Twee preventieve aanpassingen na de mislukte ST-2wj8mw8q-run, waarbij een
schema-edit (Todo-model verwijderd) groen door lint en vitest kwam maar
op `next build` brak vanwege 16 ongelinkte `prisma.todo`-references in 4
bestanden.
package.json:
- "typecheck": "tsc --noEmit" — losse script voor snelle full-project
type-check, los van eslint en next build.
- "verify": "npm run lint && npm run typecheck && npm test" — umbrella
voor agents/CI om voor de eind-build te valideren.
lib/idea-prompts/make-plan.md:
- Werkwijze-stap 3 toegevoegd: "Bij removal/refactor: doe een
dependency-cascade-grep". Voor de strikte format-sectie staat nu een
verplicht protocol:
- Removal van Prisma-model: grep `prisma.x` in actions/app/components/lib
- Removal van component/utility/type: grep paden + exports
- Hernoemen: per geraakt bestand een edit-taak
- Veld-wijziging in create/update: grep op die call-sites
- Eind-taak: `npm run typecheck` als sanity-check los van
lint/test/build.
- Toelichting waarom: eslint en vitest+esbuild slaan diepe type-check
over; next build is de eerste step die alles type-checkt en zit aan
het eind van de pijp.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Drie functies via prisma.$queryRaw: getSprintTokenHistory (per-sprint
aggregaat), getDayTokenData (dag-totalen met guard op lege sprintId),
getPbiTokenAggregates (per-PBI met guard). Tests voor alle drie.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-vmc7vpps): lib/insights/token-stats.ts — sprint KPI + per-job query
SQL-queries voor totale tokens/kosten (KPI) en per-job tabel met
ModelPrice JOIN. Guard op lege sprintId. Tests voor empty guard,
KPI-mapping en null token-data.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-vmc7vpps): TokenUsageCard — KPI-kaartjes + sorteerbare per-job tabel
Client-component met drie KPI-strips (totaal tokens, kosten USD, gem. per job)
en sorteerbare tabel op kosten of duur. Nulls als '—', MD3-tokens, geen
hardcoded kleuren.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-vmc7vpps): insights page — TokenUsageCard integreren
Voeg getTokenStats + TokenUsageCard imports toe aan insights/page.tsx.
tokenStats apart awaiten na activeSprints (kan niet in dezelfde Promise.all).
TokenUsageCard-sectie toegevoegd na AgentThroughputCard.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Voegt IdeaProduct schema toe (dependency van story-qtkvz6ly), breidt
IdeaWithProduct type en IdeaDto interface uit met secondary_products array,
en laadt de relatie mee in findMany/findFirst in page.tsx en REST GET.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- lib/job-status.ts: SKIPPED ↔ 'skipped' mapping in beide richtingen
- components/shared/job-status.ts: label "Overgeslagen" + neutrale italic styling
- actions/admin/jobs.ts: cancel-guard erkent SKIPPED als eindstatus
- app/api/cron/cleanup-agent-artifacts: SKIPPED ook opruimen na 7d
- lib/insights/agent-throughput: SKIPPED telt mee als terminal
ACTIVE_JOB_STATUSES bewust ongewijzigd — SKIPPED is afgerond.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Nav-label 'Ideeën' hernoemd naar 'Ideas'; breadcrumb idem
- Grill/Plan tabs disabled (grijs, cursor-not-allowed) zolang er geen
content is; groene stip zodra grill_md resp. plan_md beschikbaar is
- SSE hook roept router.refresh() aan bij job done/failed zodat de
Timeline automatisch de nieuwe GRILL_RESULT/PLAN_RESULT logs toont
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- requireAdmin() checkt nu de database i.p.v. session.isAdmin (was altijd undefined)
- loginAction stelt session.isAdmin in op basis van UserRole in de DB
- registerAction stelt session.isAdmin = false expliciet in
- NavBar toont 'Admin'-link conditioneel als roles.includes('ADMIN')
- UserMenu ROLE_LABELS uitgebreid met ADMIN → 'Admin'
- Tests aangepast: prismaUserRole.findFirst mock toegevoegd
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Nieuwe server action in actions/user-questions.ts
- Aanmaken UserQuestion + ClaudeJob PLAN_CHAT in transactie
- Blokkeert als idea.plan_md null is of product ontbreekt
- Idempotency-check: geen dubbele PLAN_CHAT per idee
- pg_notify claude_job_enqueued event voor SSE-realtime
- Rate-limit config uitgebreid met create-user-question
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The /ideas/[id]?tab=timeline page is server-rendered: questions are a
prop snapshot. Without a router.refresh, new questions only show after a
manual page reload — and during a grill-session that's every ~20s.
