エージェントとマルチエージェント

Embabel Agent Framework

embabel/embabel-agent

Embabel is an open-source JVM framework for building strongly typed agentic applications that combine LLM interactions, regular code, tools, and domain models. It supports dynamic planning and can be used from Kotlin or Java.

★ 4.3Kスター
⑂ 417フォーク数
65未解決の問題
Kotlin言語
Apache-2.0ライセンス
Q@project.QualityScore編集部スコア

概要

Embabel models workflows through actions, goals, conditions, domain objects, and dynamically generated plans. Its default planner uses Goal-Oriented Action Planning, while Utility AI is also supported for exploratory or open-ended tasks. The framework builds on Spring and executes agents through an AgentPlatform in focused, closed, or open modes. MCP and retrieval-augmented generation support are not stated in the repository metadata or supplied README excerpt.

主な機能

  • Strongly typed actions, goals, conditions, and domain models
  • Dynamic planning with replanning after each action
  • Default Goal-Oriented Action Planning implementation
  • Utility AI support for utility-driven action selection
  • Spring-based dependency injection and component integration
  • Java annotation model and idiomatic Kotlin DSL
  • Focused, closed, and open AgentPlatform execution modes
  • Support for mixing regular code, LLM calls, and tools
  • Pluggable planning step
  • Unit and end-to-end agent testing orientation
  • Ability to select different LLMs for different tasks

要件、インストール、クイックスタート

No complete installation command is present in the supplied repository record. The documented quick-start path is to create a repository from the Java template at https://github.com/embabel/java-agent-template or the Kotlin template at https://github.com/embabel/kotlin-agent-template. Maven artifact information is available at https://mvnrepository.com/artifact/com.embabel.agent/embabel-agent-api. Consult the documentation at https://docs.embabel.com/embabel-agent/guide/1.5.0-SNAPSHOT/ for version-specific setup.

使用方法

Start from one of the provided templates, configure OPENAI_API_KEY, and run the Maven-based project according to its template instructions. Define an agent using annotated Java or Kotlin classes, or use the Kotlin DSL. Add typed action methods that receive domain objects and return new state, identify goal-achieving actions, and invoke the agent through an AgentPlatform execution mode. Additional examples are available at https://github.com/embabel/embabel-agent-examples.

モデルの互換性とユースケース

The examples explicitly reference OpenAI models and an OPENAI_API_KEY. The framework exposes default and explicitly selected LLM usage and is designed to mix models, including local models for focused tasks. A complete list of supported providers, models, and local runtimes is not stated in the repository metadata.

ライセンスとリスクに関する注意事項

The repository metadata identifies the license as Apache-2.0. License metadata is available at https://api.github.com/licenses/apache-2.0. Users should review the repository license text and applicable dependency or model-provider terms before distribution or production use.

リリースとメンテナンス

No release version, changelog, or release notes are included in the supplied repository record. The linked documentation path references 1.5.0-SNAPSHOT, but the repository metadata does not identify it as a stable release.

Firecrawl

firecrawl/firecrawl

★ 161.1KTypeScript

LangChain

langchain-ai/langchain

★ 143.6KPython

RAGFlow

infiniflow/ragflow

★ 86.7KGo