DocGraph — GraphRAG System

General-purpose GraphRAG system integrating Neo4j knowledge graphs with FAISS vector search.

Stack

• Backend: FastAPI + Neo4j + FAISS + Groq (Llama 3)
• Frontend: React + Vite + TypeScript + Tailwind CSS
• MLOps: Docker + GitHub Actions CI/CD + Prometheus + Grafana
• Deployment: AWS EC2

Features

• LLM-based entity & relationship extraction into Neo4j
• Multi-hop Cypher graph traversal
• FAISS vector search with sentence-transformers
• Real-time SSE streaming
• Force-directed interactive knowledge graph visualization
• System reset endpoint
• Optional bearer-token auth for ingest/query/reset endpoints
• Confidence + citation metadata in query responses
• Regression evaluation harness (eval/run_eval.py) with CI gating