Fix: in use-notifications-realtime, after dispatching idea-question
events to idea-store + re-syncing the bell, call \`router.refresh()\`.
This re-runs the server-component for whichever page the user is on; on
/ideas/[id] it pulls the new question. On other pages it's a no-op (no
visible state change).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two gaps discovered during the first live grill-session of IDEA-002:
the agent posted a question, but the user had no UI to answer it.
1. Idea-questions only appeared on the Timeline-tab as read-only entries
2. Notifications-bell fetched + handled story-questions only
This fix:
**Inline answer-form in IdeaTimeline** (components/ideas/idea-timeline.tsx)
- Open questions now render an AnswerForm directly under the question text
- Multi-choice options become clickable buttons (one-click submit); free-text
fallback via collapsed details/textarea
- Plain free-text questions render textarea + Verzend
- Calls existing answerQuestion server-action; toast + router.refresh on success
**Notifications-bell extended for idea-questions**
- stores/notifications-store.ts: NotificationQuestion → discriminated union
(kind: 'story' | 'idea'); forYouCount treats idea-questions as always-for-you
(idea is strictly user_id-only — only the owner sees them)
- components/notifications/notifications-bridge.tsx: parallel fetch of
story-questions (productAccessFilter) + idea-questions (idea.user_id ===
session.userId); merged + sorted by created_at
- components/notifications/notifications-sheet.tsx: renders idea_code/title
for kind='idea'
- components/notifications/answer-modal.tsx: header + open-link branch on
kind (idea → /ideas/[id]?tab=timeline; story → existing /sprint link)
- lib/realtime/use-notifications-realtime.ts: idea-question events also
trigger close+reconnect on 'open' (loads fresh detail) and remove(id) on
non-open — same pattern story-questions already use
- components/shared/notifications-bell.tsx: badge counts idea-questions as
for-you regardless of assignee
**Security gap closed (actions/questions.ts answerQuestion)**
Before: accepted any answer if user has product-access.
After: idea-questions require idea.user_id === session.userId; story-
questions keep the existing productAccessFilter path. (Prisma 7 rejects
\`{ not: null }\` in WHERE; routing happens app-level after a single fetch.)
Tests: 546/546 still green.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- app/api/ideas/route.ts: GET (list with archived/product_id/status filters,
user_id-scope), POST (creates DRAFT with auto IDEA-NNN code, 201)
- app/api/ideas/[id]/route.ts: GET (idea + recent logs), PATCH
(ideaUpdateSchema, isIdeaEditable guard)
- lib/idea-dto.ts: API projection — converts Prisma row → DTO with
lowercase status + has_grill_md/has_plan_md flags (md content excluded
from list payloads, fetch via dedicated download action)
Auth: session OR API-token via authenticateApiRequest. Strict user_id
scope (no productAccessFilter — Idee is privé per Q8). 404 (not 403) for
foreign-user reads to prevent enumeration.
Tests: 13 cases (auth-401, demo-403, validation-422, malformed-400,
not-found-404, status-mismatch-422, filter param round-trip, DTO shape).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
PrismaClientValidationError ('Argument \`not\` must not be null') hit at
runtime when notifications-bridge mounted post-M12 schema change.
Although StringNullableFilter typings allow \`not: null\`, the v7 query
engine rejects it.
Removed the WHERE-side filter in 3 places — null-narrowing already
happens client-side via flatMap / Boolean filter:
- components/notifications/notifications-bridge.tsx
- app/api/realtime/notifications/route.ts
- lib/insights/verify-stats.ts (task_id filter)
Idea-questions / idea-jobs will be routed via separate channels in
T-502 + T-507; for now, story-question + task-job paths simply ignore
NULL rows in their post-fetch mapping.
Tests: 479/479 green; tsc clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- new enums IdeaStatus, ClaudeJobKind, IdeaLogType
- new models Idea (with @@unique([user_id, code]) + pbi_id @unique) and IdeaLog
- User.idea_code_counter Int @default(0) for IDEA-{nnn} code generation
- ClaudeJob.task_id nullable; new idea_id + kind fields + index
- ClaudeQuestion.story_id nullable; new idea_id field + index
- existing call sites narrowed to story-questions / task-jobs (idea-paths come in T-502+)
- includes the M12 plan doc copied from /Users/janpetervisser/.claude/plans
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Story 5 van PBI "Alle dialogen conform docs/patterns/dialog.md".
