> ## Documentation Index > Fetch the complete documentation index at: https://java.agentscope.io/llms.txt > Use this file to discover all available pages before exploring further. # 概览 本节汇总 AgentScope Java 与第三方系统、生态服务的集成扩展。每个扩展都是 `agentscope-extensions/` 下的独立 Maven 模块，按需引入即可。 按主题分为以下几组： ## 模型提供商 所有模型提供商已经迁移到独立模型扩展模块中，`agentscope-core` 只保留共享模型契约。完整模型创建方式、Spring Boot 配置、formatter、credential 和高级 registry 行为见 [模型文档](https://java.agentscope.io/dev/zh/docs/building-blocks/model.md)。 | 提供商 | Maven artifact | `ModelRegistry` id | 标准环境变量 | 文档 | | - | - | - | - | - | | OpenAI | `agentscope-extensions-model-openai` | `openai:` | `OPENAI_API_KEY` | [OpenAI](https://java.agentscope.io/dev/zh/integration/model/openai.md) | | OpenAI Official | `agentscope-extensions-model-openai-official` | `openai-official:` | `OPENAI_API_KEY` | [OpenAI Official](https://java.agentscope.io/dev/zh/integration/model/openai-official.md) | | DeepSeek | `agentscope-extensions-model-openai` | `deepseek:` | `DEEPSEEK_API_KEY` | [DeepSeek](https://java.agentscope.io/dev/zh/integration/model/deepseek.md) | | GLM | `agentscope-extensions-model-openai` | `glm:` | `ZAI_API_KEY` / `GLM_API_KEY` / `ZHIPUAI_API_KEY` | [GLM](https://java.agentscope.io/dev/zh/integration/model/glm.md) | | Kimi | `agentscope-extensions-model-openai` | `kimi:` | `MOONSHOT_API_KEY` / `KIMI_API_KEY` | [Kimi](https://java.agentscope.io/dev/zh/integration/model/kimi.md) | | MiniMax | `agentscope-extensions-model-openai` | `minimax:` | `MINIMAX_API_KEY` | [MiniMax](https://java.agentscope.io/dev/zh/integration/model/minimax.md) | | DashScope | `agentscope-extensions-model-dashscope` | `dashscope:` / `qwen*` | `DASHSCOPE_API_KEY` | [DashScope](https://java.agentscope.io/dev/zh/integration/model/dashscope.md) | | Gemini | `agentscope-extensions-model-gemini` | `gemini:` | `GEMINI_API_KEY` | [Gemini](https://java.agentscope.io/dev/zh/integration/model/gemini.md) | | Anthropic | `agentscope-extensions-model-anthropic` | `anthropic:` | `ANTHROPIC_API_KEY` | [Anthropic](https://java.agentscope.io/dev/zh/integration/model/anthropic.md) | | Ollama | `agentscope-extensions-model-ollama` | `ollama:` | `OLLAMA_BASE_URL`（可选） | [Ollama](https://java.agentscope.io/dev/zh/integration/model/ollama.md) | > [!NOTE] > > `agentscope-extensions-model-e2e-tests` 是仓库测试模块，不是面向用户的模型集成依赖。 ## 分布式存储（Distributed Store） 生产多副本部署所需的全链路分布式存储组件。通过 `DistributedStore` 一键配置 Agent 状态、工作区文件系统、沙箱快照与并发锁。 * [分布式存储总览](https://java.agentscope.io/dev/zh/integration/distributed/index.md) — `DistributedStore` API、能力矩阵、混合后端 * [Redis](https://java.agentscope.io/dev/zh/integration/distributed/redis.md) — `AgentStateStore` + `BaseStore` + `SandboxSnapshotSpec` + `SandboxExecutionGuard` * [JDBC](https://java.agentscope.io/dev/zh/integration/distributed/jdbc.md) — `AgentStateStore` + `JdbcStore` + `JdbcSnapshotSpec` + `JdbcSandboxExecutionGuard` * [阿里云 OSS](https://java.agentscope.io/dev/zh/integration/distributed/oss.md) — `AgentStateStore` + `OssBaseStore` + `OssSnapshotSpec` ## 沙箱执行环境（Sandbox） 隔离的代码执行环境，所有 `SandboxFilesystemSpec` 实现。Docker 内置在 harness 中，其余为独立扩展模块。 * Docker — 内置默认，无需额外依赖 * [Kubernetes](https://java.agentscope.io/dev/zh/docs/harness/sandbox.md) — `agentscope-extensions-sandbox-kubernetes` * [AgentRun（阿里云）](https://java.agentscope.io/dev/zh/docs/harness/sandbox.md) — `agentscope-extensions-sandbox-agentrun` * [Daytona](https://java.agentscope.io/dev/zh/docs/harness/sandbox.md) — `agentscope-extensions-sandbox-daytona` * [E2B](https://java.agentscope.io/dev/zh/docs/harness/sandbox.md) — `agentscope-extensions-sandbox-e2b` ## 记忆（Memory） 跨会话持久化用户偏好与事实，所有实现都符合 `LongTermMemory` 接口。 * [Mem0](https://java.agentscope.io/dev/zh/integration/memory/mem0.md) * [百炼记忆](https://java.agentscope.io/dev/zh/integration/memory/bailian.md) * [ReMe](https://java.agentscope.io/dev/zh/integration/memory/reme.md) ## RAG 知识库 通过 `Knowledge` 接口接入不同的检索后端。 * [Simple（自建 embedding + 向量库）](https://java.agentscope.io/dev/zh/integration/rag/simple.md) * [百炼知识库](https://java.agentscope.io/dev/zh/integration/rag/bailian.md) * [Dify](https://java.agentscope.io/dev/zh/integration/rag/dify.md) * [HayStack](https://java.agentscope.io/dev/zh/integration/rag/haystack.md) * [RAGFlow](https://java.agentscope.io/dev/zh/integration/rag/ragflow.md) ## 技能仓库（Skill） `AgentSkillRepository` 的多种存储实现。 * [Git 技能仓库](https://java.agentscope.io/dev/zh/integration/skill/git-repository.md) * [MySQL 技能仓库](https://java.agentscope.io/dev/zh/integration/skill/mysql-repository.md) * [PostgreSQL 技能仓库](https://java.agentscope.io/dev/zh/integration/skill/postgresql-repository.md) * 也可以使用 [Nacos 技能仓库](https://java.agentscope.io/dev/zh/integration/infrastructure/nacos.md#skill-仓库) ## Channel 适配器 通过 Harness Channel 接口将 Agent 接入消息平台。 * [钉钉](https://java.agentscope.io/dev/zh/integration/channel/dingtalk.md) * [飞书 / Lark](https://java.agentscope.io/dev/zh/integration/channel/feishu.md) * [GitHub](https://java.agentscope.io/dev/zh/integration/channel/github.md) * [GitLab](https://java.agentscope.io/dev/zh/integration/channel/gitlab.md) * [企业微信](https://java.agentscope.io/dev/zh/integration/channel/wecom.md) ## 智能体协议 让 Agent 与外部世界以标准方式交互。 * [A2A（Agent-to-Agent）](https://java.agentscope.io/dev/zh/integration/protocol/a2a.md) * [AG-UI](https://java.agentscope.io/dev/zh/integration/protocol/agui.md) * [Agent Protocol](https://java.agentscope.io/dev/zh/integration/protocol/agent-protocol.md) ## 基础设施 / 中间件 把 Agent 接到企业基础设施。 * [Higress AI 网关](https://java.agentscope.io/dev/zh/integration/infrastructure/higress.md) * [Nacos](https://java.agentscope.io/dev/zh/integration/infrastructure/nacos.md) * [Scheduler（Quartz / XXL-Job）](https://java.agentscope.io/dev/zh/integration/infrastructure/scheduler.md) ## 生态扩展 运行环境、语言生态、调试与训练流水线。 * [Chat Completions Web](https://java.agentscope.io/dev/zh/integration/ecosystem/chat-completions-web.md) * [AgentScope Studio](https://java.agentscope.io/dev/zh/integration/ecosystem/studio.md) * [在线训练（Training）](https://java.agentscope.io/dev/zh/integration/ecosystem/training.md) > [!NOTE] > > 若你正在使用 Spring Boot，绝大多数扩展都有对应的 `agentscope-spring-boot-starter-*` 一键接入版本，可减少手动装配代码。 This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.