CiteConnect is an intelligent research paper recommendation system that combines retrieval-augmented generation (RAG), vector search, and citation graph analysis to help researchers efficiently discover and explore relevant academic literature. Unlike traditional keyword-based search engines, CiteConnect provides curated recommendations with explanations and visualizes connections through interactive citation graphs.
- Semantic Search: Advanced vector-based retrieval using SPECTER2 embeddings for finding relevant papers beyond keyword matching
- Citation Network Visualization: Interactive graph visualization showing connections between papers
- Personalized Recommendations: User profiles that adapt to individual research interests over time
- Explainable Results: LLM-generated summaries explaining why each paper is relevant
- Multi-Database Architecture: Integrated storage using PostgreSQL, Neo4j, Vector DB, and Redis
- Production-Ready MLOps: Complete CI/CD pipeline with monitoring, versioning, and automated deployment
- Abhinav Aditya
- Anusha Srinivasan
- Dennis Jose
- Dhiksha Mathanagopal
- Sahil Mohanty
- Size: 10,000-20,000 papers initially (scalable to 100,000+)
- Sources:
- Storage: ~2MB per PDF (20-40GB total for 20K papers)
- Format: JSON metadata, PDF papers, embeddings, citation graphs
Infrastructure & Deployment
- Google Cloud Platform (GCP)
- Kubernetes for container orchestration
- Docker for containerization
- Apache Airflow for pipeline orchestration
Data Storage
- PostgreSQL: Metadata storage
- Neo4j: Citation graph database
- Weaviate/FAISS: Vector embeddings
- Redis: Caching layer
- Google Cloud Storage: Raw PDF storage
ML/AI Components
- SPECTER2: Academic paper embeddings
- OpenAI API: Summarization and explanation generation
- sentence-transformers: Backup embedding models
Monitoring & MLOps
- Prometheus: Metrics collection
- Grafana: Visualization dashboards
- MLflow: Experiment tracking and model versioning
- DVC: Data version control
- Cloud Logging: Centralized logging
- Recall@10: ≥ 0.75
- Precision@10: ≥ 0.60
- Query Latency (p95): < 2 seconds
- System Availability: 99.5%
- Code Coverage: 85%+
- Click-Through Rate: ≥ 25%
- Return User Rate (7-day): ≥ 35%
- Average Session Duration: ≥ 10 minutes
# Required software
- Python 3.9+
- Docker & Docker Compose
- Google Cloud SDK
- Node.js (for frontend)# Clone the repository
git clone https://github.com/DhikshaMathanagopal/CiteConnect.git
cd CiteConnect
# Set up virtual environment
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt
# Set up environment variables
cp .env.example .env
# Edit .env with your API keys and configuration# Build and start services
docker-compose up -d
# Verify services are running
docker-compose ps