🚀 Model Integration Protocol (MIP) with Spring Boot

This project demonstrates how to integrate Spring Boot methods into LLMs using Model Integration Protocol (MIP) by exposing them as JSON-RPC tools.
With @Agent and @Action annotations, any existing Spring Boot service becomes an AI-triggerable tool.

---

Screenshots

Connect to a server and start chatting with the AI:

View all the available tools and interact with them:

Click on tool to see the details

Click on the Send button to do normal chat with AI

Click on Tool button to get the closed matching tool for your prompt

Execute the tool and get the result

📦 Prerequisites

Tools4AI is here Tools4AI
Server is here Neurocaster-Server
Client is here Neurocaster-Client

🛠️ Features

• 🔥 Spring Boot + MIP Integration: Convert existing methods into JSON-RPC tools.
• ⚙️ Annotations:
  • @Agent: Declares a service as an AI-invocable agent.
  • @Action: Exposes a method as an actionable tool.
  • @Prompt: Defines custom prompts, like date formats, for better LLM interpretation.
• ⚡ Tools4AI Execution: LLMs use Tools4AI to trigger the exposed services dynamically.
• 🗨️ WebSocket Integration: All Spring services annotated with @Agent and @Action can now be used for chat interactions with LLMs.

---

⚙️ WebSocket Configuration

To integrate your Spring Boot application for chat interactions, configure WebSocket and the dependency as follows:

🧩 WebSocket Configuration Add the following configuration class to enable WebSocket support:

@Configuration
@EnableWebSocket
public class NeuroCasterWebSocketConfig implements WebSocketConfigurer {
    @Override
    public void registerWebSocketHandlers(WebSocketHandlerRegistry registry) {
        registry.addHandler(webSocketHandler(), "/chat")
                .addInterceptors(new HttpSessionHandshakeInterceptor())
                .setAllowedOrigins("*");  // Allow all origins or specific ones
    }

    @Bean
    public WebSocketHandler webSocketHandler() {
        return new NeuroCasterChatEndpoint(); // Use the ChatEndpoint class as WebSocketHandler
    }

    @Bean
    public ServerEndpointExporter serverEndpointExporter() {
        return new ServerEndpointExporter();
    }
}

📦 Add Dependency Add the following dependency to your pom.xml:

<dependency>
    <groupId>io.github.vishalmysore</groupId>
    <artifactId>neurocaster-server</artifactId>
    <version>0.0.1</version>
</dependency>

Then Connect to http://localhost:8081/actuator/tools4ai-tools using the neurocaster-client you can start chatting with the AI or you can use the tools

🔥 Example Services

🛠️ Customer Support Service

@Service
@Agent(groupName = "customer support", groupDescription = "Actions related to customer support")
public class ComplexActionService {

    public static int COUNTER = 0;

    public ComplexActionService() {
        COUNTER++;
    }

    @Action(description = "Customer has a problem, create a ticket for them")
    public String computerRepair(Customer customer) {
        return customer.toString();
    }
}

{
    "methodName": "computerRepair",
    "parameters": [
        {
            "name": "customer",
            "type": "Customer",
            "fieldValue": {
                "name": "Alice",
                "email": "alice@example.com",
                "issue": "Laptop not booting"
            }
        }
    ],
    "returnType": "String"
}

⚙️ Tools4AI Integration

To integrate with LLMs using Tools4AI, the services are exposed as JSON-RPC tools. When the LLM receives a query, it:

Selects the most relevant tool based on the context.

Generates the JSON-RPC request.

Invokes the Spring Boot service dynamically.

Returns the result back to the LLM.

🔥 Customization

You can:

Add more services with @Agent and @Action annotations.

Use @Prompt to specify custom prompts for date formatting or other field interpretation.

Scale by adding additional JSON-RPC tools, making your Spring Boot app LLM-compatible.

🚀 Contributing

Fork the repo and submit a pull request.

Add more services, improve error handling, or optimize the JSON-RPC layer.