[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-zenml-io--zenml":3,"tool-zenml-io--zenml":64},[4,17,27,35,43,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",138956,2,"2026-04-05T11:33:21",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[13,26],{"id":44,"name":45,"github_repo":46,"description_zh":47,"stars":48,"difficulty_score":23,"last_commit_at":49,"category_tags":50,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,51,52,53,15,54,26,13,55],"数据工具","视频","插件","其他","音频",{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"status":16},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,"2026-04-04T04:44:48",[15,14,13,26,54],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":76,"owner_avatar_url":77,"owner_bio":78,"owner_company":79,"owner_location":79,"owner_email":80,"owner_twitter":81,"owner_website":82,"owner_url":83,"languages":84,"stars":121,"forks":122,"last_commit_at":123,"license":124,"difficulty_score":23,"env_os":125,"env_gpu":126,"env_ram":127,"env_deps":128,"category_tags":138,"github_topics":139,"view_count":23,"oss_zip_url":79,"oss_zip_packed_at":79,"status":16,"created_at":159,"updated_at":160,"faqs":161,"releases":162},2787,"zenml-io\u002Fzenml","zenml","ZenML 🙏: One AI Platform from Pipelines to Agents. https:\u002F\u002Fzenml.io.","ZenML 是一个专为机器学习工程师和 AI 开发者打造的开源平台，旨在打通从传统模型训练管道到现代大语言模型（LLM）智能体（Agents）的全流程工作流。它核心解决了 AI 项目在落地过程中面临的基础设施复杂、实验难以追踪以及开发环境与生产环境不一致等痛点。\n\n通过 ZenML，用户只需编写标准的 Python 代码来定义业务逻辑（如模型训练或智能体循环），平台即可自动处理容器化构建、运行追踪及元数据管理。其独特之处在于强大的“栈（Stacks）”抽象能力，能够屏蔽底层基础设施的差异，让同一套代码无缝运行在本地、云端或各类集群上。同时，ZenML 原生支持集成 MLflow、LangGraph、Sagemaker 等主流工具，帮助团队快速构建可观测、可迭代的 AI 应用。\n\n无论是需要规范实验流程的研究人员，还是致力于将 AI 模型投入生产的企业工程团队，ZenML 都能提供标准化的解决方案。目前，包括空客、AXA 在内的众多知名企业已利用它来高效管理和运营其 AI 工作流，是实现 AI 工程化落地的得力助手。","\u003Cdiv align=\"center\">\n\n  \u003C!-- PROJECT LOGO -->\n  \u003Cbr \u002F>\n    \u003Ca href=\"https:\u002F\u002Fzenml.io\">\n      \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzenml-io_zenml_readme_4ccebbac9740.png\" alt=\"ZenML Header\">\n    \u003C\u002Fa>\n  \u003Cbr \u002F>\n  \u003Cdiv align=\"center\">\n    \u003Ch3 align=\"center\">One AI Platform From Pipelines to Agents \u003C\u002Fh3>\n  \u003C\u002Fdiv>\n\n  [![PyPi][pypi-shield]][pypi-url]\n  [![PyPi][pypiversion-shield]][pypi-url]\n  [![PyPi][downloads-shield]][downloads-url]\n  [![Contributors][contributors-shield]][contributors-url]\n  [![License][license-shield]][license-url]\n\n\u003C\u002Fdiv>\n\n\u003C!-- MARKDOWN LINKS & IMAGES -->\n[pypi-shield]: https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fpyversions\u002Fzenml?color=281158\n[pypi-url]: https:\u002F\u002Fpypi.org\u002Fproject\u002Fzenml\u002F\n[pypiversion-shield]: https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fzenml?color=361776\n[downloads-shield]: https:\u002F\u002Fimg.shields.io\u002Fpepy\u002Fdt\u002Fzenml?color=431D93\n[downloads-url]: https:\u002F\u002Fpypi.org\u002Fproject\u002Fzenml\u002F\n[contributors-shield]: https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fcontributors\u002Fzenml-io\u002Fzenml?color=7A3EF4\n[contributors-url]: https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fgraphs\u002Fcontributors\n[license-shield]: https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002Fzenml-io\u002Fzenml?color=9565F6\n[license-url]: https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fblob\u002Fmain\u002FLICENSE\n\n\u003Cdiv align=\"center\">\n\u003Cp>\n    \u003Ca href=\"https:\u002F\u002Fzenml.io\u002Fprojects\">Projects\u003C\u002Fa> •\n    \u003Ca href=\"https:\u002F\u002Fzenml.io\u002Froadmap\">Roadmap\u003C\u002Fa> •\n    \u003Ca href=\"https:\u002F\u002Fdocs.zenml.io\u002Fchangelog\">Changelog\u003C\u002Fa> •\n    \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fissues\">Report Bug\u003C\u002Fa> •\n    \u003Ca href=\"https:\u002F\u002Fzenml.io\u002Fpro\">Sign up for ZenML Pro\u003C\u002Fa> •\n    \u003Ca href=\"https:\u002F\u002Fwww.zenml.io\u002Fblog\">Blog\u003C\u002Fa> •\n    \u003Ca href=\"https:\u002F\u002Fdocs.zenml.io\u002F\">Docs\u003C\u002Fa>\n    \u003Cbr \u002F>\n    \u003Cbr \u002F>\n    🎉 For the latest release, see the \u003Ca href=\"https:\u002F\u002Fdocs.zenml.io\u002Fchangelog\">changelog\u003C\u002Fa>.