- lib/schemas/sprint.ts — gedeelde zod-schemas (create/dates/goal)
- actions/sprints.ts — code+fieldErrors voor 422; code: 403 voor
auth/demo errors
- StartSprintButton dialog: useDirtyCloseGuard, useDialogSubmitShortcut,
entityDialog* layout-classes; DemoTooltip op trigger; veld-niveau
errors via fieldErrors
- SprintHeader's date- en complete-dialogen: zelfde behandeling; date-
dialog krijgt dirty-guard, complete-dialog krijgt DemoTooltip op
bevestigen
- docs/specs/dialogs/sprint.md — entity-profile dat alle drie de modes
documenteert; consolidatie naar één SprintDialog component bewust
uitgesteld
- Sprint-dates tests aangepast aan nieuwe action-shape
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Vercel build was failing with TS18047 — `sprint.completed_at` is
possibly `null`. The earlier `.filter(s => s.completed_at != null)`
runtime-filtered the nulls out but did NOT narrow the element type;
TypeScript still saw `Date | null` on the result.
Add a user-defined type guard `(s): s is SprintWithCompletedAt =>` so
the narrowed array carries `completed_at: Date`. No runtime change.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(insights): add getJobsPerDay helper — agent throughput per day + KPIs
Raw SQL aggregation of claude_jobs by day and status over 14 days with
zero-fill for missing days. KPIs: todayCount, successRate7d, avgDurationSeconds7d.
Optional productId filter.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(insights): add AgentThroughputCard — stacked BarChart + KPI-strip + product filter
KPI strip (jobs today, 7d success rate, 7d avg duration), 14-day stacked
BarChart with JOB_STATUS_COLORS, and URL-bookmarkable product dropdown via
useTransition + router.replace. Empty-state when no activity.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Three read-only counters: stories without acceptance_criteria, tasks without
implementation_plan, and top-10 IN_PROGRESS tasks stuck >7 days. All scoped
via productAccessFilter.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Aggregates task.status=DONE counts across last N completed sprints
(default 5), filtered by productAccessFilter and returned in
chronological order for x-axis rendering.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Three SSE-routes (solo, backlog, notifications) each create a long-
running pg.Client that LISTENs on scrum4me_changes. On abrupt close
(Fast Refresh, browser refresh, Vercel function recycle) the
pgClient.end()-await sometimes hangs silently, leaving the underlying
socket connected to Postgres. The connection stays in 'idle' on Neon's
side and after ~10-20 reconnects the connection-pool fills up — new
SSE connects fail with ERR_INCOMPLETE_CHUNKED_ENCODING in the browser.
Fix: shared `closePgClientSafely` helper that races client.end()
against a 2 s timeout; on timeout it force-destroys the underlying
socket so the OS releases the FD and Postgres notices the disconnect.
Validated by direct DB inspection: 18 stale 'idle LISTEN'-connections
were piled up before the fix; after manual pg_terminate_backend cleanup
the SSE-stream stabilised. This change makes the pile-up impossible
going forward.
- new lib/realtime/pg-client-cleanup.ts
- 3 routes use the helper instead of bare `await pgClient.end()`
- 3 unit tests for the helper (timely-end, hang-falls-back-to-destroy,
end-rejection-is-swallowed)
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Re-introduce the 3 unique files from closed PRs #37 and #40 that
overlap-merged with already-landed sub-PRs (#34, #35, #36, #38, #39):
- app/(app)/insights/page.tsx — Server Component dat alle helpers
parallel aanroept en de 5 sectie-Cards rendert (Sprint Health,
Plan-quality, Agent throughput, Velocity, Backlog health)
- app/(app)/insights/components/sprint-info-strip.tsx — chips per
active sprint met productname + goal + dagen-over + taakcount
- app/(app)/insights/components/alignment-trend.tsx — Recharts
LineChart die % ALIGNED jobs per sprint over laatste 5 sprints toont
- lib/insights/verify-stats.ts — TrendPoint type + getAlignmentTrend
helper (uitgebreid van PR #38)
Plus dependency: recharts (was in package.json van #37/#40 die we
sloten).
Tests: 290/290 groen, tsc clean, lint clean.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Aggregeert verify_result counts (ALIGNED/PARTIAL/EMPTY/DIVERGENT) en top-5 EMPTY/DIVERGENT
jobs over de laatste N dagen voor de ingelogde gebruiker.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* chore: voeg recharts toe aan dependencies
Vereist door SprintStatusDonut component.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat: SprintStatusDonut + getSprintStatusBreakdown helper
PieChart (donut) met TO_DO/IN_PROGRESS/DONE verdeling over alle active sprints.
REVIEW wordt samengevoegd in IN_PROGRESS. MD3 status-kleuren via CSS-variabelen.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Server-side aggregatie per active sprint: bouwt time-series met remaining en ideal per dag.
