🧠 RAG Knowledge Base App

A full-stack Retrieval-Augmented Generation (RAG) application that lets you upload documents and ask AI-powered questions about their content. Claude answers using only the information in your uploaded documents — no hallucination.

Built with React, Fastify, AWS Bedrock (Claude), AWS S3, PostgreSQL, and Docker.

---

✨ Key Features

• Document Upload — Drag-and-drop files (TXT, PDF, MD, CSV) stored in AWS S3 with automatic text extraction
• RAG-Powered Q&A — Ask natural language questions answered exclusively from your document content
• Source Attribution — Every answer references which documents were used
• Query History — Browse past questions and answers with timestamps
• Feedback System — Thumbs up/down on answers for quality tracking
• Session Management — Independent sessions with separate document sets
• Fully Dockerized — One command starts React, Fastify, and PostgreSQL together

---

🏗️ Architecture

┌──────────────────────────────────────────────────────────┐
│                     React Frontend                        │
│  ┌──────────┐    ┌──────────────┐    ┌────────────────┐  │
│  │  Upload   │    │   Chat Q&A   │    │    History     │  │
│  │  View     │    │   View       │    │    View        │  │
│  └────┬─────┘    └──────┬───────┘    └───────┬────────┘  │
│       │                 │                     │           │
│       └─────────────────┼─────────────────────┘           │
│                         │ axios                           │
├─────────────────────────┼────────────────────────────────┤
│                   Nginx (proxy)                           │
│              /api/* → backend:3001                        │
├─────────────────────────┼────────────────────────────────┤
│                  Fastify Backend                          │
│                         │                                 │
│  ┌──────────┐  ┌───────┴────────┐  ┌──────────────────┐  │
│  │ POST     │  │ POST           │  │ GET              │  │
│  │ /upload  │  │ /query (RAG)   │  │ /history         │  │
│  └────┬─────┘  └───────┬────────┘  └────────┬─────────┘  │
│       │                │                     │            │
│  ┌────┴──┐     ┌───────┴────────┐    ┌───────┴────────┐  │
│  │ AWS   │     │  RAG Engine    │    │  PostgreSQL    │  │
│  │ S3    │     │  Retrieve →    │    │  Sessions,     │  │
│  │ Store │     │  Augment →     │    │  Documents,    │  │
│  │       │     │  Generate      │    │  Queries       │  │
│  └───────┘     │  (Claude)      │    └────────────────┘  │
│                └────────────────┘                         │
└──────────────────────────────────────────────────────────┘

---

🔄 The RAG Pipeline

User: "What is the return policy for electronics?"
  │
  ▼
1. RETRIEVE — Fetch all document texts from PostgreSQL for this session
  │
  ▼
2. AUGMENT — Build a prompt with document content as context:
  │  "Here are the documents: [full text]... Question: What is the return policy?"
  │
  ▼
3. GENERATE — Send to Claude via AWS Bedrock (temperature=0.0 for factual answers)
  │  Claude reads the documents and generates a grounded answer
  │
  ▼
4. RETURN — "According to the Company Policy Document, electronics have a
             15-day return window. Items must be in original packaging..."

---

🚀 Quick Start

Prerequisites

• Node.js 18+
• Docker Desktop
• AWS Account (Bedrock + S3 access)

Option 1: Docker (Recommended)

git clone https://github.com/jaymistry98/rag-knowledge-base.git
cd rag-knowledge-base

# Create .env with your AWS credentials
cp .env.example .env
# Edit .env with your values

# Start everything (React + Fastify + PostgreSQL)
docker compose up --build

# Open http://localhost:3000

Option 2: Development Mode

# Terminal 1 — PostgreSQL
docker run --name postgres-rag -e POSTGRES_PASSWORD=password -e POSTGRES_DB=ragapp -p 5432:5432 -d postgres:16

# Terminal 2 — Backend
node server/server.js

# Terminal 3 — Frontend
cd client && npm run dev

# Open http://localhost:5173

---

📡 API Endpoints

Method	Endpoint	Description
POST	/api/upload	Upload a document (multipart form-data)
POST	/api/query	Ask a question (RAG pipeline)
GET	/api/history/:sessionId	Get query history
GET	/api/documents/:sessionId	List uploaded documents
POST	/api/feedback	Submit thumbs up/down
GET	/health	Health check

Example: Ask a Question

Request:

POST /api/query
{
  "sessionId": "session_123",
  "question": "What is the shipping policy?"
}

Response:

{
  "success": true,
  "queryId": 1,
  "answer": "According to the Company Policy Document, all orders ship within 2 business days...",
  "documentsUsed": ["company-policy.txt"],
  "tokenUsage": { "inputTokens": 850, "outputTokens": 120 }
}

---

📁 Project Structure

rag-knowledge-base/
├── server/
│   ├── config.js          # Environment configuration
│   ├── db.js              # PostgreSQL connection & queries
│   ├── s3Client.js        # AWS S3 upload/download/text extraction
│   ├── ragEngine.js       # RAG pipeline (retrieve → augment → generate)
│   ├── server.js          # Fastify entry point
│   ├── routes/
│   │   ├── upload.js      # File upload endpoint
│   │   ├── query.js       # RAG query endpoint
│   │   └── history.js     # History & feedback endpoints
│   ├── Dockerfile         # Backend container
│   └── package.json       # Backend dependencies
├── client/
│   ├── src/
│   │   ├── App.jsx        # Main app with tab navigation
│   │   ├── App.css        # Styles
│   │   ├── api.js         # Centralized API client
│   │   ├── main.jsx       # React entry point
│   │   └── components/
│   │       ├── UploadView.jsx   # Document upload with drag-and-drop
│   │       ├── ChatView.jsx     # Q&A chat interface
│   │       └── HistoryView.jsx  # Query history with feedback
│   ├── nginx.conf         # Nginx reverse proxy config
│   ├── Dockerfile         # Frontend multi-stage build
│   └── package.json       # Frontend dependencies
├── docker-compose.yml     # 3-service orchestration
├── .env                   # Environment variables (not committed)
└── README.md

---

🛠️ Technologies

Technology	Purpose
React	Frontend UI (upload, chat, history views)
Vite	Frontend build tool
Fastify	Backend REST API framework
AWS Bedrock (Claude)	AI answer generation
AWS S3	Document storage
PostgreSQL	Session, document metadata, query history storage
Nginx	Static file serving + API reverse proxy
Docker Compose	Multi-container orchestration
Axios	HTTP client for frontend-backend communication

---

🧪 Database Schema

-- Sessions: Track user conversations
sessions (id, created_at, last_active)

-- Documents: Metadata for uploaded files (content stored in S3)
documents (id, session_id, original_name, s3_key, file_size, content_type, text_content, uploaded_at)

-- Queries: Question/answer history with feedback
queries (id, session_id, question, answer, documents_used, feedback, created_at)

---

📝 What I Learned

• RAG Pattern: Building the retrieve → augment → generate pipeline from scratch, understanding why grounded answers matter for enterprise AI
• Full-Stack AI Architecture: Connecting a React frontend to an AI-powered backend with persistent storage
• AWS S3 Integration: Programmatic file upload/download with the AWS SDK v3
• PostgreSQL: Relational database design, parameterized queries, and connection pooling
• Docker Multi-Stage Builds: Optimizing container images from ~500MB to ~25MB
• Nginx Reverse Proxy: Routing API traffic and serving static files in production
• Multi-Container Orchestration: Docker Compose with health checks and service dependencies