​
Overview

The iOS app is fully integrated with the Python matching API v2 deployed at Railway, providing sophisticated personalized matching instead of random school selection.

​
Architecture Components

​
1. Data Models (APIModels.swift)

  • Request/response models for API communication
  • Conversion methods between API and iOS models
  • Feedback options for post-swipe collection

​
2. API Client (MatchingAPIClient.swift)

Complete API client with error handling and timeout configuration:
  • Production URL: https://findu-matching-production.up.railway.app
  • Development URL: http://localhost:8000 (when DEBUG flag is set)

Endpoints

  • /recommendations - Get personalized matches
  • /swipe - Record swipe interactions
  • /feedback - Submit learning feedback
  • /health - Check API status

Configuration

Located in MatchingAPIClient.swift line 18Timeout: 30 secondsRetry logic: Built-in error handling

​
3. New ViewModels

​
CardsViewModelAPI

Replaces the original CardsViewModel with full API integration:
  • Manages recommendation fetching
  • Handles swipe recording
  • Coordinates with learning system

​
LearningStateManager

Tracks learning phases for improved recommendations:
  • Discovery (1-10 swipes): Initial preference learning
  • Refinement (11-30 swipes): Fine-tuning matches
  • Stable (31+ swipes): Consistent personalized results

​
FeedbackManager

Handles post-swipe feedback collection with smart triggering.

​
4. Updated Views

  • SwipeableCardsViewAPI: API-integrated swiping interface
  • CardStackViewAPI: Works with CardModel instead of School
  • SwipeFeedbackView: UI for collecting user feedback

​
5. Extended Models

​
CardModel

Enhanced with API-specific fields: 
// New API fields
var matchScore: Double
var explanations: [String]
var isProbe: Bool
var matchedMajors: [String]
var distance: Double?

​
StudentSchoolInteraction

Added API tracking fields for learning system integration.

​
Key Features

​
🎯 Personalized Matching

  • Match scores from 0-100% based on student preferences
  • Detailed explanations showing why schools were recommended
  • Grade calculation (A/B/C/D/F) based on match score

​
🧠 Learning System

1

Discovery Phase (1-10 swipes)

Initial preference learning from user behavior
2

Refinement Phase (11-30 swipes)

Fine-tuning recommendations based on patterns
3

Stable Phase (31+ swipes)

Consistent personalized results with ongoing adaptation

​
🔍 Probe Schools

  • Every 4th swipe tests preference boundaries
  • Schools marked as probes for special algorithm handling
  • Helps discover unexpected preferences

​
🌍 Geographic Intelligence

Smart geographic distribution:
  • 50% nearby schools (within preferred distance)
  • 30% slightly beyond preferences (expansion)
  • 20% discovery schools (new regions)

​
💬 Feedback Collection

  • Smart triggering after pattern breaks or probe schools
  • Minimal disruption with maximum 5 options
  • Feeds directly back into learning algorithm

​
Data Flow

​
Initial Load Process

​
Swipe Recording

​
API Response Structure

{
  "recommendations": [
    {
      "school": {
        "id": "uuid",
        "name": "Example University",
        "city": "Boston",
        "state": "MA"
        // ... full school data
      },
      "match_score": 0.75,
      "explanations": [
        "Strong Business program",
        "Perfect size match", 
        "Within your budget"
      ],
      "is_probe": false,
      "matched_majors": ["Business Administration"],
      "distance": 125.5
    }
  ],
  "learning_state": {
    "phase": "discovery",
    "progress": 0.4,
    "preference_weights": {
      "academic_selectivity": 0.8,
      "cost_weight": 0.6,
      "size_preference": 0.9
    }
  },
  "total_swipes": 4
}

​
Testing the Integration

​
1. Build and Verify

1

Build and run the app

The app should automatically use API matching
2

Check console logs

Look for “[API]” prefixed logs indicating API calls
3

Verify API connection

Should see “Finding personalized matches…” loading state

​
2. Test Learning System

  • Swipe through 10+ schools to trigger refinement phase
  • Should see occasional feedback forms
  • Match quality should improve over time
  • Console logs show learning state changes

​
3. Debug API Issues

​
Migration Guide

​
From Random to API Matching

If you need to rollback to random matching:
Only use this for debugging. API matching provides significantly better user experience.
1

Update ios_app.swift

Change CardsViewModelAPI back to CardsViewModel
2

Update ExploreView.swift

  • Change SwipeableCardsViewAPI → SwipeableCardsView
  • Change CardsViewModelAPI → CardsViewModel
3

Update HomeView.swift

Change CardsViewModelAPI back to CardsViewModel

​
Known Limitations

Network Dependency

No Offline SupportRequires active internet connection for recommendations

Profile Requirement

Complete Student ProfileAPI needs full student profile data to generate matches

Location Hardcoded

NYC CoordinatesCurrently uses hardcoded coordinates, needs geocoding integration

Error Recovery

Limited Retry LogicBasic error handling, could use more sophisticated retry mechanisms

​
Next Steps

​
Planned Enhancements

  1. Location Services Integration
    • Request location permission
    • Geocode student address
    • Pass real coordinates to API
  2. Enhanced UI Features
    • Show match explanations on cards
    • Display learning phase progress
    • Highlight matched majors with ⭐
  3. Improved Error Recovery
    • Add retry mechanisms for failed requests
    • Better error messages for users
    • Fallback to cached data when offline
  4. Performance Optimizations
    • Implement recommendation caching
    • Prefetch next batch of schools
    • Optimize image loading for better performance
For setup instructions and troubleshooting, see the iOS API Setup Guide.