Promote dev to main: inline AMA operations drain, sparser crons

[CaliCastle]

Release

Promotes #299 to production.

Neon compute fix: the */5 AMA work cron kept the database endpoint awake 24/7 (~183 CU-hrs/month). Durable operations (booking emails, Finalizing Booking recovery, refunds, artifact updates) now drain inline via waitUntil in the same invocation that enqueues them, so guest-facing work starts immediately. The cron drops to */30 (media reconcile to hourly) as a pure clock fallback for reminders, retry backoff, and expired Slot Hold bookkeeping — letting Neon autosuspend between wakeups (~85% compute reduction expected).

Review hardening included: webhook kick gated on booking_created, drain loops past full batches under a 240s budget, mutating routes at the platform-default 300s maxDuration.

Note: the new cron schedules take effect with this production deployment.

Greptile Summary

The PR starts due AMA durable operations within the mutation invocation that enqueues them, while retaining less-frequent cron sweeps for time-based and recovery work.

• Adds waitUntil-backed, multi-batch operation draining with a 240-second start budget.
• Triggers draining after successful booking-management actions and newly created Stripe bookings.
• Raises affected mutation routes to the platform’s 300-second duration limit.
• Changes the AMA worker cadence from five to thirty minutes and media reconciliation from fifteen minutes to hourly.

Confidence Score: 5/5

The PR appears safe to merge, with inline processing covering mutation-triggered AMA work and existing cron paths retaining scheduled recovery duties.

Durable-operation leasing protects concurrent drains, all changed enqueueing mutation paths trigger the runner, and the reduced schedules do not establish a broken correctness contract.

Important Files Changed

Filename	Overview
lib/ama/booking/server.ts	Adds a bounded waitUntil drain that repeatedly runs full operation batches while relying on existing lease-based concurrency controls.
lib/ama/booking/http.ts	Adds a typed webhook-outcome callback after successful event processing without changing signature verification or retry responses.
app/api/ama/stripe/webhook/route.ts	Starts inline processing only when webhook handling creates a booking and therefore enqueues finalization work.
app/api/ama/manage/[token]/cancel/route.ts	Starts operation draining after a successful guest cancellation and raises the route duration limit.
app/api/ama/manage/[token]/reschedule/route.ts	Starts operation draining after a successful guest reschedule and raises the route duration limit.
app/api/admin/ama/bookings/[bookingId]/route.ts	Starts operation draining after successful owner booking actions.
app/api/admin/ama/operations/[operationId]/route.ts	Starts the runner after successful retry or resolution actions, including a harmless empty pass for resolution-only actions.
vercel.json	Reduces AMA and media maintenance frequency, intentionally increasing clock-driven latency while allowing database autosuspension.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  M[AMA mutation] --> E[Enqueue durable operation]
  E --> K[kickAmaOperations]
  K --> W[waitUntil inline drain]
  W --> R[Runner: batches of 10]
  R --> P[Calendar / Meeting / Email / Stripe]
  C[30-minute AMA cron] --> R
  C --> H[Expired hold cleanup]
  C --> T[Reminders and retry backoff]
  MC[Hourly media cron] --> MR[Media reconciliation]

Loading

_{Reviews (1): Last reviewed commit: "perf(ama): drain durable operations inli..." | Re-trigger Greptile}

Context used (4)

• Knowledge Base — AMA booking flow
• Knowledge Base — AMA Admin Operations
• Knowledge Base — AMA Stripe Payments
• Knowledge Base — Media reconciliation, purge, geocoding, and privacy