📄 AI Document Q&A (RAG System)

An AI-powered document question-answering system built with Node.js, Next.js, and OpenAI. Upload PDFs → system extracts + embeds content → stores vectors in Postgres (pgvector) → answers queries with GPT grounded in document context.

✨ Features

• 📂 Upload PDFs via frontend
• 🔄 Background indexing with BullMQ + Redis
• 🧩 Text chunking & OpenAI embeddings
• 🗄 Vector search using Postgres + pgvector
• 🤖 Context-aware answers with GPT-4o-mini
• 🌐 Next.js + Material UI frontend for a polished UI
• 🐳 Dockerized for easy deployment

🚀 Getting Started

1. Clone repo

git clone https://github.com/abhicode/document-workflow-assistant.git
cd document-workflow-assistant

2. Environment setup

Create .env files at the root:

# Redis
REDIS_URL=redis://redis:6379

# OpenAI
OPENAI_API_KEY=sk-proj-***
EMBEDDING_MODEL=text-embedding-3-small
EMBEDDING_DIM=1536
TOP_K=5

# Server
PORT=8080
FRONTEND_URL=http://localhost:3000

POSTGRES_USER=postgres
POSTGRES_PASSWORD=postgres
POSTGRES_DB=vector_db

DATABASE_URL=postgres://postgres:postgres@db:5432/vector_db

3. Run with Docker Compose

docker-compose up --build

Frontend → http://localhost:3000

Backend API → http://localhost:8080

🖥️ Usage

• Upload a PDF from the frontend.
• Wait for it to be indexed (processed by a worker).
• Enter a question → system retrieves relevant chunks → GPT answers with context + sources.

🛠️ Tech Stack

• Backend: Node.js, Express, BullMQ, Redis
• Database: PostgreSQL + pgvector
• Workers: PDF parsing, OpenAI embeddings
• Frontend: Next.js (TypeScript) + Material UI
• AI Models: OpenAI embeddings + GPT-4o-mini

📦 Deployment

• Local: Docker Compose (frontend + backend + redis + postgres)
• Cloud: Easily deployable to AWS, GCP, Azure, Railway, or Render