MedSchools.ai System Architecture v1.0

Complete system architecture document for MedSchools.ai covering:

Tech Stack

• Frontend: SvelteKit 2.x, Svelte 5, Tailwind CSS 4, shadcn-svelte
• Backend: SvelteKit server routes, Supabase, Upstash Redis
• AI: OpenAI GPT-4, text-embedding-3-small
• Payments: Stripe
• Voice: Vapi
• Email: Resend
• Hosting: Vercel

Database (7 schemas)

• med_schools: 26 tables, 173 schools
• content: 6 tables (blog, guides, glossary)
• user_data: 9 tables (profiles, activities, essays)
• billing: 2 tables (invite_codes)
• rag: 6 tables (~10K chunks)
• ai: 2 tables (chat sessions)
• scraper: 6 tables (2,821 pages)

Key Components

1. RAG System: Hybrid GPT-4 + community data (SDN/Reddit)
2. Redis Cache: User context (15min), school data (1hr)
3. Auth: Supabase Auth (Google, Apple, Email)
4. Billing: Stripe ($49.95/mo, $399/cycle, $99 interview)
5. Invite Codes: Beta access system

API Routes

• /api/chat, /api/invites, /api/stripe, /api/user, etc.

Full document saved to: projects/medschools_ai/SYSTEM_ARCHITECTURE.md