* PBI-50 F1: SPRINT_BATCH execution-strategy + cross-repo blocker + branch-resume
Schema-migratie + Scrum4Me-side wiring voor de nieuwe SPRINT_IMPLEMENTATION-flow:
- prisma: PrStrategy ADD VALUE 'SPRINT_BATCH'; ClaudeJobKind ADD VALUE
'SPRINT_IMPLEMENTATION'; nieuwe enum SprintTaskExecutionStatus; ClaudeJob.lease_until
+ status_lease_until index; SprintRun.previous_run_id (self-relation
SprintRunChain) voor branch-hergebruik bij resume; nieuwe sprint_task_executions
tabel met frozen plan_snapshot + verify_required_snapshot per task in scope.
- actions/sprint-runs.ts startSprintRunCore: nieuwe blocker-type 'task_cross_repo'
voor SPRINT_BATCH (pre-flight rejecteert sprints met cross-repo task_url).
Bij SPRINT_BATCH: één SPRINT_IMPLEMENTATION ClaudeJob (geen per-task loop).
- actions/sprint-runs.ts resumePausedSprintRunAction: SPRINT_BATCH-pad met
remaining-execution-check; bij onafgemaakt werk → nieuwe SprintRun met
previous_run_id + run.branch hergebruikt + nieuwe SPRINT_IMPLEMENTATION-job.
Oude SprintRun → CANCELLED. Bestaande PBI-49 P0 scope-DONE pad ongewijzigd.
- actions/products.ts updatePrStrategyAction: accepteert SPRINT_BATCH.
- components/products/pr-strategy-select.tsx: drie opties met helptekst,
gebruikt @prisma/client PrStrategy ipv lokaal type.
- components/sprint/sprint-run-controls.tsx: BLOCKER_LABELS + blockerHref
voor task_cross_repo.
Migratie applied op Neon. Type-check + 532 tests groen.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* PBI-50 F5: cross-repo blocker test voor SPRINT_BATCH
- task_cross_repo blocker fires bij task.repo_url ≠ product.repo_url
- happy path: tasks zonder repo_url-override of met match → één
SPRINT_IMPLEMENTATION-job (niet per-task).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* PBI-50 F5: docs/architecture/sprint-execution-modes.md
Vergelijking PER_TASK vs SPRINT_BATCH met trade-offs, datamodel-
toevoegingen (SprintTaskExecution, lease_until, SprintRunChain) en
MCP-tools-matrix per modus. Toegevoegd aan breadcrumb in
docs/architecture.md.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
A STORY-mode MERGE_CONFLICT triggers AFTER all tasks are already DONE
(storyBecameDone is what produces the conflict event). The previous resume
logic put the SprintRun back to QUEUED, but there was no QUEUED job left for
a worker to claim — the run would hang forever.
Now: when both QUEUED and CLAIMED/RUNNING counts are zero, transition straight
to DONE (with finished_at set). The dev resolved the conflict manually and
the PR is theirs to merge. Existing behaviour preserved when active claims
or queued work remain.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Scrum4Me-side counterpart of scrum4me-mcp@f7f5a48 (PBI-9 + PBI-47):
- prisma migration: ClaudeJob.{base_sha,head_sha} + SprintRun.pause_context
- lib/pause-context.ts: Zod schema + parsePauseContext + pauseReasonLabel
helper; single source of truth for the JSON pause_context shape produced
by the mcp sprint-run flow (MERGE_CONFLICT pause)
- actions/sprint-runs.ts: resumePausedSprintRunAction — separate from the
existing FAILED-resume flow, requires SprintRun.status === PAUSED, closes
the linked ClaudeQuestion, clears pause_context, sets RUNNING/QUEUED based
on whether a claim is still active
- components/sprint/sprint-run-controls.tsx: PAUSED banner with reason label,
PR link, conflict-files list (max 5 + "+N more"), Resume button with
confirm() guard
- app/(app)/products/[id]/sprint/page.tsx: load pause_context from active
SprintRun and pass through to SprintRunControls
All MD3 tokens (warning-container, on-warning-container, primary). No raw
Tailwind utility colours.
Tests: 532 passing across 72 files (Scrum4Me side).
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* ST-1243: F1 schema + propagateStatusUpwards-helper voor sprint-flow
Schema-uitbreidingen voor de sprint-niveau jobflow (PBI-46):
- TaskStatus, StoryStatus, PbiStatus, SprintStatus krijgen FAILED
- Nieuwe enums: SprintRunStatus, PrStrategy
- Nieuw SprintRun-model dat per-task ClaudeJobs groepeert
- ClaudeJob.sprint_run_id koppeling + index
- Product.pr_strategy (default SPRINT)
- Bijhorende Prisma-migratie
propagateStatusUpwards vervangt updateTaskStatusWithStoryPromotion en
herevalueert de keten Task → Story → PBI → Sprint → SprintRun bij elke
task-statuswijziging. Bij FAILED cancelt het sibling-jobs in dezelfde
SprintRun. PBI-status BLOCKED blijft handmatig en wordt niet overschreven.
Status-mappers + theme krijgen failed-token + label-uitbreidingen.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* ST-1244: F2 sprint-runs actions + deprecate per-task enqueues
actions/sprint-runs.ts (nieuw):
- startSprintRunAction met pre-flight (impl_plan / open ClaudeQuestion / PBI BLOCKED|FAILED)
- Maakt SprintRun + ClaudeJobs in PBI→Story→Task volgorde
- resumeSprintAction zet FAILED tasks/stories/PBIs terug en start nieuwe SprintRun
- cancelSprintRunAction breekt lopende SprintRun af zonder cascade
actions/claude-jobs.ts:
- enqueueClaudeJobAction, enqueueAllTodoJobsAction, previewEnqueueAllAction,
enqueueClaudeJobsBatchAction nu deprecation-stubs (UI-cleanup volgt in F4)
- cancelClaudeJobAction blijft beschikbaar voor losse jobs
Tests bijgewerkt: 11 nieuwe sprint-runs tests, claude-jobs(-batch) tests
herzien naar deprecation-asserties.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* ST-1246: F4 UI Start/Resume/Cancel sprint + pr_strategy dropdown
- components/sprint/sprint-run-controls.tsx: knoppen Start Sprint
(sprintStatus=ACTIVE), Hervat sprint (sprintStatus=FAILED) en
Annuleer sprint-run (lopende run). Pre-flight blocker-modal toont
blockers met directe links naar de relevante pagina's.
- components/products/pr-strategy-select.tsx: dropdown SPRINT|STORY in
product-settings, met optimistic update + sonner-toast op fail.
- actions/products.ts: updatePrStrategyAction (eigenaar-only, demo-block).
- Sprint-page: query op actieve SprintRun + tonen van controls-balk.
Live cascade-visualisatie (T-634) staat als follow-up genoteerd —
huidige sprint-board statusbadges volstaan voor MVP. De Solo-board
"Voer uit"-knoppen zijn niet expliciet verwijderd; ze tonen nu de
deprecation-error van de gestubde actions tot de Solo-flow opnieuw
ontworpen 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>
