My RAG Development Journey: From Exploration to Graph RAG

January 09, 2026

RAG Graph RAG Web dev Docker DevLog

This is a journey of how I evolved from an LLM (Large Language Model) enthusiast to a RAG system developer. From understanding the core principles to implementing cloud-native microservices, every step represents my growth and thinking in technology.

1. The Beginning: First Taste of RAG

I discovered RAG (Retrieval-Augmented Generation) on GitHub and was instantly fascinated by its ability to mitigate LLM hallucinations. This marked the start of my learning journey, where I built my first experimental repo: RAG-explore. At this stage, I focused on understanding Vector Embeddings and the basic logic of retrieval.

2. Practice: Stepping into Web Development

As I acquired web development skills at school, I began thinking about how to transform RAG into a real-world application. I developed ChatYourNotes, marking my first attempt to bring a RAG system to a web interface with local network deployment. This taught me that great technology needs a good user interface to truly shine.

3. Advancement: Integrating Stack and Architecture

Next, I expanded my tech stack to pursue a more professional architecture:

  • Frontend: Built a modern UI with React.
  • Backend: Handled logic with Flask.
  • Data Storage: Integrated ChromaDB for vectors and MySQL for metadata.
  • Engineering: Leveraged Docker Compose to orchestrate full-stack services.
  • Data Processing: Introduced OCR to parse text directly from PDF files. The goal of this phase was to build a stable, end-to-end RAG application.

4. Breakthrough: Embracing Graph RAG and Cloud-Native

While exploring the open-source community, I identified a major pain point in traditional RAG: context fragmentation caused by chunking. Graph RAG, through “Entity-Relationship” modeling, provides a more robust solution for preserving complex relationships.

Consequently, I launched a new project: GraphWeaver. This project is a culmination of my recent learning:

  • High-Performance Backend: Re-architected using Golang (Gin).
  • Cloud-Native Practice: Integrated with Kubernetes (K8s) for microservice management. (planning, not yet completed, currently running with Docker Compose)
  • Automation: Explored GitHub Actions for CI/CD pipelines.

5. Future: Continuous Growth in the Open Source Community

The path of technology is endless. Moving forward, I will dedicate more time to researching outstanding open-source projects such as:

  • DeepTutor: Its powerful system integration aligns closely with my project goals.
  • LightRAG & KAG: Exploring lighter yet higher-level knowledge-oriented architectures.
  • Agentic RAG: Studying how Agents can intelligently utilize RAG tools.

Beyond code, I’ve started actively participating in community discussions. Interacting with the open-source community has been incredibly rewarding and continues to motivate me to refine my projects.

Post recorded on January 9, 2026. The development continues…

← Back to Home