介绍 Embabel：用于 Java 应用程序的高级 AI 代理开发

Embabel Agent Framework: Launched by Spring founder Rod Johnson, it's a JVM-based platform for building AI-powered apps on the Spring Framework. Aims to transform LLM integration with Java apps.
- Models "agentic flows" using actions, goals, and conditions. Leverages [Goal-Oriented Action Planning (GOAP)] for proper planning, forming an [OODA loop].
- Strong typing and object orientation ensure LLM interactions are typed, with compile-time checking, etc.
- Treats planning as a distinct capability, with GOAP implementation determining optimal action sequences. Benefits include enhanced efficiency, dynamic adaptability, increased composability, and parallelization.
- Supports three execution modes: Focused, Closed, and Open. Prioritizes ease of testing with unit-testable agents and a prompt testing library. Extensible without altering existing code.
- Provides clear separation between programming model and platform internals for local runs and better production QoS. Supports LLM mixing for cost-effectiveness.
Language Choice: Rod Johnson prefers Kotlin for Embabel development due to null safety, more readable collection usage, concise syntax, and reification. But the choice of language for the framework doesn't affect the language for building apps. Embabel's core values hold in both Kotlin and Java.
Positioning Relative to Other Initiatives: Embabel is a higher-level framework than LangChain4j and Spring AI. Builds on Spring AI but offers more abstraction and original ideas. Urgently needed for Java to maintain its position.
Adoption by Traditional Java Teams: Understanding of the domain and existing apps is crucial. Embabel aims to make it easy for JVM developers to write agents using their core skills. Java teams are well-positioned to leverage AI with minimal friction.
Evolution in Relation to Spring AI: Embabel is an independent project focused on solving JVM's Gen AI problem. It's a higher level than Spring AI and can introduce novel ideas. Eventually, it may be ported to other platforms like TypeScript and Python.
Learning Resources: The embabel repository is the starting point. Examples are gradually moving to the embabel-agent-examples subdirectory. Examples like the coding agent and travel planner are being evolved to be useful.