Spring AI Docling

A demonstration project integrating Spring AI with Docling for intelligent PDF parsing and conversational AI capabilities. This project showcases how to build a modern RAG (Retrieval-Augmented Generation) application that can extract, process, and chat about PDF document content.

📖 Blog Post

For a detailed walkthrough and explanation of this project, read the accompanying blog post:

A Practical Walkthrough for Parsing PDFs and Chatting About Their Content

🎯 Features

• PDF Processing: Extract and parse PDF documents using Docling's MCP server
• Spring AI Integration: Leverage Spring AI's MCP client for seamless tool integration
• Conversational AI: Chat about document content using OpenAI models
• REST API: Simple HTTP endpoints for document interaction
• React Frontend: Modern UI built with React and Chakra UI
• Docker Support: Easy deployment with Docker Compose

🏗️ Architecture

The project consists of three main components:

1. Spring Boot Backend (Java 21)

   • Spring AI MCP client integration
   • REST API for document queries
   • OpenAI model integration

2. Docling MCP Server (Docker)

   • PDF parsing and conversion
   • Document indexing and retrieval
   • Exposed via MCP protocol

3. React Frontend

   • Chat interface for document interaction
   • Built with Chakra UI

🚀 Getting Started

Prerequisites

• Java 21+
• Maven 3.6+
• Node.js 16+ (for frontend)
• Docker and Docker Compose
• OpenAI API key

Setup

1. Clone the repository

   git clone <repository-url>
   cd spring-ai-docling

2. Set up OpenAI API key

   Create a src/main/resources/application.properties file:

   spring.ai.openai.api-key=${OPENAI_API_KEY}
   spring.ai.openai.chat.options.model=gpt-4

   Export your API key:

   export OPENAI_API_KEY=your-api-key-here

3. Start Docling MCP server

   docker-compose up -d

   The Docling server will be available at http://localhost:8000

4. Place PDF documents

   Add your PDF files to the documents/ directory. These will be automatically mounted into the Docker container.

5. Run the Spring Boot application

   ./mvnw spring-boot:run

   The backend API will start on http://localhost:8080

6. Run the frontend (optional)

   cd frontend
   npm install
   npm start

   The React app will open at http://localhost:3000

📝 Usage

Using the REST API

Send a POST request to /ai with your query:

curl -X POST http://localhost:8080/ai \
  -H "Content-Type: application/json" \
  -d '{
    "input": "Can you extract the contents of this pdf: /data/your-document.pdf"
  }'

Example Queries

• Extract document content: "Can you extract the contents of this pdf: /data/business-plan.pdf"
• Ask about document: "Can you translate the summary to Dutch for document <document-id>?"
• List available tools: "Can you tell me which tools are available for the MCP server docling?"

See test-calls.http for more examples.

🐍 Python Experimentation

A Python virtual environment is included for experimenting with Docling directly:

source .venv/bin/activate
python

from docling.document_converter import DocumentConverter

converter = DocumentConverter()
result = converter.convert("documents/your-document.pdf")
print(result.document.export_to_markdown())

🛠️ Technology Stack

• Backend: Spring Boot 3.5.8, Spring AI 1.1.0
• Frontend: React 18, Chakra UI
• AI/ML: OpenAI GPT-4, Docling
• Infrastructure: Docker, Docker Compose
• Build Tools: Maven, npm

📂 Project Structure

spring-ai-docling/
├── src/main/java/          # Spring Boot application
│   └── org/rag4j/docling/
├── frontend/               # React application
├── documents/              # PDF files (mounted in Docker)
├── .venv/                  # Python virtual environment
├── docker-compose.yml      # Docling server configuration
├── pom.xml                 # Maven dependencies
└── test-calls.http         # Example API calls

🤝 Contributing

Contributions are welcome! This is a demonstration project, so feel free to experiment and extend it.

📄 License

This project is provided as-is for educational and demonstration purposes.

🔗 Resources

• Spring AI Documentation
• Docling Documentation
• Model Context Protocol
• Blog Post: A Practical Walkthrough for Parsing PDFs and Chatting About Their Content