Inclusief 4 Vitest-unit-tests voor de pure computeBurndownDays functie.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* chore(ST-1112): add deps for task dialog
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1112): add shared zod schema for task dialog
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1112): add missing MD3 tokens for task dialog
outline-variant, on-error-container, status-review (light + dark)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1112): add saveTask and deleteTask server actions for TaskDialog
Unified create/edit action (saveTask) replaces separate formData-based
actions for the new TaskDialog. Uses shared zod schema, structured
SaveTaskResult union type, and context-aware revalidatePath for both
sprint and backlog routes.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1112): add TaskDialog component (create & edit mode)
Builds the full TaskDialog on top of the existing @base-ui/react
Dialog primitive. Covers create mode, edit mode (status field +
created_at metadata + delete), dirty-check AlertDialog, delete
confirm AlertDialog, Cmd+Enter submit, and per-field char counters.
Uses react-hook-form + zodResolver against the shared taskSchema.
Priority and status are extracted to PrioritySegmented and
StatusSelect sub-components using MD3 tokens throughout.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1112): refactor task-list to open TaskDialog via URL params
Replaces inline create/edit forms with router.push navigation:
- Clicking a task row → ?editTask=<id>
- "+ Taak" button → ?newTask=1&storyId=<storyId>
Removes CreateTaskForm, EditSubmitButton, updateTaskAction, and
createTaskAction from the component. Status toggle and DnD remain
unchanged. Rows now have cursor-pointer and keyboard a11y.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1112): wire TaskDialog into sprint page via searchParams
Sprint page now reads ?newTask, ?storyId, and ?editTask query params.
For edit mode: fetches the task server-side with productAccessFilter
scope (invalid/foreign IDs redirect to closePath). Renders TaskDialog
when either param is present. closePath is the sprint route without
query params.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1112): add Suspense skeleton for edit-mode task loading
Extracts task fetch into EditTaskLoader (async server component) so
the sprint board renders immediately while the task loads.
TaskDialogSkeleton shows 3 grey bars during the fetch. Invalid or
out-of-scope task IDs redirect to closePath.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1112): render description as markdown in task-detail-dialog
Solo task detail now renders description via react-markdown +
remark-gfm with prose styling. Sanitizes script/iframe elements.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* test(ST-1112): add saveTask/deleteTask server action tests
Covers all three demo-policy layers and cross-tenant scope:
demo blocked (403), unauthenticated blocked, validation 422,
edit cross-tenant forbidden, create cross-tenant forbidden,
and happy-path for both edit and create.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1112): add updateTaskStatusWithStoryPromotion helper
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1112): wire story-promotion into saveTask and PATCH /api/tasks/:id
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* docs(ST-1112): add task-dialog doc and architecture note
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* chore: extend allowed tools in settings.local.json
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1113): add 200ms animation-delay to TaskDialogSkeleton to prevent flicker
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1114): add DirtyCloseGuard reusable component for dirty-form close confirmation
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1114): add shared Markdown wrapper, apply to task-detail and story-dialog
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* chore: allow grep -E pattern in settings.local.json
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1111.1): add ClaudeJob model and state-machine enum
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1111.2): add ClaudeJob status API mappers
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1111.3): add enqueue/cancel ClaudeJob server actions with idempotency + NOTIFY
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1111.4): forward ClaudeJob events on solo SSE stream + initial state
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1111.6): add 'Voer uit' + cancel buttons to task detail dialog
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1111.7): add job status pill with spinner on solo task cards
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* test(ST-1111.8): cover job-status mappers and enqueue/cancel actions
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* docs(ST-1111.9): document Claude job queue architecture and agent flow
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1111.10a): add ClaudeWorker presence model
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1111.10c): forward worker presence events on solo SSE + initial count
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1111.10d): show worker presence indicator and gate 'Voer uit' on connected workers
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(ST-1109.2): add PbiStatus enum and status field to Pbi model
- New PbiStatus enum (READY/BLOCKED/DONE) for PBI lifecycle tracking
- Pbi.status PbiStatus @default(READY)
- Index on (product_id, status) for filter queries
- Migration: 20260429150643_add_pbi_status
- ERD regenerated via prisma generate
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(ST-1109.3): add PBI status API mappers
