正在打开每日关注
正在打开每日关注
解析 Google ADK 2.0 的图工作流、任务委派与本地开发能力,并指出推文的夸大之处。
Google just killed the agent framework industry. ADK 2.0: Open-source. Free. Better than $50K enterprise tools. What it does: → Graph-based execution with routing, fan-out/fan-in, loops, retry → Structured agent-to-agent delegation via Task API → State management, dynamic nodes, human-in-the-loop, nested workflows → Interactive CLI (adk run) and Web UI (adk web) for local dev → Multi-turn task mode with single-turn controlled output → Works with Gemini 2.5 Flash, extensions via pip What it replaces: → LangChain orchestration boilerplate → LangGraph state machines → Vertex AI Agent Builder lock-in → Custom agent-to-agent delegation code Define your Agent class with instructions and tools. Compose a Workflow class as a graph. Run it locally with adk run or adk web. No hosted platform. No vendor lock-in. Customer support bots, research agents, multi-agent pipelines - same library. This is what open-source from Google looks like. → https://t.co/SSQnZZ94w6
是什么 这是 google/adk-python 的 ADK 2.0 README:一个 Apache 2.0 开源、代码优先的 Python Agent 框架,用于构建、评估和部署智能体。仓库元数据显示 21,323 stars,主要语言为 Python,创建于 2025-04-01,最近提交为 2026-08-29。
架构与核心设计 它把 Agent(指令、模型、工具与行为)和 Workflow(节点与边)分层;Workflow Runtime 以图执行支持路由、并行 fan-out/fan-in、循环、重试、状态、动态节点、人工介入与嵌套流程。Task API 提供多轮任务、单轮受控输出和结构化 Agent-to-Agent 委派,并可将任务 Agent 作为工作流节点。工具可接自定义函数、OpenAPI 与 MCP;adk run/adk web 用于本地运行和调试,adk eval 用于评估,Cloud Run 与 Vertex AI Agent Engine 用于部署。
值得借鉴 分层模型和显式执行图让复杂流程更易组合、测试和版本化;把人审、评估、工具确认及本地 Web UI 纳入开发闭环,对研究、客服和多智能体流水线有实际价值。
边界与风险 推文把它描述成取代 LangChain、LangGraph 和 Vertex AI Agent Builder,且完全无厂商锁定,这超出了 README 能证明的范围。README 明确 ADK 针对 Gemini 优化,同时提供 Google 云部署;2.0 还改变 Agent API、事件模型和会话 schema,升级前需处理兼容性,2.0 会话仅保证 ADK 1.28+ 可读。