **Platform Admin** section for super admins (`/platform/*`). |
All apps talk to the API over **CORS** with **Laravel Sanctum** tokens.
The SPA talks to the API over **CORS** with **Laravel Sanctum** tokens.
---
@@ -125,13 +124,12 @@ make db-shell
| Resource | Contents |
|----------|----------|
| [resources/design/](resources/design/) | **Canonical product specs** in Markdown. Referenced by `.cursor` and `CLAUDE.md` as source of truth for features and data model: `design-document.md`, `dev-guide.md`, `start-guide.md`. |
| [.cursor/ARCHITECTURE.md](.cursor/ARCHITECTURE.md) | System diagram, apps, multi-tenancy, roles, event lifecycle, API route map, core schema overview (summarises `resources/design` when present) |
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.