- pbiStatusToApi / pbiStatusFromApi following same pattern as task/story
- PbiStatusApi type derived from PBI_DB_TO_API
- PBI_STATUS_API_VALUES export for downstream Zod schemas
- Lowercase API surface (ready/blocked/done), DB stays UPPER_SNAKE
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(ST-1109.4): support status in PBI create/update actions
- Optional status field in Zod schemas (lowercase API: ready/blocked/done)
- pbiStatusFromApi() maps to DB enum before persistence
- Status omitted on create => Prisma @default(READY) takes effect
- Update preserves existing status when not provided
- Demo-check unchanged
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(ST-1109.5): auto-mark PBI as DONE when all its stories are DONE on sprint close
Extends completeSprintAction's $transaction with PBI status cascade:
- Pre-transaction: identify PBIs touched by this close (via stories.pbi_id),
fetch each with all its stories
- Skip PBIs already DONE; skip PBIs with 0 stories
- Mark PBI DONE only when every story (post-decision) is DONE — stories
outside the sprint are evaluated against their current DB status
- Promote-only: never demotes a PBI that becomes "incomplete" again
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(ST-1109.6): add Popover primitive (base-ui wrapper)
- Mirrors the Tooltip pattern: render-prop composition, data-slot attrs
- Exports Popover (Root), PopoverTrigger, PopoverContent (Portal+Positioner+Popup)
- MD3 popover/popover-foreground tokens, animated open/close states
- Will be used to consolidate the backlog filter UI in ST-1109.8
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(ST-1109.7): add status select to PBI dialog
- New components/shared/pbi-status-select.tsx mirrors PrioritySelect:
PBI_STATUS_LABELS (NL), PBI_STATUS_COLORS, PbiStatusSelect component
- Reuses existing --status-todo/blocked/done MD3 tokens
- PbiDialog: status state with sync-on-open; default 'ready' for create,
pbi.status for edit; hidden input submits lowercase API value
- Priority + Status sit side-by-side in 2-col grid
- PbiDialogPbi.status is optional; pbi-list.tsx will populate in ST-1109.8
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(ST-1109.8): show PBI status badge and consolidate filters into popover
- Pbi.status (lowercase API) flows from page.tsx via pbiStatusToApi
- Status badge rendered in BacklogCard's badge slot using PBI_STATUS_COLORS
- Two old Select dropdowns replaced by single Popover with three pill-button
sections (Sorteren, Prioriteit, Status) and a "Wis filters" footer
- Filter trigger shows active count "(n)" badge in label
- Active priority/status filters still surface as dismissable chips next to
the trigger for at-a-glance feedback
- onEdit passes the full Pbi (incl. status) so the dialog opens with the
correct current status — closes the data flow loop opened in ST-1109.7
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(ST-1109.9): cover PBI status mappers and sprint-close cascade
- __tests__/lib/task-status.test.ts: 11 cases incl. round-trip + invalid
input for task/story/pbi mappers; verifies PBI_STATUS_API_VALUES shape
- __tests__/actions/sprints-cascade.test.ts: 8 cases for completeSprintAction:
promote on all-DONE, no promote on partial OPEN, respect out-of-sprint
story status, skip already-DONE PBIs, multi-PBI cascade, 0-story guard,
demo-user block
- Full vitest run: 170/170 green across 21 files
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* docs(ST-1109.10): document PbiStatus enum, sprint-close cascade, and filter UI
- docs/scrum4me-architecture.md: pbis-table updated with status column +
index; PbiStatus enum + Pbi model in the Prisma schema sample;
cascade-on-sprint-close rule documented inline
- docs/scrum4me-styling.md: short note pointing to PBI_STATUS_LABELS /
PBI_STATUS_COLORS in components/shared/pbi-status-select.tsx so future
components don't ad-hoc-copy the color map
- docs/plans/ST-1109-pbi-status.md: in-repo mirror of the approved plan
(per feedback_plan_location memory) with cascade pseudo-code and
end-to-end verification checklist
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(ST-1109.11): persist backlog filters in localStorage
Filters reset op reload was verwarrend. Nu net als sortMode:
- scrum4me:pbi_filter_priority — 'all' | '1' | '2' | '3' | '4'
- scrum4me:pbi_filter_status — 'all' | 'ready' | 'blocked' | 'done'
useState-init met SSR-guard; ongeldige waarden vallen terug op 'all'.
Wis filters reset alle drie de keys correct (sortMode -> 'priority',
beide filters -> 'all'), waardoor de localStorage-staat consistent wordt.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* docs(ST-1101..1108): add M11 — Claude question-channel milestone to backlog
Plant acht stories ST-1101..ST-1108 voor het persistente vraag-antwoord-kanaal
tussen Claude (MCP) en de actieve gebruiker. Eerste concrete uitwerking van
de AI-driven dev-flow-richting (strategisch besluit "B" uit overleg na M10).
Beveiligingsuitgangspunt: atomic answer via updateMany WHERE status='open',
demo-blok op write-tools, access-check via productAccessFilter in DB-query én
SSE-filter, cron-endpoint via Bearer-secret, geen vraag/antwoord-tekst in logs.
Hergebruikt bestaande scrum4me_changes-channel (uitgebreid met entity:'question')
en het LISTEN/NOTIFY+ReadableStream-pattern uit M8/M10. Nieuw: user-scoped SSE
op /api/realtime/notifications zodat de bell globaal werkt over producten heen.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore(M11): swap demo-active sprint from M10 to M11
M10 is gemerged en afgesloten — M11 wordt de nieuwe demo-actieve milestone
zodat get_claude_context (via MCP) ST-1101 als next-story teruggeeft.