\n\u003C\u002Fp>\n\u003C\u002Fdiv>\n\n---\n\nZenML is built for ML or AI Engineers working on traditional ML use-cases, LLM workflows, or agents, in a company setting.\n\nAt it's core, ZenML allows you to write **workflows (pipelines)** that run on any **infrastructure backend (stacks)**. You can embed any Pythonic logic within these pipelines, like training a model, or running an agentic loop. ZenML then operationalizes your application by:\n\n1. Automatically containerizing and tracking your code.\n2. Tracking individual runs with metrics, logs, and metadata.\n3. Abstracting away infrastructure complexity.\n4. Integrating your existing tools and infrastructure e.g. MLflow, Langgraph, Langfuse, Sagemaker, GCP Vertex, etc.\n5. Allowing you to quickly iterate on experiments with an observable layer, in development and in production.\n\n...amongst many other features.\n\nZenML is used by thousands of companies to run their AI workflows. Here are some featured ones:\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fwww.airbus.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F66e826c67966c0e639be6591_airbus.svg\" alt=\"Airbus\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fwww.axa.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F66c84308916684f0d07b57ff_axa-min.svg\" alt=\"AXA\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fwww.jetbrains.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F682337dd23ca98ec293c2dc6_jetbrains-min.svg\" alt=\"JetBrains\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Frivian.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F66e9897d1b1dc28e560c0c07_rivian-min.svg\" alt=\"Rivian\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fwww.wisetechglobal.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F65ddeac90f19eb6c4cd715f9_wisetech_logo-min.svg\" alt=\"WiseTech Global\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fwww.brevo.com\u002F\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzenml-io_zenml_readme_62aca4a32447.webp\" alt=\"Brevo\" height=\"50\"\u002F>\u003C\u002Fa>\n\u003C\u002Fdiv>\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fwww.leroymerlin.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F65ddeac9b83eea2954b5a561_leroy_merlin_logo-min.svg\" alt=\"Leroy Merlin\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fwww.koble.ai\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F67e673e60161f59b5db6554c_koble.svg\" alt=\"Koble\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fwww.playtika.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F66e959c0c5f8422ecac8d81a_Playtika-min.svg\" alt=\"Playtika\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fnielseniq.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F65ddeac959d7ca93745e8130_nielsen_iq_logo-min.svg\" alt=\"NIQ\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fwww.enel.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F66c84308b1e802ab9a246134_enel-min.svg\" alt=\"Enel\" height=\"50\"\u002F>\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n\u003Csub>\u003Ci>(please email support@zenml.io if you want to be featured)\u003C\u002Fi>\u003C\u002Fsub>\n\n## 🚀 Get Started (5 minutes)\n\n```bash\n# Install ZenML with server capabilities\npip install \"zenml[server]\"  # pip install zenml will install a slimmer client\n\n# Initialize your ZenML repository\nzenml init\n\n# Start local server or connect to a remote one\nzenml login\n```\n\nYou can then explore any of the [examples](examples\u002F) in this repo. We recommend starting with the [quickstart](examples\u002Fquickstart\u002F), which demonstrates core ZenML concepts: pipelines, steps, artifacts, snapshots, and deployments.