Koog Playground

This repository contains examples and utilities for working with the Koog AI agent framework.

Overview

Koog Playground is a project that demonstrates various capabilities of the Koog framework for building AI agents. It includes examples of:

• AI agents implementation
• Strategy patterns for AI interactions
• API integrations
• Tool sets for extending agent capabilities
• Utility functions

Koog Links

• https://github.com/JetBrains/koog
• https://docs.koog.ai/

Project Structure

Main Source Code

• src/main/kotlin/Main.kt - Main application entry point
• src/main/kotlin/QuickstartExample.kt - Quick start example for the Koog framework
• src/main/kotlin/me/bossm0n5t3r/LoggerExtension.kt - Logger utility extensions

Core Modules

• src/main/kotlin/me/bossm0n5t3r/agent/ - AI agent implementations
  • AIAgentSample.kt - Sample AI agent implementation
• src/main/kotlin/me/bossm0n5t3r/api/ - API integration examples
  • PromptApi.kt - Prompt API integration
  • SingleApi.kt - Single API call examples
• src/main/kotlin/me/bossm0n5t3r/di/ - Dependency injection configuration
  • AppModule.kt - Koin dependency injection module
• src/main/kotlin/me/bossm0n5t3r/mcp/ - Model Context Protocol implementations
  • GoogleMaps.kt - Google Maps integration for location-based queries
  • Playwright.kt - Playwright for browser automation tasks
• src/main/kotlin/me/bossm0n5t3r/opentelemetry/ - OpenTelemetry and observability
  • langfuse/ - Langfuse tracing integration
    • AgentWithLangfuseTracing.kt - Agent with Langfuse tracing capabilities
    • docker-compose.yml - Docker setup for Langfuse
• src/main/kotlin/me/bossm0n5t3r/strategies/ - Strategy patterns for AI interactions
  • CustomStrategyGraphs.kt - Custom strategy graph implementations
  • HistoryCompression.kt - History compression strategies
  • PredefinedStrategiesAndCommonStrategyPatterns.kt - Common strategy patterns
  • StructuredDataProcessing.kt - Structured data processing strategies
• src/main/kotlin/me/bossm0n5t3r/tools/ - Tool sets for extending agent capabilities
  • AgentWithWeatherToolSet.kt - Agent with weather tool integration
  • DiagnosticToolSet.kt - Diagnostic tools for debugging
  • MyToolSet.kt - Custom tool set implementation
  • WeatherToolSet.kt - Weather-related tools
• src/main/kotlin/me/bossm0n5t3r/utils/ - Utility functions
  • Helper.kt - Helper utilities

Test Code

• src/test/kotlin/me/bossm0n5t3r/api/ - API tests
  • PromptApiTest.kt - Tests for Prompt API

Getting Started

Prerequisites

• JDK 21 or higher
• Gradle

Setup

1. Clone the repository
2. Build the project:

   ./gradlew build
   

3. Run examples:

   ./gradlew run
   

Examples

Check out QuickstartExample.kt for a simple demonstration of how to use the Koog framework.

Running the Application

The application provides an interactive menu to run different examples:

1. Run the application:

   ./gradlew run
   

2. Select an example to run from the menu by entering the corresponding number:

   > Task :run
    00:00:00.249 [main] INFO Application -- Hello, Kotlin!
    00:00:00.311 [main] INFO Application -- Please select the example to run:
    00:00:00.314 [main] INFO Application -- 1. PromptApi
    00:00:00.314 [main] INFO Application -- 2. SingleApi
    00:00:00.314 [main] INFO Application -- 3. AIAgent
    00:00:00.314 [main] INFO Application -- 4. AgentWithWeatherToolSet
    00:00:00.314 [main] INFO Application -- 5. StructuredDataProcessing
    00:00:00.314 [main] INFO Application -- 6. Google Maps
    00:00:00.314 [main] INFO Application -- 7. Playwright
    00:00:00.314 [main] INFO Application -- 0. Exit
    Enter your choice:
    <==========---> 83%00:00:00.773 [main] INFO Application --
   

3. Follow the prompts for the selected example.

Each example demonstrates different capabilities of the Koog framework:

• PromptApi: Shows how to use the prompt API for generating responses
• SingleApi: Demonstrates single API calls to LLM models
• AIAgent: Basic AI agent implementation
• AgentWithWeatherToolSet: AI agent with weather-related tools
• StructuredDataProcessing: Processing structured data with AI
• Google Maps: Using Google Maps MCP for location-based queries
• Playwright: Browser automation using Playwright MCP

Dependencies

This project uses:

• Koog Agents library
• Koin for dependency injection
• KotlinX Serialization for JSON handling
• JUnit 5 and MockK for testing