This project demonstrates a simple Retrieval Augmented Generation (RAG) pipeline using Spring AI, PGVector as the vector store, and OpenAI for embedding and chat models.

Or watch the tutorial on YouTube.

Components

Spring Boot Application: The main application that orchestrates the RAG pipeline.
PGVector: A PostgreSQL extension used as a vector database to store and retrieve document embeddings.
OpenAI Embedding Model: Used to convert text documents and user queries into numerical vector representations.
OpenAI Chat Model: Used to generate responses based on the user's query and retrieved relevant information.
DocumentLoader: A Spring component that loads sample documents into the PGVector store on application startup.
RagService: A service that handles the RAG logic:
- Takes a user query.
- Performs a similarity search in the PGVector store to find relevant documents.
- Constructs a prompt for the OpenAI chat model, incorporating the original query and the retrieved document content.
- Returns the generated response from the OpenAI chat model.
RagController: A REST controller that exposes an endpoint (/ai/rag) to interact with the RagService.
rag-prompt.st: A prompt template used by the RagService to guide the AI model's response, ensuring it uses the provided information.

How to Run

Prerequisites

Java 21: Ensure you have Java 21 installed.
Gradle: This project uses Gradle for dependency management and building.
PostgreSQL with PGVector Extension:
- Install PostgreSQL.
- Install the pgvector extension. You might need to compile it from source or use a pre-built image (e.g., pgvector/pgvector).
- Create a database (e.g., rag_demo). CREATE DATABASE rag_demo;
- Enable the vector extension in your database (after connecting to your database \c rag_demo):
```
CREATE EXTENSION IF NOT EXISTS vector;
CREATE EXTENSION IF NOT EXISTS "uuid-ossp";
CREATE TABLE IF NOT EXISTS vector_store (
    id uuid DEFAULT uuid_generate_v4() PRIMARY KEY,
    content text,
    embedding vector(768)
);
CREATE INDEX ON vector_store USING HNSW (embedding vector_cosine_ops);
```
OpenAI API Key: Obtain an API key from OpenAI.

Configuration

application.properties:
- Update src/main/resources/application.properties with your PostgreSQL connection details if they differ from the defaults.
- Set your OpenAI API key as an environment variable named OPENAI_API_KEY. Alternatively, you can directly paste it into application.properties (not recommended for production):
```
spring.ai.openai.api-key=YOUR_OPENAI_API_KEY
```

Build and Run

Build the application:
```
./gradlew clean build
```
Run the application:
```
OPENROUTER_API_KEY=YOUR_OPENROUTER_API_KEY ./gradlew bootRun
```
The DocumentLoader will automatically load sample documents into your PGVector database on startup.

Test the RAG Endpoint

Once the application is running, you can access the RAG endpoint:

curl "http://localhost:8080/ai/rag?message=What is RAG?"

You should receive a response generated by the OpenAI chat model, augmented with information retrieved from your PGVector store.