Four items are blocked on your input. These must be resolved before development can begin.
Request sent 2026-03-26. Jonathan is linked to the project but cannot read environment variables. Without this, the full 63-variable configuration must be manually reconstructed.
Steps: Vercel Dashboard → holistic-health-marketplace → Settings → General → Project Members → Add Member (noboxcreatives email) → Role: Member
Also needed: confirm if any env vars are marked "Sensitive" (Owner-only access).
Option A: Shared Neon DB — risk of breaking Blake's data during development.
Option B: Separate Neon project — isolated, safe, uses seed data (6 practitioners, 5 clients, 8 bookings). Recommended
Option C: Both — read-only access to Blake's + own dev DB. Best for Phase 2.
Key question: Have the 60K+ imported listings been loaded into Blake's DB, or is that pending?
The schema has 80+ models. An estimated 30% serve core marketplace functions (search, booking, payments, claiming). The remaining 70% support features explicitly deferred in the CONTEXT.md kickoff notes.
Decision needed: Which domains must work in the Apr 24 prototype?
| Feature | In Codebase | MVP? | Decision |
|---|---|---|---|
| Practitioner search + profiles | Full | Yes | Proceed |
| Appointment booking (Cal.com) | Full | Yes | Proceed |
| Payment processing (Whop) | Full | Yes | Proceed |
| Profile claiming | Full | Yes | Proceed |
| AI search agent | Partial | Yes | Needs input |
| Enterprise B2B | Full | Deferred | Strip? |
| GLP-1 program | Full | Deferred | Strip? |
| LMS (courses) | Full | Deferred | Strip? |
| Lab orders | Full | Deferred | Strip? |
| Health journaling | Full | Deferred | Strip? |
| In-app messaging | Full | Not specified | Keep? |
Currently all Vercel deployments are Preview — no Production deployment exists. Questions:
Five work items that can proceed now or are blocked only on the decisions above.
Create Neon DB, generate .env.local, install dependencies, push schema, seed data, verify app loads.
Run production build locally, compare against Blake's Vercel output, identify env-dependent failures.
Fork to jgatlit/holistic-health-marketplace with dual remotes (origin = fork, upstream = Blake's).
Create new project under ai-chemist scope linked to fork. Push env vars, deploy independently.
Map which schema domains are active in UI vs. scaffolded vs. dormant. Recommend minimum viable schema.
20+ integrated services across 6 infrastructure layers.
| Layer | Technology | Role | Domain |
|---|---|---|---|
| Frontend | Next.js 14.2 (App Router) | TypeScript, Tailwind CSS, src/ directory | Core |
| ORM | Prisma 6.x | 84KB schema, 80+ models, db push strategy | Core |
| Database | Neon PostgreSQL | Serverless, auto-scaling | Core |
| Auth | Auth.js v5 | Email/password + Google OAuth, JWT, MFA | Core |
| Payments | Whop | Direct charges, checkout sessions, refunds | Financial |
| Business Credit | Capital OS | Wallets, auto-sweep, card management | Financial |
| Scheduling | Cal.com v2 | Managed users, event types, slot availability | Core |
| Resend | Transactional (verification, bookings, reviews) | Core | |
| SMS | Twilio | Booking reminders with credit system | Clinical |
| AI | Claude Haiku | Session notes (SOAP), rate-limited 20/day | Clinical |
| Video | Daily.co | HIPAA-compliant telehealth rooms | Clinical |
| Health Data | Terra API | 200+ wearable providers | Clinical |
| Insurance | Thimble | Embedded GL + professional liability | Enterprise |
| Errors | Sentry | Server, client, edge (10% sample rate) | Infra |
| Rate Limiting | Upstash Redis | Falls back to in-memory | Infra |
| Storage | Vercel Blob | File uploads and attachments | Infra |
| Analytics | GTM + GA4 + Meta + TikTok + LinkedIn | 5-platform conversion tracking | Infra |
| Enrichment | practicenear-engine (Python) | Google Maps, static pages, llms.txt | Core |
The three critical pipelines that define the marketplace: booking lifecycle, payment/earnings, and listing-to-profile claiming.