\n\n### 🏗️ Architecture Overview\n\nZenML uses a [**client-server architecture**](https:\u002F\u002Fdocs.zenml.io\u002Fgetting-started\u002Fsystem-architectures) with an integrated web dashboard ([zenml-io\u002Fzenml-dashboard](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml-dashboard)):\n\n- **Local Development**: `pip install \"zenml[local]\"` - runs both client and server locally\n- **Production**: Deploy server separately, connect with `pip install zenml` + `zenml login \u003Cserver-url>`\n\n## 🎮 Demo\n\nHere is a short demo:\n\n [![Watch the video](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzenml-io_zenml_readme_067280c79799.png)](https:\u002F\u002Fzenml.io\u002Fdemo-video)\n\n## 🖼️ Resources\n\nThe best way to learn about ZenML is through our comprehensive documentation and tutorials:\n\n- **[Documentation](https:\u002F\u002Fdocs.zenml.io\u002F)** - Complete product documentation\n- **[Your First AI Pipeline](https:\u002F\u002Fdocs.zenml.io\u002Fgetting-started\u002Fyour-first-ai-pipeline)** - Build and evaluate an AI service in minutes\n- **[Starter Guide](https:\u002F\u002Fdocs.zenml.io\u002Fuser-guides\u002Fstarter-guide)** - From zero to production in 30 minutes\n- **[LLMOps Guide](https:\u002F\u002Fdocs.zenml.io\u002Fuser-guides\u002Fllmops-guide)** - Specific patterns for LLM applications\n- **[SDK Reference](https:\u002F\u002Fsdkdocs.zenml.io\u002F)** - Complete SDK reference\n\n ## 📚 More examples\n\n1. **[Agent Architecture Comparison](examples\u002Fagent_comparison\u002F)** - Compare AI agents with LangGraph workflows, LiteLLM integration, and automatic visualizations via custom materializers\n2. **[Deploying ML Models](examples\u002Fdeploying_ml_model\u002F)** - Deploy classical ML models as production endpoints with monitoring and versioning\n3. **[Deploying Agents](examples\u002Fdeploying_agent\u002F)** - Document analysis service with pipelines, evaluation, and embedded web UI\n4. **[E2E Batch Inference](examples\u002Fe2e\u002F)** - Complete MLOps pipeline with feature engineering\n5. **[LLM RAG Pipeline](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml-projects\u002Ftree\u002Fmain\u002Fllm-complete-guide)** - Production RAG with evaluation loops\n6. **[Agentic Workflow (Deep Research)](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml-projects\u002Ftree\u002Fmain\u002Fdeep_research)** - Orchestrate your agents with ZenML\n7. **[Fine-tuning Pipeline](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml-projects\u002Ftree\u002Fmain\u002Fgamesense)** - Fine-tune and deploy LLMs\n\n## 🗣️ Chat With Your Pipelines: ZenML MCP Server\n\nStop clicking through dashboards to understand your ML workflows. The **[ZenML MCP Server](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fmcp-zenml)** lets you query your pipelines, analyze runs, and trigger deployments using natural language through Claude Desktop, Cursor, or any MCP-compatible client.\n\n```\n💬 \"Which pipeline runs failed this week and why?\"\n📊 \"Show me accuracy metrics for all my customer churn models\"  \n🚀 \"Trigger the latest fraud detection pipeline with production data\"\n```\n\n**Quick Setup:**\n1. Download the `.dxt` file from [zenml-io\u002Fmcp-zenml](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fmcp-zenml)\n2. Drag it into Claude Desktop settings\n3. Add your ZenML server URL and API key\n4. Start chatting with your ML infrastructure\n\nThe MCP (Model Context Protocol) integration transforms your ZenML metadata into conversational insights, making pipeline debugging and analysis as easy as asking a question. Perfect for teams who want to democratize access to ML operations without requiring dashboard expertise.\n\n### 🎓 Books & Resources\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fwww.amazon.com\u002FLLM-Engineers-Handbook-engineering-production\u002Fdp\u002F1836200072\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzenml-io_zenml_readme_94a1402b13bb.jpg\" alt=\"LLM Engineer's Handbook Cover\" width=\"200\"\u002F>\n  \u003C\u002Fa>\n  \u003Ca href=\"https:\u002F\u002Fwww.amazon.com\u002F-\u002Fen\u002FAndrew-McMahon\u002Fdp\u002F1837631964\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzenml-io_zenml_readme_cf5a2c138a4c.jpg\" alt=\"Machine Learning Engineering with Python Cover\" width=\"200\"\u002F>\n  \u003C\u002Fa>\n  \u003Ca href=\"https:\u002F\u002Fwww.amazon.com\u002FDeepSeek-Practice-fine-tuning-distillation-engineering\u002Fdp\u002F1806020858\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzenml-io_zenml_readme_9f2a1efe61df.jpg\" alt=\"DeepSeek in Practice Cover\" width=\"200\"\u002F>\n  \u003C\u002Fa>\n\u003C\u002Fdiv>\n\n[ZenML](https:\u002F\u002Fzenml.io) is featured in these comprehensive guides to production AI systems.