How sai0499/diffusion-page-reranking-aided-dense-retrieval-for-rag- works
Project type and purpose
• This is a full-stack RAG (Retrieval-Augmented Generation) application.
• Purpose: Provide a chat interface (React) that interacts with a Node.js REST API, which in turn calls a Python Flask service for document retrieval (via Milvus) and LLM calls (Google’s Gemini API + optional SERP API).
File structure
• Top level:
• public/, src/ → React front-end
• server/ → Node.js REST API
• Python-Server/ → Flask-based retrieval & LLM microservice
• SyntheticDatasetforIntentClassifier/ → offline data‐generation scripts
• server/: MVC-style folders (models/, controllers/, routes/, middleware/), config/db.js for MongoDB, utils for email/encryption.
• Python-Server/: Flask entrypoint (app.py), intent-classification pipelines, retrieval modules, gemini_api.py, decryption utilities, embedding ingestion.
• Config files: package.json/npm lock for Node projects, requirements implied by README, .env for environment variables.
README for additional insights
• Tech stack: React, Node.js+Express, Python+Flask, MongoDB, Milvus, Google Generative AI, SERP API.
• Ports & interaction flow: React (3000) → Node API (4000) → Flask service (5000).
• External dependencies: MongoDB, Milvus vector database, email SMTP, Google API.