melukootto/raggin repository overview

⁠🧠 RAGGIN: RAG-Powered Assistant for Next.js Developers

⁠✨ Overview

RAGGIN is an open-source, Docker-deployable system that enhances the web development experience within Visual Studio Code using Large Language Models (LLMs) and RAG techniques. Specifically designed for the fast-evolving Next.js framework, it helps developers with:

• 🔍 Semantic documentation search
• ✍️ Intelligent code suggestions
• 🧠 Context-aware question answering
• 💡 Framework migration and version comparison assistance

⁠⚙️ Architecture

The system consists of two core components:

⁠📥 Retriever

• Uses BGE-M3 embeddings to encode content from versioned Next.js documentation.
• Stores vectors in Milvus, an open-source vector database supporting hybrid search with both dense and sparse retrieval.
• Returns the most contextually relevant documentation chunks in response to developer queries.

⁠🧾 Generator

• Lets users select their preferred local LLM via Ollama⁠ or compatible LangChain-backed runtimes.
• Supports multiple local models (e.g., Qwen, LLaMA, Mistral, etc.) based on user needs and available resources.
• Generates actionable suggestions and detailed answers by combining user queries with retrieved docs.
• Fully integrated into VSCode for seamless inline interaction and minimal context switching.

⁠📦 What's in the Docker Container?

• milvus-standalone vector DB with HNSW + inverted indexing
• Pre-processed Next.js documentation (v13.0.0–15.x.x) with semantic embeddings
• REST API powered by FastAPI and LangChain
• Ready-to-integrate backend for VSCode extension frontend
• No cloud dependencies — runs fully offline

⁠🚀 Getting Started

docker pull melukootto/raggin

Run with Docker Compose:

docker compose -f docker-compose.yml up

You can then install the companion RAGGIN⁠.

⁠📂 Open Dataset

All versioned Next.js documentation used in this project is:

• 📚 Publicly accessible on Kaggle⁠
• 🔄 Regularly updated
• 🧪 Structured and pre-embedded for semantic search

⁠📍 Availability

• 🔗 GitHub: https://github.com/silam741852963/RAGGIN⁠
• 🧩 VSCode Marketplace: RAGGIN⁠
• 🐳 Docker Hub: melukootto/raggin⁠
• 📊 Kaggle Dataset: Next.js Documentation for RAGGIN⁠

⁠🛠️ Built With

• Milvus⁠ – Vector DB
• BGE-M3⁠ – Embedding model
• LangChain⁠ – RAG orchestration
• Ollama⁠ – Local LLM launcher
• VS Code⁠ – IDE integration
• Next.js⁠ – Target framework

⁠🎓 Project Info

This project is the result of a graduation thesis at Ho Chi Minh City University of Technology, under the supervision of Dr. Nguyen An Khuong and Mr. Pham Nhut Huy (Zalo AI).

⁠📄 License

MIT License
Use freely in research, commercial, or educational settings. Contributions welcome!