Reliable AI Agents Using Domain Modeling with Koog in Java
summary
Koog ships a new idiomatic Java API for building JVM-native agents, with a claim of >4,000 GitHub stars and real deployments at banks and Mercedes‑Benz. The demo shows registering Java methods as tools, composing type-safe subtasks and graph strategies that restrict tool access, checkpointing execution for crash recovery via persistence, and a history compression feature to shrink long contexts. Examples include Spring Boot integration, multi‑model routing (OpenAI, Anthropic, local), and compile‑time checks for subtask inputs/outputs.
AI-written summary · about
chapters (15)
- 0:00Introduction to Cook
- 2:56The concept of AI agents
- 6:54Building an AI agent
- 9:40Developing agent tools
- 12:26Integration with Spring
- 14:02Managing chat history
- 16:33Customizing agent strategies
- 19:46Implementing complex flows
- 24:21Graph-based strategy coding
- 29:14Advanced agent techniques
- 31:10Strategy types explained
- 33:51Persistence and recovery
- 35:52Observability and monitoring
- 38:47History compression techniques
- 43:39Summary and Q&A
related talks
51mSupercharging Spring AI: Scalable AI Agents with Koog + Spring AI + Spring Boot
10mAI agents on the JVM
53mUnleash the Power of AI Agents with Java Building Intelligent Application
54mGenAI Grows Up: Building Production-Ready Agents on the JVM
02h 28mEvent-driven business automation powered by cloud native Java
34mBuilding a Weather Agent on Android with Koog
30mLessons From Building and Deploying a Multi-Agent System at Scale: An...
02h 49mJava meets AI: Build LLM-Powered Apps with LangChain4j