Spring AI Example

Project Overview

This project demonstrates various implementation patterns and best practices for using Spring AI tools. It consists of two main modules:

- mcp-server: Implements the Model Context Protocol (MCP) server with both WebFlux and WebMvc SSE support
- proposal-agent: Implements the MCP client for making AI-powered proposals

Project Structure

.
├── mcp-server/           # MCP Server implementation
│   ├── src/             # Server source code
│   └── README.md        # Server documentation
├── proposal-agent/      # MCP Client implementation
│   ├── src/             # Client source code
│   └── README.md        # Client documentation
└── src/                 # Common source code

Tools Implementation Patterns

Methods as Tools

Spring AI supports using methods as tools by annotating them with @Tool. Example from DateTimeTools:

@Tool(name = "getCurrentDateTime", description = "Get the current date and time")
public String getCurrentDateTime() {
    return LocalDateTime.now().atZone(LocaleContextHolder.getTimeZone().toZoneId()).toString();
}

Tool Result Converter

Custom result converters can be implemented to control how tool results are formatted. Example from CustomToolCallResultConverter:

@Tool(name = "getCustomer",
      description = "Retrieve customer information",
      resultConverter = CustomToolCallResultConverter.class)
public Customer getCustomer(String name, ToolContext context) {
    return new Customer(name, "example@email.com");
}

Tool Context

Spring AI provides a ToolContext parameter that can be injected into tool methods to access contextual information:

public Customer getCustomerByEmail(String email, ToolContext context) {
    log.info("Context: {}", context);
    return new Customer("Demo", email);
}

Tool Parameters

Tool parameters can be annotated with @ToolParam to provide descriptions:

@Tool(name = "setAlarm")
public void setAlarm(@ToolParam(description = "Time in ISO-8601 format") String time) {
    // Implementation
}

Configuration

The project uses Spring Boot with the following key configurations:

- Ollama AI model integration
- Vector store with PGVector
- H2 database for development
- CORS configuration for web access
- MCP Server implementation
- MCP Client implementation
- SSE implementation

SSE implementation

an implementation of SSE (Server-Sent Events) for real-time updates. This is achieved by using the SseEmitter class.

Technology Stack

- Spring Boot
- Spring AI
- Ollama AI Model
- PGVector Vector Store
- H2 Database
- CORS Configuration
- SSE (Server-Sent Events) for real-time updates