RagWiser

RagWiser is a Retrieval Augmented Generation (RAG) system built with Spring Boot that enables users to upload PDF documents, process them, and ask questions about their content using natural language.

Project Overview

RagWiser uses Spring AI and PGVector to create an advanced document question-answering system. It processes PDF documents, stores their vectorized representation in a PostgreSQL database with pgvector extension, and answers user queries by retrieving relevant context and generating responses using OpenAI's GPT models.

Features

- PDF Document Upload: Upload and process PDF documents through a REST API
- Document Vectorization: Automatically extracts text from PDFs, splits it into chunks, and stores embeddings
- Semantic Search: Query documents using natural language
- RAG-powered Response Generation: Get accurate answers based on the content of your documents
- Spring AI Integration: Leverages Spring AI for vector stores and LLM integration
- Docker Support: Containerized PostgreSQL with pgvector extension

Technology Stack

- Java 21
- Spring Boot 3.3.2
- Spring AI 1.0.0-M1
- PostgreSQL with pgvector extension
- Docker
- OpenAI GPT-4

Getting Started

Prerequisites

- Java Development Kit (JDK) 21
- Docker and Docker Compose
- OpenAI API Key

Setup and Installation

1. Clone the repository:

   git clone https://github.com/yourusername/RagWiser.git

   cd RagWiser

   

2. Configure your OpenAI API key in src/main/resources/application.yaml:

   spring:

     ai:

       openai:

         api-key: YOUR_OPENAI_API_KEY

   

3. Start the PostgreSQL database with pgvector:

   docker-compose up -d

   

4. Build and run the application:

   ./mvnw spring-boot:run

   

API Endpoints

Upload a PDF Document

POST /api/rag/upload
Content-Type: multipart/form-data

Ask a Question

GET /api/rag?question=YOUR_QUESTION_HERE

How It Works

1. Document Processing:
- PDF documents are uploaded via the /api/rag/upload endpoint
- The application uses PagePdfDocumentReader to extract text from PDFs
- Text is split into chunks using TokenTextSplitter
- Text chunks are embedded and stored in the vector database

2. Question Answering:
- User submits a question via the /api/question endpoint
- The system retrieves the most relevant document chunks using vector similarity search
- A prompt template combines the question and retrieved documents
- OpenAI's GPT model generates an answer based on the context

3. MCP Integration:
- The application also provides a Tool-based integration for RAG capabilities using Spring AI's Tool Callbacks
- This enables the RAG functionality to be used as a tool by other AI systems

Database Schema

The application uses a PostgreSQL database with the pgvector extension for storing document embeddings:

CREATE TABLE vector_store (
    id uuid DEFAULT uuid_generate_v4() PRIMARY KEY,
    content text,
    metadata json,
    embedding vector(1536)
);
CREATE INDEX ON vector_store USING HNSW (embedding vector_cosine_ops);

Configuration

Key configuration options in application.yaml:

spring:
  datasource:
    url: jdbc:postgresql://localhost:5432/rag_db
    username: postgres
    password: postgres
  ai:
    openai:
      api-key: YOUR_OPENAI_API_KEY
      chat:
        options:
          model: gpt-4
  vectorstore:
    pgvector:
      index-type: HNSW
      distance-type: COSINE_DISTANCE
      dimensions: 1536
  servlet:
    multipart:
      enabled: true
      max-file-size: 100MB
      max-request-size: 100MB

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgements

- Spring AI Team for their excellent framework
- PostgreSQL and pgvector for vector storage capabilities
- OpenAI for their powerful language models