AN MINH
← Back to Projects

GraphRAG / AI

Vietnam History Chatbot based on Knowledge Graph (GraphRAG)

A GraphRAG system utilizing Neo4j, vector embeddings, and LLMs for accurate Vietnam history queries with response latency under 1.2 seconds.

Client
Personal Project
Role
AI Engineer
Stack
Python, React TypeScript, Neo4j, Redis, LLMs, Docker
Date
Apr 2026 — Jun 2026
Repository
123456789101112
Dataset
Preprocess
Graph
Embedding
Neo4j DB
User
Embedding
GenAI App
Cache
Prompt
LLM Model
Fig 1.0 — Vietnam History Chatbot GraphRAG system architecture

Technological & Algorithmic Highlights

Knowledge Graph Integration

Processed over 1,000,000 text chunks into 20,000+ entities and 50,000+ relationships using Neo4j to build a robust GraphRAG system.

High-Performance Query Pipeline

Developed a cached query pipeline achieving sub-1.2 second response latency, containerized and deployed using Docker Compose.

Operation Flow

  1. 01User submits a query related to Vietnam History.
  2. 02The query is embedded and matched against vector embeddings.
  3. 03Graph traversal retrieves related entities and relationships in Neo4j.
  4. 04Retrieved context is synthesized using LLMs.
  5. 05Response is cached in Redis for fast subsequent retrievals.

System Infrastructure

The system runs entirely in Docker containers, utilizing Neo4j for graph storage and Redis for caching to ensure high availability and low latency.

Infrastructure parameters

  • • Database: Neo4j
  • • Cache: Redis
  • • Backend: Python
  • • Frontend: React TypeScript
  • • Deployment: Docker Compose