RelateAI is an elegant AI-powered relationship assistant built for iOS using SwiftUI and modern Apple frameworks.
It helps you understand conversations, decode emotional tone, and craft the perfect replies — all powered by AI ✨
RelateAI blends psychology, AI, and communication science into a seamless mobile experience.
With RelateAI, you can:
- Analyze conversation screenshots or pasted messages
- Detect emotional tone & subtle attachment patterns
- Identify toxic / warm / mixed communication signals
- Generate replies with AI (gentle, calm, flirty, direct, confident)
- Store and revisit insights
- Enjoy a beautiful minimal dark UI focused on clarity and comfort
| Feature | Technology |
|---|---|
| UI Framework | SwiftUI |
| Architecture | MVVM |
| AI Backend | HuggingFace Inference API + Meta LLaMA 3/3.1 |
| Networking | Async/Await + URLSession |
| Text Processing | Custom Prompt Engineering |
| State Management | @StateObject / @ObservedObject |
| Animations | Smooth SwiftUI transitions & haptics |
| Design System | Custom palette (Orange / Amber / Soft Purple) |
| Font | MontserratAlternates |
| Backend | None (client-side AI only) |
- 🔍 Conversation Insight: Emotional tone analysis, interest level, confusion, avoidance, anxiety.
- 💬 AI Reply Generator: Built on top of Meta LLaMA + HF models.
- 📊 Psychology-backed patterns: Secure vs avoidant vs anxious message behavior.
- 🌙 Dark UI: Modern, clean, distraction-free.
- 🔄 Reply Refinement: Re-write softer, shorter, longer, or more playful responses.
- 🎯 On-device hints: Smart suggestions for how to communicate better.
RelateAI uses a centralized design system (DesignSystem.swift) that ensures consistent UI across the entire app.
| Element | Description |
|---|---|
| 🎨 Colors | rOrange, rAmber, rPurple, rSoftDark |
| 🌈 Gradient | LinearGradient.rCTA |
| 🖋️ Font | .montserrat() (MontserratAlternates default) |
| 🧱 Components | RPrimaryButton, RCard, RChatBubble |
RelateAI is powered by a hybrid AI system:
Used for:
- Deep emotional understanding
- Relationship tone interpretation
- Long-text conversation pattern detection
- Human-like reply generation
Meta’s models provide:
- More natural relationship-oriented language
- Better understanding of subtle emotional nuance
- Low hallucination in conversational context
- Excellent short-form message generation
Used for:
- Emotion classification
- Toxicity detection
- Sentiment scoring
- Attachment-style pattern hints
HF_API_KEY=your_key_here
META_API_KEY=your_key_here
And added to .gitignore:
.env
Batuhan Küçükaydın
Software Engineer | iOS Developer | AI Builder
📫 LinkedIn • GitHub • Medium
If you like this project, please consider giving it a star 🌟
It truly supports my work and future projects!