80+ models across 15 domains. Estimated ~30% serve the MVP. The remaining 70% support deferred features that are already scaffolded in the codebase.
| Domain | Models | MVP Status | Notes |
|---|---|---|---|
| Auth & Users | 5 | Active | User, Account, Session, tokens |
| Practitioner & Services | 5 | Active | Practitioner (60+ fields!), Service, Credential, Availability |
| Bookings | 1 | Active | Full lifecycle: PENDING → COMPLETED/NO_SHOW |
| Payments | 5 | Active | Payment, Refund, PackageCredit, Wallet, WalletTransaction |
| Listings & Search | 2 | Active | ImportedListing, ListingSearchDocument — Typesense target |
| Communication | 8 | Partial | Notifications MVP; messaging, email campaigns may defer |
| Clinical | 5 | Partial | SessionNote (AI) likely MVP; IntakeForm, TreatmentProtocol may defer |
| Health Tools | 6 | Deferred | Journal, symptoms, supplements, food diary, photos |
| Health Data (Terra) | 3 | Deferred | Wearable connections, snapshots, sharing |
| Lab Integration | 2 | Deferred | 3 providers: Microbiome Labs, Rupa, Everlywell |
| Enterprise B2B | 5 | Deferred | Organization, WellnessProgram, BulkBooking, ApiKey |
| GLP-1 Program | 5 | Deferred | Affiliate, referral, payout (80/20 split) |
| LMS | 4 | Deferred | Course, Lesson, Enrollment, Completion |
| HIPAA Compliance | 3 | Partial | Schema ready; enforcement for Phase 2 |
| Platform & Audit | 6+ | Active | PlatformConfig, AuditLog, ProcessedWebhook, etc. |
27 API routes, 20+ server action files, 8 cron jobs, 6 webhook handlers — all following consistent patterns.
Practitioner search, list, services, slot availability, tracking, health check, auth register/verify
Auto-complete, auto-sweep, reminders, SMS reminders, no-shows, payouts, reindex, review follow-ups
Cal.com, Whop, Capital OS, GLP-1, Lab Results, Terra — all HMAC-SHA256 verified with idempotency
Bulk listing import (max 1000/batch), v1 API for organizations with scoped API keys
Three primary user journeys through the marketplace.
Five enhancement tracks that activate after Phase 1 gates clear.
Replace Prisma-based ListingSearchDocument with Typesense parametric filtering. Amazon-style intelligent search at 60K+ scale.
Semantic search layer for AI agent-guided discovery. "Find someone near me for lower back pain."
Infrastructure migration eval (Railway), compliant LLM selection, encryption enforcement. Schema already prepared.
Agentic browsers simulating 1,000+ concurrent users. No test infra beyond Playwright smoke tests currently.
Strip PMS-level complexity from practitioner dashboard. Assess first, then simplify with Blake's agreement.
| Risk | Likelihood | Impact | Mitigation |
|---|---|---|---|
| Schema too large for MVP — 80+ models create cognitive overhead and slow iteration | High | Medium | CP-5: Assess and identify minimum viable schema. Hide non-MVP routes. |
| Vercel-provisioned integrations (Neon, KV, Blob) may not be duplicable to Jonathan's scope | Medium | High | CPI-3: Get Member access. Fallback: provision own integrations independently. |
| 95 stale branches in Blake's repo create merge conflict risk | Medium | Low | Work from main only. Clean up branches after alignment with Blake. |
| No Production deployment — go-live process undefined | Medium | High | CPI-4: Define production deployment workflow before Phase 2. |
| Prisma db push without migrations — schema changes can cause data loss with real data | High | High | Implement migration strategy before real practitioner data enters the system. |
| 6% platform fee hardcoded in PlatformConfig seed — no admin UI to adjust | Low | Medium | Verify admin dashboard can modify or plan direct DB update path. |