A sophisticated content recommendation system powered by VMware Tanzu Greenplum, pgvector, and AI-generated embeddings. This project demonstrates how to build a complete AI-driven recommendation pipeline with a stunning modern web interface featuring real-time clustering and metallic design.

• Glass morphism design with metallic gradients and effects
• Interactive cluster rating system with custom metallic sliders
• Real-time re-clustering with adjustable parameters (5-25 clusters)
• Responsive design that works beautifully on all devices
• Dark theme with gold accents and shimmer animations

• Embedding generation using configurable AI models (OpenAI or local)
• KMeans clustering with PL/Python and scikit-learn
• Vector similarity search using pgvector for recommendations
• AI-generated cluster summaries for better user understanding
• Personalized preference modeling based on cluster ratings

• Dynamic re-clustering - Change cluster count and regenerate groupings
• Configurable embedding dimensions (supports 768D, 1536D, etc.)
• Local or cloud AI models via OpenAI-compatible APIs
• Export functionality for recommendations and analytics
• Comprehensive pre-flight checks for data integrity

• Greenplum Database 7.x with pgvector extension
• Python 3.8+ with virtual environment support
• AI Model Access (OpenAI API or local model server)

1. Clone and setup:

   git clone <repository-url>
   cd blog-recommendations
   python3 -m venv venv
   source venv/bin/activate
   pip install -r requirements.txt

2. Download the dataset:

   # Easy one-click download
   ./download_data.sh

   Alternative manual download:

   # Manual download and extract
   curl -o medium_data.csv.zip "https://storage.googleapis.com/kaggle-data-sets/748442/1294572/compressed/medium_data.csv.zip?X-Goog-Algorithm=GOOG4-RSA-SHA256&X-Goog-Credential=gcp-kaggle-com%40kaggle-161607.iam.gserviceaccount.com%2F20250915%2Fauto%2Fstorage%2Fgoog4_request&X-Goog-Date=20250915T195945Z&X-Goog-Expires=259200&X-Goog-SignedHeaders=host&X-Goog-Signature=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"
   unzip medium_data.csv.zip
   mv medium_data.csv medium_post_titles.csv
   rm medium_data.csv.zip

3. Configure environment (edit .env):

   # Database Configuration
   GP_HOST=localhost
   GP_PORT=15432
   GP_USER=gpadmin
   DB_NAME=demo
   
   # AI Model Configuration
   USE_LOCAL_MODELS=true
   LOCAL_API_BASE=http://127.0.0.1:1234/v1
   EMBEDDING_MODEL=text-embedding-nomic-embed-text-v2
   EMBEDDING_DIMENSIONS=768
   
   # Web Interface
   WEB_PORT=8081

4. Run the complete demo:

   ./run_demo.sh

┌─────────────────┐    ┌──────────────────┐    ┌─────────────────┐
│   Raw CSV Data │ -> │  Greenplum + AI  │ -> │  Web Interface  │
│   (Blog Posts)  │    │   Processing     │    │   (Metallic UI) │
└─────────────────┘    └──────────────────┘    └─────────────────┘
                              │
                              ▼
                    ┌──────────────────┐
                    │  Recommendation  │
                    │     Engine       │
                    └──────────────────┘

1. Load → Blog posts imported into Greenplum
2. Embed → AI generates vector embeddings for each post
3. Cluster → KMeans groups similar content (configurable clusters)
4. Summarize → AI creates thematic summaries for each cluster
5. Recommend → Vector similarity finds personalized content

• View AI-generated cluster summaries with sample articles
• Adjust cluster count (5-25) using the metallic slider
• Click "Re-cluster Data" to dynamically reorganize content

• Rate each cluster from 1-10 using metallic sliders
• Build your personalized preference vector in real-time
• Visual feedback with gold accent colors

• Get top 25 most interesting articles for you
• See 25 least relevant articles (for comparison)
• Export results with metadata and system information

blog-recommendations/
├── 🌐 web_app.py              # Flask web server with REST API
├── ⚙️ config.py               # Configuration management
├── 🔧 run_demo.sh             # One-click demo launcher
├── 📊 genvec.py               # Embedding generation
├── 🎯 cluster.sql             # KMeans clustering query
├── 📝 summarize.py            # AI cluster summaries
├── 🗄️ load_data.sh            # Data pipeline setup
├── 🔧 create_kmeans_function.sql  # PL/Python KMeans UDF
├── 📋 generate_schema.py      # Dynamic schema generation
├── 🌐 templates/
│   └── index.html             # Metallic UI template
├── 📱 static/
│   └── js/app.js              # Frontend JavaScript
├── 📜 scripts/                # Utility scripts
└── 📸 screenshots/            # Demo screenshots

// Via Web UI - adjust slider and click "Re-cluster Data"
// Or via API:
curl -X POST http://localhost:8081/api/recluster \
  -H "Content-Type: application/json" \
  -d '{"num_clusters": 20}'

# In .env file:
EMBEDDING_MODEL=text-embedding-3-large
EMBEDDING_DIMENSIONS=3072

# Complete pipeline:
bash load_data.sh              # Load CSV data
python genvec.py               # Generate embeddings
psql -f create_kmeans_function.sql demo  # Create KMeans function
psql -f cluster.sql demo       # Run clustering
python summarize.py            # Generate summaries
python web_app.py              # Start web interface

• GP_HOST, GP_PORT, GP_USER - Database connection parameters
• DB_NAME - Target database name

• USE_LOCAL_MODELS - Toggle between local and cloud AI models
• EMBEDDING_MODEL - Model for generating embeddings
• EMBEDDING_DIMENSIONS - Vector dimensions (768, 1536, 3072, etc.)
• CHAT_MODEL - Model for generating cluster summaries

• WEB_PORT - Flask server port (default: 8081)
• CLUSTER_SAMPLE_SIZE - Articles per cluster for summaries

The metallic theme uses CSS custom properties for easy customization:

/* Metallic color scheme */
.metallic-bg { /* Silver gradients */ }
.metallic-blue { /* Blue metallic accents */ }
.metallic-gold { /* Gold highlights */ }
.glass-panel { /* Glassmorphism effects */ }

• Vector Search: pgvector provides fast similarity search at scale
• Clustering: PL/Python enables in-database ML processing
• Caching: Configuration display caching for faster startups
• Responsive: UI adapts from mobile to large displays

• Media & Publishing: Personalized article feeds and content discovery
• E-commerce: Product recommendations based on user behavior
• Education: Adaptive learning content recommendations
• Research: Academic paper recommendation systems
• Corporate: Knowledge base content suggestion

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Test with your Greenplum setup
4. Commit your changes (git commit -m 'Add amazing feature')
5. Push to the branch (git push origin feature/amazing-feature)
6. Submit a pull request

This project demonstrates advanced Greenplum + AI capabilities for educational and commercial use.

• Greenplum Documentation
• pgvector Extension
• OpenAI API Documentation
• Flask Documentation

Built with ❤️ using Greenplum, pgvector, and modern web technologies

Showcasing the power of in-database AI and vector search for next-generation recommendation systems