Sources Conference Talk: From 0 to H-AI-ro: fast track to AI for Kotlin developers.

Preparations

• Install the Kotlin Notebooks plugin in IntelliJ
• You need an OpenAI API key in your environment: OPENAI_API_KEY=<your-key>

LangChain4J with Spring Boot in module: /langchain4j

Run LangChain4J sample application

• Provide the OpenAI key OPENAI_API_KEY in the environment
• Run application: LangChain4JDemoApplication.kt
• Run Samples in the notebook: langchain4j-demo.ipynb

Play with Embeddings and Similarity

• Run Samples in the notebook: langchain4j-similarity.ipynb

Data Ingestion

• Start docker-compose: docker-compose up
• Open the notebook: langchain4j-ingest.ipynb
• Run code snippets to ingest data

LangChain4J with ktor in module: /langchain4j-ktor

Run LangChain4J-ktor sample application

• Provide the OpenAI key OPENAI_API_KEY in the environment
• Run application: Application.kt
• Run Samples in the notebook: /langchain4j/langchain4j-demo.ipynb

SpringAI in module: /spring-ai

Run SpringAI sample application

• Provide the OpenAI key OPENAI_API_KEY in the environment
• Start docker-compose: docker-compose up
• Run application: SpringAIDemoApplication.kt
  • The first time it takes a while since the vector db is ingested using the spring-ai way
• Run Samples in the notebook: spring-ai-demo.ipynb

Run KMP chat / audio client

• Run KMP Chat Client main in Main.kt in the chatclient-kmp module

• Chat Feature:

  

• Voice Feature:

  

Kotlin MCP server + integration with Claude Desktop: /kotlin-mcp-server

• MCP Server that exposes:

  • prompt: similar to system prompt in Spring-AI but tailored for MCP capabilities (resources and tools)
  • resource: the menu as MCP resource
  • tools: tools to classify, find dishes via vector search and ordering dish, all forwarding to the Spring-AI rest endpoints

• Build the MCP Server with: ./gradlew jar

• Open the Claude Desktop application -> Settings -> Developer and press Edit Config to locate the claude_desktop_config.json file.

• Add the following configuration in claude_desktop_config.json:

  {
   "mcpServers": {
      "mcp-italian-delaight": {
       "command": "java",
       "args": ["-jar", "<project-path>/kotlin-mcp-server/build/libs/mcp-kt-jar-0.1.0.jar"]
      }
    } 
  }
  
  
  

• Restart Claude Desktop and manually select the italian-meal-agent prompt that the MCP server exposes:

• 

• Now you can order your favorite dish. Buon Appetito 🍕🇮🇹🍝!