Намери подобно съдържание

#python#adk#agent_samples#agents The Agent Development Kit (ADK) offers ready-made sample agents in Python and Java to help you quickly build AI-powered agents for various tasks, from simple chatbots to complex multi-agent workflows. It supports flexible design, letting you combine multiple specialized agents, use diverse tools, and create adaptable workflows. ADK also includes developer tools for easy testing, debugging, and deployment, and works well with Google’s AI models and other large language models. Using these samples can save you time and effort by providing practical examples and a strong foundation to develop your own intelligent agents efficiently. This helps you focus on your agent’s logic while ADK handles orchestration and scaling. https://github.com/google/adk-samples

Google DeepMind 发布 Gemini API 开发者技能：用 Agent Skills 弥合模型知识鸿沟 Google DeepMind 于 3 月 25 日发文介绍其构建的 Gemini API 开发者技能（gemini-api-dev），旨在通过 Agent Skills 规范弥合大语言模型的知识鸿沟。测试显示，Gemini 3 系列模型在启用该技能后，代码生成正确率从不足 7% 大幅提升至接近 100%，但前提是模型需具备强大的推理能力。 ⚙️ 技能内容 该技能包含四部分： - API 高级功能集概览 - 当前模型和各语言 SDK 说明 - 各 SDK 基础示例代码 - 文档入口点列表（作为真实信息源） 技能已开源在 GitHub，支持通过 Vercel skills 和 Context7 两种方式安装： npx skills add google-gemini/gemini-skills --skill gemini-api-dev --global npx ctx7 skills install /google-gemini/gemini-skills gemini-api-dev 📊 评测结果 - 评测集：117 个 prompt，覆盖 Agent 编码、聊天机器人、文档处理、流式内容等场景 - 失败标准：生成的代码使用了旧版 SDK - Gemini 3.0 Pro/Flash：基线仅 6.8%，启用技能后大幅提升 - Gemini 3.1 Pro：基线 28%，启用技能后几乎全部通过 - Gemini 2.5 系列：也有提升，但远不如 3.x 系列——强推理能力是关键 - SDK Usage 类别通过率最低（95%），部分失败来自明确要求使用 Gemini 2.0 模型的 prompt 📎 相关链接 原文：https://developers.googleblog.com/closing-the-knowledge-gap-with-agent-skills/ GitHub：https://github.com/google-gemini/gemini-skills Agent Skills 规范：https://agentskills.io Google ADK Skills 文档：https://google.github.io/adk-docs/skills/ #Google#AI#AgentSkills#GeminiAPI#ADK#开发者工具