\n\n## 🤝 Join ML Engineers Building the Future of AI\n\n**Contribute:**\n- 🌟 [Star us on GitHub](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fstargazers) - Help others discover ZenML\n- 🤝 [Contributing Guide](CONTRIBUTING.md) - Start with [`good-first-issue`](https:\u002F\u002Fgithub.com\u002Fissues?q=is%3Aopen+is%3Aissue+archived%3Afalse+user%3Azenml-io+label%3A%22good+first+issue%22)\n- 💻 [Write Integrations](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fblob\u002Fmain\u002Fsrc\u002Fzenml\u002Fintegrations\u002FREADME.md) - Add your favorite tools\n\n**Stay Updated:**\n- 🗺 [Public Roadmap](https:\u002F\u002Fzenml.io\u002Froadmap) - See what's coming next\n- 📰 [Blog](https:\u002F\u002Fzenml.io\u002Fblog) - Best practices and case studies\n- 🎙 [Slack](https:\u002F\u002Fzenml.io\u002Fslack) - Talk with AI practitioners\n\n## ❓ FAQs from ML Engineers Like You\n\n**Q: \"Do I need to rewrite my agents or models to use ZenML?\"**\n\nA: No. Wrap your existing code in a `@step`. Keep using `scikit-learn`, PyTorch, LangGraph, LlamaIndex, or raw API calls. ZenML orchestrates your tools, it doesn't replace them.\n\n**Q: \"How is this different from LangSmith\u002FLangfuse?\"**\n\nA: They provide excellent observability for LLM applications. We orchestrate the **full MLOps lifecycle for your entire AI stack**. With ZenML, you manage both your classical ML models and your AI agents in one unified framework, from development and evaluation all the way to production deployment.\n\n**Q: \"Can I use my existing MLflow\u002FW&B setup?\"**\n\nA: Yes! ZenML integrates with both [MLflow](https:\u002F\u002Fdocs.zenml.io\u002Fstacks\u002Fexperiment-trackers\u002Fmlflow) and [Weights & Biases](https:\u002F\u002Fdocs.zenml.io\u002Fstacks\u002Fexperiment-trackers\u002Fwandb). Your experiments, our pipelines.\n\n**Q: \"Is this just MLflow with extra steps?\"**\n\nA: No. MLflow tracks experiments. We orchestrate the entire development process – from training and evaluation to deployment and monitoring – for both models and agents.\n\n**Q: \"How do I configure ZenML with Kubernetes?\"**\n\nA: ZenML integrates with Kubernetes through the native Kubernetes orchestrator, Kubeflow, and other K8s-based orchestrators. See our [Kubernetes orchestrator guide](https:\u002F\u002Fdocs.zenml.io\u002Fstacks\u002Forchestrators\u002Fkubernetes) and [Kubeflow guide](https:\u002F\u002Fdocs.zenml.io\u002Fstacks\u002Forchestrators\u002Fkubeflow), plus [deployment documentation](https:\u002F\u002Fdocs.zenml.io\u002Fgetting-started\u002Fdeploying-zenml\u002Fdeploy-with-helm).\n\n**Q: \"What about cost? I can't afford another platform.\"**\n\nA: ZenML's open-source version is free forever. You likely already have the required infrastructure (like a Kubernetes cluster and object storage). We just help you make better use of it for MLOps.\n\n### 🛠 VS Code \u002F Cursor Extension\n\nManage pipelines directly from your editor:\n\n\u003Cdetails>\n  \u003Csummary>🖥️ VS Code Extension in Action!\u003C\u002Fsummary>\n  \u003Cdiv align=\"center\">\n  \u003Cimg width=\"60%\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzenml-io_zenml_readme_b1a121a03e70.gif\" alt=\"ZenML Extension\">\n\u003C\u002Fdiv>\n\u003C\u002Fdetails>\n\nInstall from [VS Code Marketplace](https:\u002F\u002Fmarketplace.visualstudio.com\u002Fitems?itemName=ZenML.zenml-vscode).\n\n## 📜 License\n\nZenML is distributed under the terms of the Apache License Version 2.0. See\n[LICENSE](LICENSE) for details.\n\n---\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fwww.linuxfoundation.org\u002F\">\u003Cimg src=\"docs\u002Fbook\u002F.gitbook\u002Fassets\u002Flf-member-silver.svg\" alt=\"Linux Foundation Silver Member\" height=\"100\"\u002F>\u003C\u002Fa>\n      \n  \u003Ca href=\"https:\u002F\u002Fwww.cncf.io\u002F\">\u003Cimg src=\"docs\u002Fbook\u002F.gitbook\u002Fassets\u002Fcncf-member-silver.svg\" alt=\"CNCF Silver Member\" height=\"100\"\u002F>\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n\u003Cimg referrerpolicy=\"no-referrer-when-downgrade\" src=\"https:\u002F\u002Fstatic.scarf.sh\u002Fa.png?x-pxid=0fcbab94-8fbe-4a38-93e8-c2348450a42e\" \u002F>\n","\u003Cdiv align=\"center\">\n\n  \u003C!