Spring AI POC - Java AI Assistant

Objective

This project is a demonstration of integrating Spring AI with OpenAI to create a Java-specialized AI assistant. The chatbot is capable of:

• Answering questions about Java development
• Memorizing conversation context
• Accessing custom tools (date/time)
• Integrating MCP (Model Context Protocol) clients to extend functionality

---

Overview of Main Spring AI Classes

1. ChatClient

• Role: Main client for interacting with the AI model (OpenAI)
• Usage: Building prompts and calling the language model
• Configuration: Defined in the controller with advisors for memory management

2. MessageWindowChatMemory

• Role: Stores and manages conversation message history
• Configuration: Limited to a maximum of 10 messages to optimize API calls
• Benefit: Allows the model to maintain conversation context

3. MessageChatMemoryAdvisor

• Role: Advisor that automatically injects history into each model call
• Usage: Ensures that each message to the AI includes previous context

4. SystemPromptTemplate

• Role: Template to define system role and behavior
• Configuration: Defines the AI as a Java expert and enforces Markdown responses

5. SyncMcpToolCallbackProvider

• Role: Tool provider via MCP protocol
• Usage: Integrates external tools available via the MCP server
• Configuration: Connected to the server https://mcp.context7.com

6. @Tool Annotation

• Role: Decorator to expose Java methods as tools for the AI
• Example: DateTimeTools exposes getCurrentDateTime() to the AI

---

Architecture

ChatController
├── POST /api/v1/chat          → Simple chat with memory
├── POST /api/v1/chat/tools    → Chat with custom tools (DateTimeTools)
└── POST /api/v1/chat/mcp      → Chat with MCP tools + custom tools

---

Running the Application

Prerequisites

• Java 25+
• Maven
• OpenAI API key (defined via the API_KEY environment variable)

Configuration

Setting the API key:

export API_KEY="your-openai-key"

Startup

# With Maven
./mvnw spring-boot:run

# Or with the compiled JAR
java -jar target/SpringAIArticle-0.0.1-SNAPSHOT.jar

The application will start on http://localhost:8080

---

Possible Requests

1. Simple Chat (with memory)

curl -X POST http://localhost:8080/api/v1/chat \
  -H "Content-Type: application/json" \
  -d '"Hello, my name is Alexandre, and you re Bottest!"'

2. Chat with Custom Tools

curl -X POST http://localhost:8080/api/v1/chat/tools \
  -H "Content-Type: application/json" \
  -d '"What time is it now?"'

3. Chat with MCP Tools + Custom Tools

curl -X POST http://localhost:8080/api/v1/chat/mcp \
  -H "Content-Type: application/json" \
  -d '"Answer these two requests: - What are the latest Spring Boot versions? Find them in MCP Context7. - Get the time and display it."'

Technologies Used

• Framework: Spring Boot 4.0.3
• AI: Spring AI 2.0.0 + OpenAI
• Protocol: MCP (Model Context Protocol)
• Language: Java 25