Skip to main content
logo
LiveWorkshop project — Google + MongoDB

GCP + MongoDB RAG chatbot — live in Whitelabel

RAG chatbot — built during a Google + MongoDB workshop, now live as part of the Whitelabel product.

GCP RAG chatbot architecture overview

About

A retrieval-augmented chatbot originally built during a Google + MongoDB workshop, then productized inside the Whitelabel project where it now runs in production. React + Material UI on the front, Express on the back, Google Vertex AI for embeddings and generation, and MongoDB Atlas Vector Search holding 768-dim embeddings of ingested PDFs.

What's interesting

  • PDF ingestion → sentence chunking → 768-dim embeddings
  • Vector search with score-based filtering
  • Toggleable RAG mode for side-by-side comparison
  • Server-side API key isolation

Stack

React 18Material UIExpress.jsMongoDB Atlas (Vector Search)Google Cloud Vertex AINode.js