Document Q&A System 🤖

An AI-powered document analysis platform with real-time chat and intelligent search capabilities.

✨ Features

• 📄 PDF Upload & Analysis - Upload PDFs and chat with your documents
• 🔍 Smart Search - WebSocket & Elasticsearch-powered fuzzy search with autocomplete
• 🔐 Privacy First - 100% local processing with Ollama (no external APIs)
• 🎯 Session Management - Multiple concurrent document conversations

🛠️ Technology Stack

Backend

• Spring Boot 3.5.7 + Java 21
• Spring AI (LLM integration)
• Elasticsearch 8.11 (search)
• MongoDB 7.0 (storage)
• WebSocket/STOMP (real-time)
• Apache PDFBox (PDF parsing)

Frontend

• React 18 + Vite
• STOMP.js + SockJS (WebSocket client)
• Custom design system

Infrastructure

• Docker Compose
• Ollama (Llama 3.1 8B model)
• Single-port deployment

📐 Architecture & Data Flow

┌─────────────────────────────────────────────────────────────────────┐
│                           USER BROWSER                              │
│                     React 18 + Vite Frontend                        │
│  ┌──────────────┐  ┌──────────────┐  ┌──────────────────────────┐   │
│  │ Session List │  │ Document     │  │ Chat Interface           │   │
│  │ (Search)     │  │ Upload       │  │ (WebSocket Streaming)    │   │
│  └──────────────┘  └──────────────┘  └──────────────────────────┘   │
└─────────────┬────────────┬────────────────────┬────────────────── ──┘
              │            │                    │
              │  WebSocket │ REST API           │ REST API
              │ (Search)   │ (Upload)           │ (Real-time Chat)
              │            │                    │
┌─────────────▼────────────▼────────────────────▼────────────────────┐
│                    SPRING BOOT APPLICATION                         │
│                      (Single Port: 8080)                           │
│  ┌──────────────────────────────────────────────────────────────┐  │
│  │                     Controllers Layer                        │  │
│  │  • WebSocketController  • DocumentController • ChatController│  │
│  └────────────┬─────────────────────┬──────────────────┬────────┘  │
│               │                     │                  │           │
│  ┌────────────▼─────────────────────▼──────────────────▼─────────┐ │
│  │                      Service Layer                            │ │
│  │  • ChatService                                                │ │
│  │  • DocumentService                                            │ │    
│  │  • ElasticsearchSearchService                                 │ │
│  │                                                               │ │
│  └────────┬──────────────┬──────────────┬─────────────┬──────────┘ │
│           │              │              │             │            │
│  ┌────────▼────┐  ┌──────▼──────┐ ┌────▼─────┐ ┌────▼──────────┐   │
│  │  AI         │  │   PDFBox    │ │ MongoDB  │ │ Elasticsearch │   │
│  │ Integration │  │  Text       │ │ Repo     │ │    Client     │   │
│  └────────┬────┘  │  Extraction │ └────┬─────┘ └────┬──────────┘   │
│           │       └─────────────┘      │            │              │
└───────────┼────────────────────────────┼────────────┼──────────────┘
            │                            │            │
    ┌───────▼────────┐         ┌─────────▼──────┐ ┌──▼─────────────┐
    │  OLLAMA LLM    │         │    MONGODB     │ │ ELASTICSEARCH  │
    │  (Llama 3.1)   │         │                │ │                │
    └────────────────┘         └────────────────┘ └────────────────┘

🔄 Data Flow Scenarios

1. Document Upload Flow:

User → Upload PDF → DocumentController → PDFBox (Extract) 
     → Save to MongoDB → Create Session → Sync to Elasticsearch 
     → Return Session ID

2. Chat Message Flow (WebSocket):

User → Send Message → WebSocketController → ChatService 
     → Retrieve Context from MongoDB → Spring AI → Ollama LLM
     → Stream Tokens → WebSocket → User (Real-time Display)
     → Save to MongoDB → Sync to Elasticsearch

3. Search Flow (Real-time):

User → Type Query → WebSocketController → ElasticsearchSearchService
     → Fuzzy Search in Elasticsearch → Return Results → WebSocket
     → Display Matches (Live Update)

4. Session Retrieval Flow:

User → Select Session → REST API → ChatService 
     → Fetch from MongoDB → Return History → Display Chat

🚀 Quick Start

Prerequisites

• 8GB RAM minimum (for Ollama LLM)
• Docker Desktop or Podman
• Maven 4.x+
• Java 21+
• IntelliJ IDEA or VSCode (optional)

One-Command Setup

docker-compose up -d

Wait 1-2 minutes for services to start and Ollama model to download.

mvn clean install

4. Build & Run Application

spring-boot:run 
(or)
java -jar target/PDFChatBot.jar

Then open: http://localhost:8080

📖 How It Works

1. Upload PDF → System extracts text and creates a chat session
2. Ask Questions → LLM analyzes document context and responds
3. Real-Time Streaming → Responses stream token-by-token like ChatGPT
4. Search Sessions → Find past conversations with fuzzy search
5. Continue Conversations → Resume any chat session

📦 Project Structure

document-summary/
├── src/main/java/com/docqa/
│   ├── config/         # Spring & Elasticsearch config
│   ├── controller/     # REST & WebSocket endpoints
│   ├── service/        # Business logic & LLM integration
│   ├── repository/     # MongoDB repositories
│   └── model/          # Domain entities
├── frontend/src/
│   ├── components/     # React components
│   ├── api/           # API client
│   └── design-system.css  # UI styles
└── docker-compose.yml  # Infrastructure setup

🔍 Architecture Highlights

• Single-Port Deployment - Frontend served from Spring Boot
• Async Streaming - Non-blocking WebSocket responses
• Search Indexing - Auto-sync MongoDB → Elasticsearch
• Session Isolation - Each document gets separate context

📄 License

MIT License - Feel free to use for personal or commercial projects!

🤝 Contributing

Contributions welcome! Please open an issue or PR.

---

Built with ❤️ using Spring Boot • React • Ollama • Elasticsearch • MongoDB