Drie maps in parse-backlog.ts uitgebreid: M11 priority=4, goal omschrijving,
sprint_status='ACTIVE'. M10 → COMPLETED.
Vereist npx prisma db seed na deze commit zodat de live DB de nieuwe
sprint-state weerspiegelt.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* docs(ST-1108): add F-11b — Claude question-channel to functional spec
Voegt feature-omschrijving toe naast bestaande F-11 (Claude Code REST API).
Beschrijft het verloop (Claude → MCP-tool → DB → trigger → SSE → user → answer
→ trigger → Claude polls), acceptatiecriteria (8 items), randgevallen (offline-
Claude, assignee-change, expiry, abuse) en datamodel (claude_questions tabel).
Persona Lars als primair, Dina secundair voor klant-werk.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore(M11): drop parser ACTIVE-flip; sprint goes via UI from now on
Bij M9/M10 hebben we de seed-flip (MILESTONE_SPRINT_STATUS pivot) gebruikt om
nieuwe stories als IN_SPRINT in een verse sprint te krijgen. Dat werkt maar
is fragiel:
- npm run seed wist user-data
- de "sprint" die de seed maakt is geen echte planning-actie
- bij multi-product scenario's breekt het model
Vanaf M11 gebruiken we de bestaande Sprint-creatie-UI van Scrum4Me. Stories
voor M11 worden via scripts/insert-milestone.ts (idempotent insert, geen
seed-reset) aan de DB toegevoegd; de gebruiker maakt zelf een Sprint aan in
/products/[scrum4me]/sprint en sleept ST-1101..1108 ernaartoe.
Parser-map M11 dus terug naar COMPLETED zodat een eventuele re-seed niet meer
een fake sprint aanmaakt voor M11-stories.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(ST-1101): add ClaudeQuestion model + notify_question_change trigger
Schema (prisma/schema.prisma):
- Nieuw model ClaudeQuestion: id (cuid), story_id (FK Cascade), task_id?
(FK SetNull), product_id (FK Cascade — gedenormaliseerd uit story.product_id
voor SSE-filter zonder join), asked_by (FK Restrict — Claude-token-houder),
question (Text), options (Json? — string[] voor multi-choice), status
('open'|'answered'|'cancelled'|'expired'), answer (Text?), answered_by
(FK SetNull), answered_at?, created_at, expires_at
- Indexes: (story_id, status), (product_id, status), (status, expires_at)
- Back-relations: User.asked_questions (ClaudeQuestionAsker),
User.answered_questions (ClaudeQuestionAnswerer), Story.claude_questions,
Task.claude_questions, Product.claude_questions
Migratie (20260427224849_add_claude_questions):
- Prisma-gegenereerde DDL voor claude_questions + indexes + 5 FK's
- Toegevoegde notify_question_change() functie + claude_questions_notify trigger
op AFTER INSERT/UPDATE
- Emit op BESTAANDE scrum4me_changes-channel met entity:'question' (i.t.t. M10
dat eigen scrum4me_pairing-channel kreeg) — solo-route in ST-1104 moet
entity='question' wegfilteren om regressie op solo-board te voorkomen
- Trigger leest story.assignee_id voor "wacht op jou"-emphase in payload
- DELETE niet ondersteund — questions gaan naar answered/cancelled/expired
Verification: Node pg-client roundtrip via DATABASE_URL toonde correcte payloads
bij INSERT (op=I, status=open) en UPDATE (op=U, status=answered) met alle FK-IDs
en assignee_id correct uit story-join.
Volgende stap M11: ST-1102 — vier MCP-tools in scrum4me-mcp-repo
(ask_user_question, get_question_answer, list_open_questions, cancel_question).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(ST-1103): add answerQuestion server action
actions/questions.ts:
- answerQuestion(questionId, answer) — auth + Zod + demo-blok + access-check
via productAccessFilter (anyone met product-membership mag antwoorden,
consistent met Scrum self-organizing — niet alleen story-assignee)
- Atomic prisma.claudeQuestion.updateMany WHERE id + status='open' +
expires_at>now → status='answered'; concurrent dubbele submit: één wint
(count=1), rest count=0 met disambiguatie via second findFirst
- revalidatePath('/', 'layout') refresh't NavBar bell-count voor SSR-paths;
realtime updates voor andere clients gaan via SSE in ST-1104/1105
- Begrijpelijke NL-foutmeldingen voor elk faalpad
Tests __tests__/actions/questions.test.ts (6 cases):
- happy: status update + revalidatePath called
- demo-block: error + geen DB-call + geen revalidate
- geen access: error + geen update
- al-answered: race-error 'is al answered'
- expired: race-error 'is verlopen'
- lege answer: Zod-validatie
Quality gates: lint 0 errors, tsc clean, vitest 145/145 (17 files).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(ST-1104): add user-scoped /api/realtime/notifications + filter solo-route
Twee delen:
1. Solo-route filter (1-regel-fix in app/api/realtime/solo/route.ts):
- NotifyPayload uitgebreid met entity:'question'
- shouldEmit returnt direct false bij entity='question'
Voorkomt dat solo-clients M11 question-events ontvangen (geen lekkage naar
het Solo-bord; geen onnodig netwerk-verkeer; loose coupling tussen features).