-- 项目Logo -->\n  \u003Cbr \u002F>\n    \u003Ca href=\"https:\u002F\u002Fzenml.io\">\n      \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzenml-io_zenml_readme_4ccebbac9740.png\" alt=\"ZenML页眉\">\n    \u003C\u002Fa>\n  \u003Cbr \u002F>\n  \u003Cdiv align=\"center\">\n    \u003Ch3 align=\"center\">从流水线到智能体的一站式AI平台\u003C\u002Fh3>\n  \u003C\u002Fdiv>\n\n  [![PyPi][pypi-shield]][pypi-url]\n  [![PyPi][pypiversion-shield]][pypi-url]\n  [![PyPi][downloads-shield]][downloads-url]\n  [![Contributors][contributors-shield]][contributors-url]\n  [![License][license-shield]][license-url]\n\n\u003C\u002Fdiv>\n\n\u003C!-- Markdown链接与图片 -->\n[pypi-shield]: https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fpyversions\u002Fzenml?color=281158\n[pypi-url]: https:\u002F\u002Fpypi.org\u002Fproject\u002Fzenml\u002F\n[pypiversion-shield]: https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fzenml?color=361776\n[downloads-shield]: https:\u002F\u002Fimg.shields.io\u002Fpepy\u002Fdt\u002Fzenml?color=431D93\n[downloads-url]: https:\u002F\u002Fpypi.org\u002Fproject\u002Fzenml\u002F\n[contributors-shield]: https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fcontributors\u002Fzenml-io\u002Fzenml?color=7A3EF4\n[contributors-url]: https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fgraphs\u002Fcontributors\n[license-shield]: https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002Fzenml-io\u002Fzenml?color=9565F6\n[license-url]: https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fblob\u002Fmain\u002FLICENSE\n\n\u003Cdiv align=\"center\">\n\u003Cp>\n    \u003Ca href=\"https:\u002F\u002Fzenml.io\u002Fprojects\">项目\u003C\u002Fa> •\n    \u003Ca href=\"https:\u002F\u002Fzenml.io\u002Froadmap\">路线图\u003C\u002Fa> •\n    \u003Ca href=\"https:\u002F\u002Fdocs.zenml.io\u002Fchangelog\">变更日志\u003C\u002Fa> •\n    \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fissues\">报告Bug\u003C\u002Fa> •\n    \u003Ca href=\"https:\u002F\u002Fzenml.io\u002Fpro\">注册ZenML Pro\u003C\u002Fa> •\n    \u003Ca href=\"https:\u002F\u002Fwww.zenml.io\u002Fblog\">博客\u003C\u002Fa> •\n    \u003Ca href=\"https:\u002F\u002Fdocs.zenml.io\u002F\">文档\u003C\u002Fa>\n    \u003Cbr \u002F>\n    \u003Cbr \u002F>\n    🎉 如需了解最新版本，请参阅\u003Ca href=\"https:\u002F\u002Fdocs.zenml.io\u002Fchangelog\">变更日志\u003C\u002Fa>。\n\u003C\u002Fp>\n\u003C\u002Fdiv>\n\n---\n\nZenML专为在企业环境中从事传统机器学习用例、大语言模型工作流或智能体开发的机器学习或人工智能工程师打造。\n\n其核心功能是允许您编写可在任何**基础设施后端（堆栈）**上运行的**工作流（流水线）**。您可以在这些流水线中嵌入任何基于Python的逻辑，例如训练模型或执行智能体循环。随后，ZenML会通过以下方式将您的应用投入生产：\n\n1. 自动对代码进行容器化并跟踪；\n2. 记录每次运行的指标、日志和元数据；\n3. 抽象化基础设施的复杂性；\n4. 集成您现有的工具和基础设施，如MLflow、Langgraph、Langfuse、Sagemaker、GCP Vertex等；\n5. 在开发和生产环境中，借助可观测层帮助您快速迭代实验。\n\n……以及许多其他功能。\n\nZenML已被数千家公司用于运行其AI工作流。以下是一些典型案例：\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fwww.airbus.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F66e826c67966c0e639be6591_airbus.svg\" alt=\"空中客车\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fwww.axa.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F66c84308916684f0d07b57ff_axa-min.svg\" alt=\"安盛\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fwww.jetbrains.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F682337dd23ca98ec293c2dc6_jetbrains-min.svg\" alt=\"JetBrains\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Frivian.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F66e9897d1b1dc28e560c0c07_rivian-min.svg\" alt=\"Rivian\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fwww.wisetechglobal.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F65ddeac90f19eb6c4cd715f9_wisetech_logo-min.svg\" alt=\"WiseTech Global\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fwww.brevo.com\u002F\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzenml-io_zenml_readme_62aca4a32447.webp\" alt=\"Brevo\" height=\"50\"\u002F>\u003C\u002Fa>\n\u003C\u002Fdiv>\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fwww.leroymerlin.