Spring Boot RAG SQL Assistant

This Spring Boot application implements an advanced Retrieval-Augmented Generation (RAG) system that connects language models with SQL database retrieval using LangChain4j.

Features

• Natural Language to SQL: Ask questions in plain English and get database insights
• Multiple LLM Support: Works with Google Gemini, Ollama, or LocalAI
• Interactive Interface: Simple web UI for asking questions
• Spring Boot Integration: Leverages Spring's dependency injection and configuration

Prerequisites

• Java 17 or higher
• Maven
• MariaDB database (or adjust configurations for your preferred database)
• API keys for language models (Google AI Gemini, etc.)

Getting Started

Clone the Repository

git clone https://github.com/DIGI-UW/rag.git
cd rag

Sign in to Google Studio to get an API key for Gemini and add it to the env file

Quick start with docker

Run App with Built docker images

docker compose up --build

Run App with published docker images

docker compose up

The application will be available at http://localhost:8080

Configuration

Set up the following environment variables or modify application.properties:

# Database configuration
DB_URL=jdbc:mariadb://localhost:3306/your_database
DB_USER=your_username
DB_PASSWORD=your_password

# Language model API keys
GEMINI_API_KEY=your_gemini_api_key

# Ollama configuration (if using)
OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_MODEL_NAME=llama3

# LocalAI configuration (if using)
LOCAL_AI_BASE_URL=http://localhost:8080/v1
LOCAL_AI_MODEL_NAME=gpt-4

Build and Run

# Build the application
mvn clean install

# Run the application
mvn spring-boot:run

The application will be available at http://localhost:8080

Usage

1. Open your browser and navigate to http://localhost:8080
2. Type your database question in the chat interface
3. Receive answers based on your database content

Example questions:

• "How many patients do we have?"
• "What are the names of the patients?"
• "what was the most prescribed drug?"

Architecture

• Controller Layer: Handles HTTP requests and responses
• Service Layer: Coordinates between components
• Content Retriever: Translates questions to SQL and retrieves data
• LLM Integration: Uses language models to generate SQL and format responses

Security Warning

The SqlDatabaseContentRetriever is potentially dangerous and should be used with caution:

• READ-ONLY ACCESS: The database user should have very limited read-only permissions
• DEVELOPMENT USE: Not recommended for production without thorough security measures
• SQL VALIDATION: Basic validation is performed but cannot guarantee safety

Extending the Application

Adding More Language Models

Modify AppConfig.java to integrate additional language models:

@Bean
public ChatLanguageModel newModel() {
    return NewModelProvider.builder()
            .apiKey("your-api-key")
            .modelName("model-name")
            .build();
}

Customizing the UI

The frontend interface is located in src/main/resources/static/index.html. Modify this file to customize the user experience.

Acknowledgments

• Built with LangChain4j
• Powered by Spring Boot