The starting point
Make reflective support more structured than an open-ended chatbot through gated exercises, reflection capture, progress synthesis, and crisis-resource handoff.
A public-source portfolio case study for a reflective/wellness AI companion. The product intent is to make support more structured than an open-ended chatbot: listen first, identify a user-articulated challenge, offer a pre-authored exercise only after gate conditions, guide it stepwise, capture reflection, synthesize progress, and route crisis language toward human or emergency resources. It must not be represented as medical care, diagnosis, efficacy, deployment, or a compliance-certified system. All listed service opportunities are proposals, not shipped functionality.
What’s implemented
- Landing, auth, onboarding, chat, journal, progress/insights, settings, and billing-related routes
- Prisma persistence models for users, conversations/messages, frameworks, exercises, completions, insights, journals, institutions, subscriptions, and streaks
- Discovery-first AI chat orchestration with exercise gates, autonomy rules, crisis-mode rules, and non-medical prohibitions
- Database-backed exercise facilitation with phase tracking, reflection submission, completion, and insight creation
- Lexical crisis detection and a U.S. 988/911 crisis-resource path
- Aggregated mood/exercise/framework/streak/journal/conversation analytics
- Stripe customer, portal, webhook, and subscription-tier integration code
- Seeded framework and exercise assets
Technical approach & boundaries
Technology
- Next.js 14 App Router
- React 18 and TypeScript
- Tailwind CSS, Radix UI, Framer Motion, Zustand
- Prisma 5 with PostgreSQL-oriented schema
- NextAuth
- AI SDK with OpenAI/Anthropic/Google integrations
- Stripe
- Jest and Playwright
Limitations & readiness
- Active-looking MVP/prototype; implementation is visible, while deployment and critical-flow validation remain open checklist work.
- No verified production URL, deployment record, customers, or measured outcomes
- Critical-flow, auth/security, monitoring, payment, support, policy, and performance checks remain open in deployment documentation
- E2E suite has placeholders and commented auth setup for onboarding/crisis scenarios
- PWA/offline is explicit future work
- External AI, database, auth, and Stripe services require configured environments
- Lexical crisis detection is not clinical assessment or emergency care
- Privacy, encryption, HIPAA, audit-log, and data-sovereignty statements are unverified aspirations
- Source prohibits diagnosis, medical advice, outcome promises, and replacing professional care
Inside the idea.
Public project notes and a separate visual interpretation.
Stack
- Next.js 14 App Router
- React 18 and TypeScript
- Tailwind CSS, Radix UI, Framer Motion, Zustand
- Prisma 5 with PostgreSQL-oriented schema
- NextAuth
- AI SDK with OpenAI/Anthropic/Google integrations
- Stripe
- Jest and Playwright
Implemented notes
- Landing, auth, onboarding, chat, journal, progress/insights, settings, and billing-related routes
- Prisma persistence models for users, conversations/messages, frameworks, exercises, completions, insights, journals, institutions, subscriptions, and streaks
- Discovery-first AI chat orchestration with exercise gates, autonomy rules, crisis-mode rules, and non-medical prohibitions
- Database-backed exercise facilitation with phase tracking, reflection submission, completion, and insight creation
- Lexical crisis detection and a U.S. 988/911 crisis-resource path
- Aggregated mood/exercise/framework/streak/journal/conversation analytics
- Stripe customer, portal, webhook, and subscription-tier integration code
- Seeded framework and exercise assets
Repository links
Conceptual provenance. The alternate view is an original portfolio diagram based only on this public project record. It is not a product screenshot, live data, or a claim about a shipped interface.
What this could become for you.
These are proposed client adaptations—not claims that they have already been delivered.
Potential engagements & deliverables
- branded guided-reflection assistant with onboarding/consent, configurable prompts, persistence, crisis copy, and an exercise-library admin surface
- coach/facilitator toolkit with a multi-tenant catalog, phase runner, reflection capture, history, and progress exports
- employee/wellbeing journaling pilot with opt-in trends, aggregate-only dashboards, RBAC, retention/deletion controls, and security review
- recovery-program companion with consent/disclaimer flows, crisis escalation, human handoff, and reporting without treatment or efficacy claims
- AI product-hardening engagement covering auth/routes, provider abstraction, rate limits, observability, CI/deployment, env validation, Stripe webhooks, and release readiness
Related work
- Standalone reflective/wellness companion; no observable AppArchitect reference or shared product-family marker.