2. Nieuwe SSE-route app/api/realtime/notifications/route.ts:
- User-scoped (geen ?product_id=); query alle accessible product-IDs één keer
bij connect via productAccessFilter
- LISTEN scrum4me_changes; filter entity='question' && product_id ∈ accessible
- Initial-state-event NA LISTEN actief (race-fix conform M10 ST-1004):
query open vragen voor deze user's accessible products, stuur als event:state
met summary (id, story_code/title, assignee_id, question, options, expires_at)
- Hergebruikt het pg.Client + ReadableStream + heartbeat 25s + hard-close 240s +
abort-cleanup-pattern uit solo-route
Tests __tests__/api/notifications-stream.test.ts:
- 401 zonder iron-session cookie (en geen DB-call)
- Solo-route filter wordt visueel/E2E gedekt in ST-1108-acceptatie
Quality gates: lint 0 errors, tsc clean, vitest 146/146 (18 files).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(ST-1105): add NavBar bell + sheet + answer-modal + Zustand store + SSE hook
UI-volledig voor de Claude vraag-antwoord-flow (M11). Bel-icon links van avatar
in NavBar; klik opent slide-over rechts met openstaande vragen; klik op een vraag
opent een modal voor antwoord. Story-assignee = current user krijgt visuele
"voor jou"-emphase met primary-container accent en error-color badge-ring.
Bestanden:
- stores/notifications-store.ts — Zustand store met init/upsert/remove +
openCount/forYouCount selectors (vereenvoudigd vs solo-store: geen pendingOps,
geen optimistic-echo-onderdrukking)
- lib/realtime/use-notifications-realtime.ts — EventSource hook met state-
event en message-event handling, exponential-backoff reconnect, Page
Visibility pause-resume
- components/notifications/notifications-bridge.tsx — Server Component die
initial open-questions fetcht via productAccessFilter
- components/notifications/notifications-realtime-mount.tsx — tiny client
island dat de store hydrateert + de hook activeert
- components/notifications/notifications-sheet.tsx — shadcn Sheet met item-
lijst, "voor jou"-accent voor assignee-vragen, lege staat
- components/notifications/answer-modal.tsx — Dialog met options-radio of
free-text Textarea (max 4000), char-counter, demo-blok via Tooltip; bij
succes optimistisch remove + sheet blijft open zodat meerdere vragen
achter elkaar te beantwoorden zijn
- components/shared/notifications-bell.tsx — Bell-icon met badge (count >9 → "9+"),
ring-accent als forYouCount > 0, ARIA-label voor screenreaders
Wiring:
- components/shared/nav-bar.tsx — <NotificationsBell /> rechts naast <UserMenu>
- app/(app)/layout.tsx — <NotificationsBridge /> naast <SoloRealtimeBridge />,
user.id (server-side) als prop
base-ui-aanpassingen: SheetTrigger/TooltipTrigger gebruiken render-prop ipv
asChild (geen Radix).
Quality gates: lint 0 errors, tsc clean, vitest 146/146, npm run build groen.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(ST-1106): add cross-product access-isolation test for notifications SSE
Demo-policy + assignee-emphase zaten al in eerdere stories:
- answerQuestion demo-blok in actions/questions.test.ts (ST-1103)
- AnswerModal demo-tooltip in components/notifications/answer-modal.tsx (ST-1105)
- requireWriteAccess in MCP write-tools (ST-1102)
Deze story voegt expliciet een access-isolation-test toe op de notifications-
SSE-route: productAccessFilter wordt met de echte userId aangeroepen, en
prisma.product.findMany filter't op archived=false + user_id-scope. Dat
garandeert dat een gebruiker geen question-events ontvangt voor producten waar
hij geen membership op heeft.
Story-assignee-emphase blijft visueel-only (NotificationsBell ring-accent +
Sheet primary-container) — toegang werkt product-membership-breed zodat een
team-lid kan invallen als de assignee niet beschikbaar is.
