jiggly-bubble Sprint — Completion Log

Goal: Get all 7 portfolio surfaces (cwa, pro-website, pewsearch, itw, sermonwise, funeralwiseai, vetwiseai, wiseaiagency) "rock solid" and unblock cold-outreach launch.

Plan target: First cold-outreach batch on Day 7 of ~10. Reality: Day 1 = code/infra functionally complete. Cold-outreach can launch as soon as founder's organic operational items finish (Resend reputation maturation, prospect source disclosure decisions).

Streams shipped (5 PRs, 1 day)

Stream	PR	Scope
A — Demo engine multi-vertical	#297	FuneralPreview + VetPreview templates; `/api/chatbot/stream` vertical-aware routing; 3 draft migrations
B — Browser voice demo	#300	`<BrowserVoiceDemo />` widget; 60s cap; capacity UX; embedded in funeral + vet previews
C — /ensure-solid 9 surfaces	knowledge#96	Per-property scorecards + `portfolio-20260504-v7.md` rollup; first-ever scoring for funeral, vet, agency
D — Per-vertical content	#298 + knowledge#96	3 sales playbooks; 4-email cold-outreach × 3 verticals; lifecycle TS files (funeral + vet); brand voice extension
E+F — SEO + agency layer	#299 + knowledge#96	4 city pages (Ingersoll/London ON/Woodstock/Tillsonburg); robots.ts; agency contact form; playground vertical picker; social strategy + 30-day calendar

Mid-sprint customer-impacting fixes (4 PRs)

Surfaced during founder's hands-on testing while sprint was in flight:

PR	Issue	Resolution
#301	Admin cancel flow dead-ended at Settings (the "cancel" link redirected to a `#cancellation` anchor that was never built; Settings tab had been converted to a slide-over 2026-04-18)	Redirect to `/api/stripe/portal?token=...` (existing Stripe Customer Portal)
#302	Cancel flow had 3 clicks (modal → modal → "Cancel in Stripe" link)	Killed second confirmation screen; modal click → straight to Stripe portal
#303	After Stripe cancellation succeeded, dashboard still showed "ACTIVE". Webhook was setting top-level `cancel_at_period_end=true` but not the per-product `<product>_cancel_at_period_end` flags that the dashboard reads	Webhook fanout: when whole subscription is scheduled for cancellation, fanout to every active per-product flag based on `multiItemUpdate.has_*_subscription`
#307	3 active P0 ops_errors dragging CWA Ops -15pp from v6 to v7 scorecard	(a) Sweep last-item guard (cancel vs item-delete); (b) bundle item-id dedup in sweep loop; (c) change-tier `resource_missing` → graceful 410 not P0

Audit follow-up PRs (3)

PR	Audit finding
#304	CASL physical address gap in lifecycle email footers (church + funeral + vet)
#306	`cold-outreach-director-transfer` spec broken since 2026-04-29 (hero button rewired to open modal; spec was clicking hero, not inner WebRTCDemoButton)
pewsearch#32	Admin SettingsTab said "Basic chatbot included with Premium" instead of "Pro Website" — churn-risk overpromise
knowledge#97	Test registry entries for the 3 new specs (preview-funeral, preview-vet, chatbot-vertical-aware)

Total PRs merged: 12 (8 churchwiseai-web + 1 pewsearch + 2 knowledge + 1 + voice-agent deploy)

Migrations applied to prod Supabase (5)

20260504_product_knowledge_vet_funeral_drift.sql — funeral pricing drift fix ($999 → $745), 9 new vet rows
20260505_drop_church_voice_agents_fk.sql — idempotent (FK didn't exist; audit-confirmed safe)
20260505_create_vet_knowledge_base.sql — per-tenant vet Q&A table
20260505_create_vet_clinics.sql — vet identity table (mirrors funeral_homes + species_supported TEXT[])
20260505_create_premium_vet_clinics.sql — vet subscription table (FK to vet_clinics); supersedes the never-applied 20260502 draft

Mid-sprint DDL discoveries: the previously-feared church_voice_agents_church_id_fkey constraint did NOT exist in production — comment at provision.ts:449-463 was stale. Audit-confirmed via pg_constraint query before applying drop migration.

Production data writes (founder-approved)

39 funeral product_knowledge rows retagged from vertical='church' → vertical='funeral' (brand_scopes: funeralwiseai, playground-funeral)
3 customer rows backfilled with per-product cancel flags to match webhook intent (Erin Reformed Church, Melvindale CoG bundle, E2E test row)
5 P0 ops_errors marked resolved with fix_summary audit trail (FA-082 unknown tier ×2, sweep race, change-tier resource_missing, voice-health orphan rule)
$79.95 refund issued via Stripe live mode to Debora Moelker (pyr_1TTR2VFaoK5IPzNoBezTmzCd)

Founder-side organic operational items completed

Vetwiseai bare→www redirect inverted to www→bare (matches funeral/church pattern)
veterinarywiseai.com added to Resend (DKIM/SPF/DMARC verified live via DNS)
veterinarywiseai.com added as Google Workspace domain alias
4 Cal.com event types confirmed live: church/funeral/vet/agency demos at cal.com/john-moelker/<vertical>-demo
funeralwiseai.com Resend verification confirmed (was completed prior to sprint)