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F65ddeac9b83eea2954b5a561_leroy_merlin_logo-min.svg\" alt=\"乐罗梅林\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fwww.koble.ai\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F67e673e60161f59b5db6554c_koble.svg\" alt=\"Koble\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fwww.playtika.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F66e959c0c5f8422ecac8d81a_Playtika-min.svg\" alt=\"Playtika\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fnielseniq.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F65ddeac959d7ca93745e8130_nielsen_iq_logo-min.svg\" alt=\"NIQ\" height=\"50\"\u002F>\u003C\u002Fa>    \n  \u003Ca href=\"https:\u002F\u002Fwww.enel.com\u002F\">\u003Cimg src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F64a817a2e7e2208272d1ce30\u002F66c84308b1e802ab9a246134_enel-min.svg\" alt=\"Enel\" height=\"50\"\u002F>\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n\u003Csub>\u003Ci>（如需被收录为案例，请发送邮件至support@zenml.io）\u003C\u002Fi>\u003C\u002Fsub>\n\n## 🚀 快速入门（5分钟）\n\n```bash\n# 安装带有服务器功能的ZenML\npip install \"zenml[server]\"  # 若仅安装“zenml”，则会安装精简版客户端\n\n# 初始化您的ZenML仓库\nzenml init\n\n# 启动本地服务器或连接远程服务器\nzenml login\n```\n\n之后，您可以探索本仓库中的任意[示例](examples\u002F)。我们建议从[快速入门](examples\u002Fquickstart\u002F)开始，它演示了ZenML的核心概念：流水线、步骤、工件、快照和部署。\n\n### 🏗️ 架构概览\n\nZenML采用[**客户端-服务器架构**](https:\u002F\u002Fdocs.zenml.io\u002Fgetting-started\u002Fsystem-architectures)，并集成了Web仪表盘（[zenml-io\u002Fzenml-dashboard](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml-dashboard))：\n\n- **本地开发**：`pip install \"zenml[local]\"` - 在本地同时运行客户端和服务器\n- **生产环境**：单独部署服务器，使用`pip install zenml` + `zenml login \u003Cserver-url>`进行连接\n\n## 🎮 演示\n\n这里有一个简短的演示：\n\n [![观看视频](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzenml-io_zenml_readme_067280c79799.png)](https:\u002F\u002Fzenml.io\u002Fdemo-video)\n\n## 🖼️ 资源\n\n了解 ZenML 的最佳方式是通过我们全面的文档和教程：\n\n- **[文档](https:\u002F\u002Fdocs.zenml.io\u002F)** - 完整的产品文档\n- **[你的第一个 AI 管道](https:\u002F\u002Fdocs.zenml.io\u002Fgetting-started\u002Fyour-first-ai-pipeline)** - 在几分钟内构建并评估一个 AI 服务\n- **[入门指南](https:\u002F\u002Fdocs.zenml.io\u002Fuser-guides\u002Fstarter-guide)** - 30 分钟内从零开始到生产环境\n- **[LLMOps 指南](https:\u002F\u002Fdocs.zenml.io\u002Fuser-guides\u002Fllmops-guide)** - 针对 LLM 应用程序的特定模式\n- **[SDK 参考](https:\u002F\u002Fsdkdocs.zenml.io\u002F)** - 完整的 SDK 参考\n\n ## 📚 更多示例\n\n1. **[智能体架构比较](examples\u002Fagent_comparison\u002F)** - 使用 LangGraph 工作流、LiteLLM 集成以及自定义物料化器实现自动可视化，比较不同 AI 智能体\n2. **[部署机器学习模型](examples\u002Fdeploying_ml_model\u002F)** - 将传统机器学习模型部署为带有监控和版本控制的生产端点\n3. **[部署智能体](examples\u002Fdeploying_agent\u002F)** - 基于管道、评估和嵌入式 Web UI 的文档分析服务\n4. **[端到端批处理推理](examples\u002Fe2e\u002F)** - 包含特征工程的完整 MLOps 流程\n5. **[LLM RAG 管道](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml-projects\u002Ftree\u002Fmain\u002Fllm-complete-guide)** - 带有评估循环的生产级 RAG\n6. **[代理式工作流（深度研究）](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml-projects\u002Ftree\u002Fmain\u002Fdeep_research)** - 使用 ZenML 协调你的智能体\n7. **[微调管道](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml-projects\u002Ftree\u002Fmain\u002Fgamesense)** - 微调并部署 LLM\n\n## 🗣️ 与你的管道对话：ZenML MCP 服务器\n\n不要再通过仪表板来理解你的机器学习工作流了。**[ZenML MCP 服务器](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fmcp-zenml)** 允许你使用自然语言，通过 Claude Desktop、Cursor 或任何兼容 MCP 的客户端查询你的管道、分析运行情况并触发部署。\n\n```\n💬 “本周哪些管道运行失败了？为什么？”\n📊 “给我看看所有客户流失预测模型的准确率指标”\n🚀 “用生产数据触发最新的欺诈检测管道”\n```\n\n**快速设置：**\n1. 从 [zenml-io\u002Fmcp-zenml](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fmcp-zenml) 下载 `.dxt` 文件\n2. 将其拖入 Claude Desktop 设置中\n3. 添加你的 ZenML 服务器 URL 和 API 密钥\n4. 开始与你的机器学习基础设施对话\n\nMCP（模型上下文协议）集成将你的 ZenML 元数据转化为对话式的洞察，使管道调试和分析变得像提问一样简单。非常适合希望在无需仪表板专业知识的情况下， democratize 访问机器学习运维的团队。