RAG Backend - Hybrid Search with Node.js

This is a minimal Retrieval-Augmented Generation (RAG) backend that supports document ingestion and hybrid (semantic + keyword) search using Node.js and a local embedding model.

✨ Features

• Local embeddings with sentence-transformers (MiniLM)
• Hybrid search using:
  • BM25-style keyword frequency
  • Cosine similarity with embedding vectors
• Simple JSON-based document storage
• Clean, minimal Express API

🚀 Endpoints

POST /knowledgebase

Ingest a document (a text) into the knowledge base.

Request:

{
    "text": "Node.js is a JavaScript runtime built on Chrome's V8 engine."
}

Search in document (text)

POST /query

{
    "query": "What is Node.js?"
}

Run project

1. Go to embedding server direcotry

• Install dependencies

pip install Flask sentence-transformers

• Run the project

python embeddingServer.py

2. In other terminal, inside main direcotry

• Install dependencies

npm install

• Run project

node app.js