Decisions locked (in order made)

Browser voice demo built now, not deferred (Stream B)
Demo phone strategy: shared Telnyx + 4-char code (no per-vertical numbers; cheaper + ships today)
Local SEO: wiseaiagency.com only, 4 cities (not mirrored to per-vertical sites)
Affiliate / pay-per-lead deferred post-launch — re-visit at 10+ paying customers per vertical OR credible partner offer (knowledge/decisions/2026-05-04-affiliate-deferred.md)
wiseaiagency.com kept as hostname rewrite, not split to separate Vercel project (knowledge/decisions/2026-05-04-wiseaiagency-rewrite.md)
CASL physical address: 125 Concession Street, Ingersoll, ON, Canada N5C 1G2 — hardcoded across 3 cold-outreach templates + 3 lifecycle email footer helpers
Vet voice demo widget shows "Demo not found" until vet identity tables exist — accepted as a soft-fall fallback for the first integration; tables shipped same day in PR #305
vetwiseai.com domain migration scheduled for May 12 when auction transfer completes (knowledge/decisions/2026-05-04-vetwiseai-domain-migration.md to be written; tracked in DECISION_LOG)

What was discovered (and would have shipped wrong without the sprint)

TypeScript build blocker in Stream A's chatbot route (maxTokens instead of maxOutputTokens on Vercel AI SDK 6) — would have failed pnpm build on first deploy. Caught by Stream B's audit before merge.
Vet vertical had NO identity layer despite a fully-shipped marketing site + Stripe products. provision.ts wrote vet prospects to outreach_contacts only; /api/livekit/token returned 404 for vet slugs. Caught by Stream B; closed in PR #305.
data.py two-step join bug in voice-agent-livekit/verticals/vet/ — _fetch_premium_vet_clinic was selecting identity columns from the wrong table. Pre-existing latent bug; corrected as part of #305.
Stale provision.ts comment flagged a non-existent FK as the funeral-provisioning blocker for weeks. Audit verified the FK was already absent.
tenant_voice_agents.vertical enum inconsistency — 'vet' in this table vs 'veterinary' in organization_settings.vertical. Pre-existing; documented; pre-launch normalization recommended.

Outstanding (not blocking cold-outreach launch)

Severity	Item	Notes
P2	`cwa-cancelled` + `cwa-trial-expired` journey specs	Authoring + run against demo tokens
P2	`pro-website-cancelled-tombstone` spec	Same
P2	Viewport Playwright runs stale on PewSearch + ITW	Need Chromium-installed run from main checkout
P2	`voice-routing-integration` test broken since 2026-04-22	Voice-agent-engineer territory
P2	`consistency-price-matching` spec re-run	After today's vet/funeral pricing updates
P2	`tenant_voice_agents.vertical` enum normalization (`'vet'` ↔ `'veterinary'`)	Pre-launch normalization
P2	DMARC tightening for funeralwiseai + vetwiseai (`p=none` → `p=quarantine`)	After 2-4 weeks of clean sending
Future	May 12 — `veterinarywiseai.com` → `vetwiseai.com` migration	30-min code + DNS work; 301-forever the old domain

Cold-outreach launch readiness

All 4 verticals ready to send. Recommended first-batch sequencing:

Day 1 (today / tomorrow): Founder reviews 30-day social calendar (knowledge/marketing/social-calendar-30day.md) and per-vertical cold-email sequences in knowledge/narrative/cold-outreach-emails-{church,funeral,vet}.md for tone/personal-detail adjustments
Day 2: First send — small test batch (5-10 prospects) per vertical from outreach_contacts already-scraped pool. Validate Resend sending, link tracking, reply handling.
Day 3-4: Scale to per-day cadence. Watch bounce rates + replies.
Day 5+: First demo bookings via Cal.com.

Process lessons (for future sprints)

Audit before migrate (feedback_never_migrate_before_audit.md) — saved a false-positive FK drop and surfaced 4+ pre-existing bugs hiding behind the assumed broken state.
Founder's hands-on testing during the sprint surfaces customer-facing UX bugs that no audit catches — the 3 cancel-flow PRs (#301, #302, #303) all came from the founder trying to use the dashboard mid-sprint.
Worktree-edit-only subagent pattern continues to work well — no agent ever committed; orchestrator owns commits + PR creation + merge timing. No worktree collisions across 7 parallel agents in one day.
Same-day full-stack vet vertical completion (DB schema + provisioning code + chatbot route + voice agent + browser voice widget + cold-outreach copy + lifecycle emails) was achievable because each layer was a 1-stream subagent with clear scope.

Sprint complete. No further code work required for cold-outreach launch.

Streams shipped (5 PRs, 1 day)​

Mid-sprint customer-impacting fixes (4 PRs)​

Audit follow-up PRs (3)​

Migrations applied to prod Supabase (5)​

Production data writes (founder-approved)​

Founder-side organic operational items completed​

Decisions locked (in order made)​

What was discovered (and would have shipped wrong without the sprint)​

Outstanding (not blocking cold-outreach launch)​

Cold-outreach launch readiness​

Process lessons (for future sprints)​