\n\n### 🎓 书籍与资源\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fwww.amazon.com\u002FLLM-Engineers-Handbook-engineering-production\u002Fdp\u002F1836200072\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzenml-io_zenml_readme_94a1402b13bb.jpg\" alt=\"LLM 工程师手册封面\" width=\"200\"\u002F>\n  \u003C\u002Fa>\n  \u003Ca href=\"https:\u002F\u002Fwww.amazon.com\u002F-\u002Fen\u002FAndrew-McMahon\u002Fdp\u002F1837631964\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzenml-io_zenml_readme_cf5a2c138a4c.jpg\" alt=\"Python 机器学习工程手册封面\" width=\"200\"\u002F>\n  \u003C\u002Fa>\n  \u003Ca href=\"https:\u002F\u002Fwww.amazon.com\u002FDeepSeek-Practice-fine-tuning-distillation-engineering\u002Fdp\u002F1806020858\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzenml-io_zenml_readme_9f2a1efe61df.jpg\" alt=\"DeepSeek 实践手册封面\" width=\"200\"\u002F>\n  \u003C\u002Fa>\n\u003C\u002Fdiv>\n\n[ZenML](https:\u002F\u002Fzenml.io) 被收录在这些关于生产级 AI 系统的综合指南中。\n\n## 🤝 加入正在构建 AI 未来的人工智能工程师社区\n\n**贡献：**\n- 🌟 [在 GitHub 上为我们点赞](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fstargazers) - 帮助更多人发现 ZenML\n- 🤝 [贡献指南](CONTRIBUTING.md) - 从 [`good-first-issue`](https:\u002F\u002Fgithub.com\u002Fissues?q=is%3Aopen+is%3Aissue+archived%3Afalse+user%3Azenml-io+label%3A%22good+first+issue%22) 开始\n- 💻 [编写集成](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fblob\u002Fmain\u002Fsrc\u002Fzenml\u002Fintegrations\u002FREADME.md) - 添加你喜欢的工具\n\n**保持更新：**\n- 🗺 [公开路线图](https:\u002F\u002Fzenml.io\u002Froadmap) - 了解接下来的计划\n- 📰 [博客](https:\u002F\u002Fzenml.io\u002Fblog) - 最佳实践和案例研究\n- 🎙 [Slack](https:\u002F\u002Fzenml.io\u002Fslack) - 与人工智能从业者交流\n\n## ❓ 来自像你一样的机器学习工程师的常见问题解答\n\n**Q: “我需要重写我的智能体或模型才能使用 ZenML 吗？”**\n\nA: 不需要。只需将现有代码包裹在 `@step` 中即可。你可以继续使用 `scikit-learn`、PyTorch、LangGraph、LlamaIndex 或原生 API 调用。ZenML 协调你的工具，而不是取代它们。\n\n**Q: “这和 LangSmith\u002FLangfuse 有什么不同？”**\n\nA: 它们为 LLM 应用提供了出色的可观测性。而我们则编排你整个 AI 技术栈的 **完整 MLOps 生命周期**。借助 ZenML，你可以在一个统一的框架中管理传统的机器学习模型和 AI 智能体，从开发、评估到生产部署。\n\n**Q: “我可以使用现有的 MLflow\u002FW&B 配置吗？”**\n\nA: 当然可以！ZenML 与 [MLflow](https:\u002F\u002Fdocs.zenml.io\u002Fstacks\u002Fexperiment-trackers\u002Fmlflow) 和 [Weights & Biases](https:\u002F\u002Fdocs.zenml.io\u002Fstacks\u002Fexperiment-trackers\u002Fwandb) 都能无缝集成。你的实验，我们的管道。\n\n**Q: “这不就是多了几个步骤的 MLflow 吗？”**\n\nA: 不是。MLflow 主要用于跟踪实验。而我们则编排整个开发流程——从训练、评估到部署和监控——无论是模型还是智能体。\n\n**Q: “如何将 ZenML 配置为与 Kubernetes 一起使用？”**\n\nA: ZenML 可以通过原生 Kubernetes 编排器、Kubeflow 以及其他基于 K8s 的编排器与 Kubernetes 集成。请参阅我们的 [Kubernetes 编排器指南](https:\u002F\u002Fdocs.zenml.io\u002Fstacks\u002Forchestrators\u002Fkubernetes) 和 [Kubeflow 指南](https:\u002F\u002Fdocs.zenml.io\u002Fstacks\u002Forchestrators\u002Fkubeflow)，以及 [部署文档](https:\u002F\u002Fdocs.zenml.io\u002Fgetting-started\u002Fdeploying-zenml\u002Fdeploy-with-helm)。\n\n**Q: “成本方面呢？我负担不起另一个平台。”**\n\nA: ZenML 的开源版本永久免费。你很可能已经具备所需的基础设施（如 Kubernetes 集群和对象存储）。我们只是帮助你更好地利用这些资源进行 MLOps。\n\n### 🛠 VS Code \u002F Cursor 扩展\n\n直接从编辑器管理管道：\n\n\u003Cdetails>\n  \u003Csummary>🖥️ VS Code 扩展实操演示！\u003C\u002Fsummary>\n  \u003Cdiv align=\"center\">\n  \u003Cimg width=\"60%\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzenml-io_zenml_readme_b1a121a03e70.gif\" alt=\"ZenML 扩展\">\n\u003C\u002Fdiv>\n\u003C\u002Fdetails>\n\n可从 [VS Code Marketplace](https:\u002F\u002Fmarketplace.visualstudio.com\u002Fitems?itemName=ZenML.zenml-vscode) 安装。\n\n## 📜 许可证\n\nZenML 根据 Apache 许可证第 2.0 版的条款进行分发。详情请参阅 [LICENSE](LICENSE)。\n\n---\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fwww.linuxfoundation.org\u002F\">\u003Cimg src=\"docs\u002Fbook\u002F.gitbook\u002Fassets\u002Flf-member-silver.svg\" alt=\"Linux 基金会银牌会员\" height=\"100\"\u002F>\u003C\u002Fa>\n      \n  \u003Ca href=\"https:\u002F\u002Fwww.cncf.io\u002F\">\u003Cimg src=\"docs\u002Fbook\u002F.gitbook\u002Fassets\u002Fcncf-member-silver.svg\" alt=\"CNCF 银牌会员\" height=\"100\"\u002F>\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n\u003Cimg referrerpolicy=\"no-referrer-when-downgrade\" src=\"https:\u002F\u002Fstatic.scarf.sh\u002Fa.png?x-pxid=0fcbab94-8fbe-4a38-93e8-c2348450a42e\" \u002F>","# ZenML 快速上手指南\n\nZenML 是一个专为机器学习（ML）和 AI 工程师设计的开源平台，支持从传统 ML 用例到 LLM 工作流及智能体（Agents）的全流程编排。它允许你编写可运行在任何基础设施后端的工作流（Pipelines），并自动处理容器化、实验追踪和部署等运维复杂性。\n\n## 环境准备\n\n在开始之前，请确保你的开发环境满足以下要求：\n\n*   **操作系统**：Linux, macOS 或 Windows (WSL2 推荐)。\n*   **Python 版本**：Python 3.8 - 3.12。\n*   **包管理工具**：`pip`。\n*   **前置依赖**：建议在一个干净的虚拟环境中进行安装（如 `venv` 或 `conda`），以避免依赖冲突。