Spring Boot AI

What is Spring Boot AI?

Spring Boot AI is a sample application demonstrating how to build an AI agent using Spring AI, the Model Context Protocol (MCP), and Retrieval Augmented Generation (RAG). It shows how to create tools (Weather, Booking), integrate with AI models (Ollama, Gemini, Bedrock), and add chat memory and vector store context.

How to use Spring Boot AI?

Configure the application for your chosen AI provider (Ollama, Gemini, or Bedrock) via Spring profiles. Run the vector database (PGVector) and start the MCP server and chat server. Send requests to the REST endpoint POST /{chatId}/chat?question=... to interact with the agent.

Key features of Spring Boot AI

• Local and remote MCP tools for weather and booking
• RAG with PGVector vector store
• Chat memory via MessageChatMemoryAdvisor
• Support for Ollama, Gemini, and AWS Bedrock
• AI Model Evaluation for testing

Use cases of Spring Boot AI

• Building an AI-powered travel booking assistant
• Learning how to integrate MCP tools into a Spring Boot application
• Experimenting with different AI models and embedding providers

FAQ from Spring Boot AI

What AI models does Spring Boot AI support?

It supports Google Gemini, AWS Bedrock, and local Ollama models, switchable via Spring profiles.

How are tools implemented in Spring Boot AI?

Tools are created as Spring Beans annotated with @Tool and @ToolParam, and registered as MethodToolCallbackProvider.

Does Spring Boot AI use a vector store?

Yes, it uses PGVector for RAG – the vector store is loaded with sample city data at startup.

What transport does the MCP server use?

The remote Booking MCP server uses SSE transport, configured via spring.ai.mcp.client.sse.connections.booking-tool.url.

Can I test the AI agent?

Yes, the project includes AI Model Evaluation using Spring AI’s testing support.