Quality gates: lint 0 errors, tsc clean, vitest 147/147 (was 146).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(ST-1107): add Vercel cron expire-questions (+ M10 pairing cleanup)
POST /api/cron/expire-questions:
- Auth via Authorization: Bearer ${CRON_SECRET} (Vercel injecteert dit
automatisch wanneer de env-var op de project-omgeving staat); 401 als secret
niet matcht of niet is gezet (faal-veilig — geen open endpoint in dev)
- updateMany op claude_questions WHERE status='open' AND expires_at<now →
'expired'
- Bonus: zelfde route ruimt M10 login_pairings op (status='pending' AND
expires_at<now → 'cancelled'). Eén cron-job is goedkoper qua Vercel-budget
en houdt cleanup-strategie centraal — opvolg-actie uit M10 dat geparkeerd was.
Config:
- vercel.json: crons-entry { path: '/api/cron/expire-questions', schedule: '0 */6 * * *' } (4x/dag)
- lib/env.ts: CRON_SECRET als optional in Zod-schema
- .env.example: documentatie + openssl rand-tip
Tests __tests__/api/cron-expire-questions.test.ts (4 cases):
- 401 zonder Authorization-header
- 401 met verkeerde secret
- 401 als CRON_SECRET niet is gezet (faal-veilig)
- 200 met juiste secret: response { expired_questions, expired_pairings, ran_at }
+ beide updateMany WHERE/data correct
Quality gates: lint 0 errors, tsc clean, vitest 151/151.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* docs(ST-1108): document M11 question-channel — API + architecture + pattern
docs/API.md — twee nieuwe secties:
- 'Notifications' met /api/realtime/notifications SSE-endpoint (event-shapes,
filter-rules, voorbeeld)
- 'Cron — Expire questions' met /api/cron/expire-questions (Bearer-auth,
schedule, response-shape, manual curl)
docs/scrum4me-architecture.md — nieuw hoofdstuk 'Vraag-antwoord-kanaal Claude
↔ user' tussen QR-pairing-flow en Projectstructuur:
- Mermaid sequence-diagram (Claude → DB → trigger → SSE → user → answer →
trigger → Claude polls)
- Threat-model-tabel (race, demo-misbruik, cross-product leak, cron-misbruik,
growth, log-leakage)
- Subsectie 'Waarom hergebruik scrum4me_changes-kanaal' met trade-off vs M10's
eigen-kanaal-aanpak
docs/patterns/claude-question-channel.md — herbruikbaar pattern 'Bidirectionele
async-comms tussen MCP-agent en interactieve user' met de vier eindpunten,
vier security-uitgangspunten, channel-strategie-tabel, TTL-richtlijn, en
sjabloon-bestanden per laag (DB / server / client / MCP-tools).
CLAUDE.md — extra rij in Implementatiepatronen-tabel die naar het nieuwe
pattern-doc verwijst.
Acceptatie 6 scenario's:
1. Sync happy path (MCP wait_seconds + UI submit) — handmatig getest tijdens
ST-1105 acceptance-loop met de q-test injection
2. Async happy path — gedekt door get_question_answer-tool in ST-1102 +
list_open_questions
3. Demo-block — actions/questions.test.ts (case 2: demo-user) + AnswerModal
tooltip (visueel)
4. Access-isolation — notifications-stream.test.ts (case 'access-isolation')
5. Expiry — cron-expire-questions.test.ts (case '200 met juiste secret')
6. Race — actions/questions.test.ts (case 'al-answered' via atomic updateMany)
Quality gates: lint 0 errors, tsc clean, vitest 151/151 (19 files), npm run
build groen.
M11 is hiermee feature-compleet. feat/M11-claude-questions heeft 12 commits
lokaal, klaar voor user-acceptatie en PR.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(ST-1107): cron schedule daily — Vercel Hobby allows only 1 run/day
Vercel deploy faalde met:
> Hobby accounts are limited to daily cron jobs.
> This cron expression (0 */6 * * *) would run more than once per day.
Schedule van 4×/dag (0 */6 * * *) naar 1×/dag (0 4 * * * — 04:00 UTC, rustig
tijdstip). Functioneel acceptabel: ClaudeQuestion TTL is 24u, dus daily
cleanup pakt alles dat in de afgelopen 24u verlopen is. Login-pairings TTL
is 2 min — die zijn al onbruikbaar zodra ze expiren, cron is alleen voor
status-housekeeping.
Schedule-referenties consistent bijgewerkt in docs (API.md, architecture,
backlog M11-sectie, plan-doc, pattern-doc) + comment in route.ts. Vermelding
overal dat dit een Hobby-plan-beperking is en Pro fijnmaziger ondersteunt.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
