LanguagesConference50min
Koog your own AI! Delicious recipe
The proposal introduces Koog, a framework for building scalable, production-ready AI agents in Java and Kotlin. It covers managing context and token usage, creating type-safe, composable workflows, ensuring persistence and fault tolerance, and applying observability and tracing with OpenTelemetry to support enterprise-level agent development on the JVM.
talk.summaryAiDisclaimer
Simon VergauwenJetBrains
talkDetail.whenAndWhere
Wednesday, May 6, 12:00-12:50
Room B
If you’re not cooking your own AI agents yet, you should! Doing so isn’t always easy, but Koog allows you to build them directly in your favorite JVM languages: Java and Kotlin. Creating a simple agent is fairly straightforward, but how do we take these systems to production at scale? There are many things to take into account, and the Koog team has been working hard to make this both simple and enjoyable. So join me on a tour of Koog and learn how to create enterprise-level AI agents on the JVM.
You’ll learn how to:
• Prevent your context window from growing and your token usage from exploding
• Build type-safe and composable workflows and agents
• Guarantee fault tolerance by making agents persistent and durable
• Keep a close eye on your agents using observation and tracing techniques with OpenTelemetry
You’ll learn how to:
• Prevent your context window from growing and your token usage from exploding
• Build type-safe and composable workflows and agents
• Guarantee fault tolerance by making agents persistent and durable
• Keep a close eye on your agents using observation and tracing techniques with OpenTelemetry