Medfold

Your health records, intelligently folded.

Medfold is a native iOS app that gives you a single, secure place to store all your health documents. Upload a lab result, prescription, or imaging report and AI automatically summarizes it, extracts key values, and makes it searchable. Then ask questions about your records in a conversational chat.

Features

Document Vault - Upload health records via camera, photo library, or file picker. Organize across 7 categories (Labs, Rx, Imaging, Insurance, Visits, Immunization, Other). Search across titles, tags, AI summaries, and provider names.

AI Analysis - Every uploaded document is automatically processed: a 2-3 sentence summary, category suggestion, extracted key values (lab metrics, dosages, dates, reference ranges), and provider identification.

AI Chat - Ask questions about your health records in natural language. The assistant loads your recent documents as context and can reference specific records in its answers. Includes suggested prompts like "Summarize my recent lab results" and "Are there any concerning patterns?"

Security First - Row Level Security on every database table, per-user storage folder isolation, AES-256 encryption at rest, Sign in with Apple, email/password, and Face ID/Touch ID biometric unlock.

Tech Stack

Layer	Technology
iOS App	Swift, SwiftUI, MVVM (iOS 17+)
Backend	Supabase (Postgres 15, Auth, Storage, Edge Functions)
AI	Provider-agnostic: Anthropic Claude or OpenAI GPT, switchable via env var
Auth	Sign in with Apple, email/password, Face ID/Touch ID
Edge Functions	Deno/TypeScript with SSE streaming

Architecture Highlights

• Protocol-oriented services: Every service (Auth, Documents, Storage, AI) is defined as a Swift protocol with a concrete implementation, making it easy to swap backends or mock for testing.
• Provider-agnostic AI: A shared TypeScript AIProvider interface with Anthropic and OpenAI implementations. Switch providers by changing one environment variable.
• Streaming chat: The ai-chat Edge Function returns Server-Sent Events, normalizing both Anthropic and OpenAI streaming formats into a unified event shape.
• JSONB for extracted data: AI-extracted values are stored in a PostgreSQL JSONB column with a GIN index, allowing flexible schema-free storage of lab metrics, dosages, and reference ranges.
• Strict Swift concurrency: SWIFT_STRICT_CONCURRENCY: complete enabled, all ViewModels are @MainActor, async/await throughout.

Project Structure

Medfold/                  # iOS app source
  App/                    # Entry point, tab navigation
  Models/                 # Document, Profile, AIConversation
  Views/                  # SwiftUI views (Auth, Documents, AI Chat, Profile)
  ViewModels/             # MVVM view models
  Services/               # Protocol-based service layer + Supabase implementations
  Utilities/              # Supabase client singleton, constants

supabase/                 # Backend
  migrations/             # Postgres schema, enums, RLS policies, triggers
  functions/
    _shared/              # AI provider abstraction (Anthropic + OpenAI)
    analyze-doc/          # Document analysis Edge Function
    ai-chat/              # Streaming chat Edge Function

Getting Started

Prerequisites

• Xcode 16+
• A Supabase project (free tier works)
• An AI provider API key (Anthropic or OpenAI)

Setup

1. Clone the repo

   git clone https://github.com/shaan-ad/Medfold.git
   cd Medfold

2. Configure Supabase credentials in Medfold/Utilities/Constants.swift:

   static let supabaseURL = "https://your-project.supabase.co"
   static let supabaseAnonKey = "your-anon-key"

3. Run the database migration: Copy supabase/migrations/001_create_tables.sql into your Supabase SQL Editor and execute it. This creates all tables, enums, RLS policies, triggers, and the storage bucket.

4. Set Edge Function secrets in your Supabase dashboard:

   AI_PROVIDER=anthropic
   ANTHROPIC_API_KEY=your_key_here
   

5. Deploy Edge Functions:

   supabase functions deploy analyze-doc
   supabase functions deploy ai-chat

6. Open in Xcode: Open Medfold.xcodeproj, select a simulator or device, and run.

Note: The project uses XcodeGen. To regenerate the Xcode project from project.yml, run xcodegen generate.

Roadmap

• Secure document vault with upload, search, and category filtering
• AI document analysis with summary and key value extraction
• AI chat with health record context and streaming responses
• Multi-method auth (Apple, email, biometric)
• Apple HealthKit integration (Apple Watch data)
• Whoop and Oura wearable sync
• Structured data entry (medications, allergies, vitals)
• Trend charts and health dashboards

License

MIT