Samsung Reflect – AI-Powered Journaling Companion

Preview

Copy.of.SAMSUNG.REFLECT.3.mp4

Youtube Demo

🔗 Watch the full demo on YouTube

A comprehensive MERN stack application focused on smart journaling and self-reflection, powered by multi-modal AI for seamless use of text, photos, audio, and intelligent media processing.

---

✨ Features

📔 Core Journaling Experience

• Intelligent Journal System
  Create rich stories combining text, photos, videos, and audio, enhanced with AI mood detection.

• Smart Media Suggestions
  Get relevant photo, video, or audio recommendations from your gallery as you type.

• Flexible Canvas
  Drag, resize, and crop journal elements to design your stories your way.

• Memories Library
  Save, revisit, edit, and favorite your journals with privacy controls like pass-key-locked folders.

• Insights Dashboard
  View streaks, activity reports, and mood trends to stay motivated.

• Settings & Privacy
  Manage preferences, data, and security with full control.

• Standby Mode
  Access and use your journals in standby mode across all your Samsung devices, keeping your reflections available anytime, even when your device is idle.

---

🤖 Enhanced AI Features

• Generative AI Storytelling Instantly turn keywords into rich, descriptive diary entries using Zephyr-7B (via Hugging Face).

• Smart "Thinking" Autocomplete Context-aware text suggestions that finish your sentences as you type, powered by LLMs.

• Text-to-Image Generation Create custom cover art and illustrations for your journals using Stable Diffusion XL.

• Semantic Media Search
  Quickly find the right photos, videos, or audio by location, objects, context, and more to enrich your journal entries.

• Object & Scene Detection
  Automatically identify objects, scenes, and faces in your media to improve recommendations and organization.

• AI-Powered Media Processing
  Analyze and process images, audio, and text in real time for smart suggestions, mood insights, and seamless journal creation.

• Real-Time Performance
  Fast and efficient pipeline for media clustering, retrieval, and personalized recommendations.

• Android WebView Support
  Smooth and optimized experience for mobile devices.

---

🚀 About Samsung Reflect

Samsung Reflect is designed to make journaling more meaningful and intelligent, transforming memories into personalized stories while keeping your data secure and accessible across Samsung devices.

Architecture

Frontend (React + TypeScript)

• Framework: Vite + React 18 + TypeScript
• UI Components: shadcn/ui with Tailwind CSS
• State Management: TanStack Query
• Routing: React Router v6
• AI Features:
  • Image upload with drag & drop
  • Real-time detection overlays
  • Search interface with filters
  • Results visualization

Backend (Node.js + Express)

• Runtime: Node.js with Express and TypeScript
• Database: MongoDB with Mongoose ODM
• Job Queue: BullMQ with Redis
• File Processing: Sharp for image manipulation
• Vector Search: Qdrant vector database
• API Design: RESTful with comprehensive error handling

ML Service (Python + FastAPI)

• Framework: FastAPI with async support
• Generative AI (New):
  • Text & Logic: Hugging Face Inference API (Zephyr-7B-Beta) for storytelling and autocomplete.
  • Image Generation: Stable Diffusion XL Base 1.0.
• Computer Vision (Local):
  • Object Detection: YOLOv8
  • Scene Classification: Places365 ResNet50
  • Embeddings: CLIP ViT-B/32
• Security: Environment-based secret management (.env) for API keys.

Quick Start

Prerequisites

• Node.js 18+ (for local development)
• Python 3.11+ (for ML service development)
• Docker v20.10+ – for consistent environment setup
• Hugging Face API Token – Required for Generative AI features (Free tier is sufficient)

Development Setup

1. Clone and Configure

   git clone https://github.com/JasmithaTholeti/Samsung_Reflect.git
   cd Samsung_Reflect

2. Set up ML Service Secrets

    cd ml-service
    touch .env
    echo "HF_API_KEY=your_hugging_face_token_here" >> .env
    cd ..

3. Download AI Models

   # Create models directory
   mkdir -p models/{yolo,places365,clip}
   
   # Download YOLOv8 nano model (6MB)
   wget https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8n.pt -O models/yolo/yolov8n.pt
   
   # Places365 and CLIP models will be downloaded automatically on first run

4. Start Services

   docker-compose up -d

5. Verify Installation

   # Check API health
   curl http://localhost:3001/api/health
   
   # Check AI models
   curl http://localhost:3001/api/models/health

6. Access Application

   • Frontend: http://localhost:5173
   • API: http://localhost:3001
   • ML Service: http://localhost:8000
   • AI Search: http://localhost:5173/search

Configuration

Environment Variables

# Database
MONGO_URI=mongodb://localhost:27017/samsung-reflect

# AI Configuration
SERVER_INFERENCE=true
MODEL_DIR=./models
MIN_DETECTION_SCORE=0.25
VECTOR_DB_URL=http://localhost:6333

# Feature Flags
ENABLE_CLIENT_INFERENCE=false
ENABLE_RERANKING=true

# Ranking Weights
OBJECT_WEIGHT=0.7
SCENE_WEIGHT=0.3
AGGREGATION_METHOD=weighted

# Generative AI Secrets (Required)
HF_API_KEY=your_hugging_face_token_here

Development

Project Structure

SamsungReflect/
├── Frontend/                 # React application
│   ├── src/
│   │   ├── features/search/  # AI search components
│   │   ├── components/       # Shared UI components
│   │   └── pages/           # Application pages
├── backend/                  # Node.js API
│   ├── src/
│   │   ├── models/          # Database models
│   │   ├── routes/          # API routes
│   │   └── services/        # Business logic
├── ml-service/              # Python ML service
│   ├── main.py             # FastAPI application
│   └── requirements.txt    # Python dependencies
├── models/                 # AI model files
├── config/                # Configuration files
└── docker-compose.yml     # Development setup