\n\n> **国内加速提示**：如果下载速度较慢，建议使用国内镜像源（如清华源或阿里源）进行安装。\n\n## 安装步骤\n\n### 1. 安装 ZenML\n\n你可以选择安装完整版本（包含本地服务器功能）或精简客户端版本。对于初学者，推荐安装包含服务器功能的版本以便在本地体验完整流程。\n\n**使用官方源安装：**\n```bash\npip install \"zenml[server]\"\n```\n\n**使用国内镜像源加速安装（推荐）：**\n```bash\npip install \"zenml[server]\" -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n```\n\n> 注：如果仅需客户端连接远程服务器，可运行 `pip install zenml`。\n\n### 2. 初始化项目\n\n在安装完成后，需要在你的项目目录中初始化 ZenML 仓库。这将创建必要的配置文件和目录结构。\n\n```bash\nzenml init\n```\n\n### 3. 启动服务\n\n初始化后，登录并启动本地 ZenML 服务器。这将同时在后台运行服务器并打开 Web 仪表盘，方便你可视化查看管道运行状态。\n\n```bash\nzenml login\n```\n\n*   如果是首次运行，该命令会自动启动本地服务器。\n*   如果需要连接已有的远程服务器，请使用 `zenml login \u003Cserver-url>`。\n\n## 基本使用\n\nZenML 的核心概念是将代码封装为 **Steps（步骤）** 并将它们组合成 **Pipelines（管道）**。以下是一个最简单的示例，展示如何定义并运行一个管道。\n\n### 1. 创建第一个管道脚本\n\n新建一个文件 `run_pipeline.py`，写入以下代码：\n\n```python\nfrom zenml import pipeline, step\n\n@step\ndef load_data() -> str:\n    return \"Hello, ZenML!\"\n\n@step\ndef process_data(data: str) -> str:\n    return f\"Processed: {data}\"\n\n@pipeline\ndef my_first_pipeline():\n    raw_data = load_data()\n    processed_data = process_data(raw_data)\n    return processed_data\n\nif __name__ == \"__main__\":\n    # 运行管道\n    result = my_first_pipeline()\n    print(result)\n```\n\n### 2. 运行管道\n\n在终端中执行该脚本：\n\n```bash\npython run_pipeline.py\n```\n\n### 3. 查看结果\n\n*   **终端输出**：你将看到打印的结果 `Processed: Hello, ZenML!`。\n*   **Web 仪表盘**：运行 `zenml login` 后打开的浏览器页面（通常位于 `http:\u002F\u002F127.0.0.1:8237`）将显示此次运行的详细信息，包括步骤执行情况、日志和元数据。\n\n### 下一步探索\n\n完成上述步骤后，你可以尝试运行官方提供的示例来深入学习：\n\n```bash\n# 克隆仓库获取示例代码\ngit clone https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml.git\ncd zenml\u002Fexamples\u002Fquickstart\n\n# 运行快速入门示例\npython run.py\n```\n\n该示例涵盖了管道、步骤、工件（Artifacts）、快照和部署等核心概念。更多高级用法（如 LLM RAG 管道、Agent 编排）请参考 [ZenML 官方文档](https:\u002F\u002Fdocs.zenml.io\u002F)。","某电商公司的 AI 团队正在构建一个动态定价系统，需要频繁迭代从数据清洗、模型训练到在线推理的完整流程，并逐步引入 LLM 代理进行市场舆情分析。\n\n### 没有 zenml 时\n- **环境地狱与复现困难**：数据科学家在本地笔记本上能跑通的代码，部署到服务器时往往因依赖冲突或路径问题失败，排查环境问题耗费大量时间。\n- **实验追踪混乱**：每次调整超参数或更换数据集，团队依靠手动记录 Excel 表格来管理模型版本和指标，难以回溯哪次提交产生了最佳效果。\n- **基础设施耦合严重**：代码中硬编码了特定的云服务商（如 AWS S3）接口，一旦需要切换到内部集群或另一家云平台，必须重写大量底层逻辑。\n- **协作壁垒高**：算法工程师与运维人员缺乏统一语言，交付过程依赖口头沟通和零散的脚本，导致从实验到生产上线的周期长达数周。\n\n### 使用 zenml 后\n- **一键容器化与标准化**：zenml 自动将管道步骤封装为容器，确保本地开发环境与生产运行环境完全一致，彻底消除了“在我机器上是好的”这类问题。\n- **全链路可观测性**：每一次管道运行都自动记录详细的元数据、日志和性能指标，团队成员可通过可视化界面直接对比不同实验的效果，快速定位最优模型。\n- **基础设施无缝抽象**：通过切换简单的配置栈（Stack），同一套代码即可灵活运行在本地 Docker、Kubernetes 或各大云厂商（如 GCP Vertex、Sagemaker）上，无需修改业务逻辑。\n- **高效协作与快速迭代**：标准化的管道定义让算法与运维团队在同一框架下协作，新策略从构思到上线验证的时间从数周缩短至几天。\n\nzenml 通过统一从流水线到智能代理的开发标准，让企业能够专注于核心算法创新，而非被繁琐的基础设施运维所拖累。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzenml-io_zenml_067280c7.png","zenml-io","ZenML","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fzenml-io_64694f68.png","Building production MLOps tooling.",null,"support@zenml.io","zenml_io","https:\u002F\u002Fzenml.io","https:\u002F\u002Fgithub.com\u002Fzenml-io",[85,89,93,97,101,104,107,111,114,118],{"name":86,"color":87,"percentage":88},"Python","#3572A5",98.2,{"name":90,"color":91,"percentage":92},"Shell","#89e051",0.7,{"name":94,"color":95,"percentage":96},"JavaScript","#f1e05a",0.3,{"name":98,"color":99,"percentage":100},"Go Template","#00ADD8",0.2,{"name":102,"color":103,"percentage":100},"CSS","#663399",{"name":105,"color":106,"percentage":100},"Dockerfile","#384d54",{"name":108,"color":109,"percentage":110},"Jinja","#a52a22",0.1,{"name":112,"color":113,"percentage":110},"HTML","#e34c26",{"name":115,"color":116,"percentage":117},"Jupyter Notebook","#DA5B0B",0,{"name":119,"color":120,"percentage":117},"Mako","#7e858d",5309,598,"2026-04-03T07:33:08","Apache-2.0","未说明 (通常支持 Linux, macOS, Windows)","未说明 (取决于具体工作负载，如训练或推理)","未说明",{"notes":129,"python":130,"dependencies":131},"ZenML 是一个编排框架而非单一模型，因此无固定 GPU\u002F内存硬性要求，资源需求取决于用户运行的具体流水线（如传统 ML、LLM 或 Agent）。核心安装命令为 'pip install \"zenml[server]\"'。支持本地开发模式及连接远程服务器生产模式。可集成 Kubernetes、Kubeflow、Sagemaker、GCP Vertex 等多种基础设施后端。提供 MCP Server 功能以通过自然语言查询流水线状态。","3.8+ (根据 PyPI badge 推断)",[67,132,133,134,135,136,137],"MLflow (可选集成)","LangGraph (可选集成)","Langfuse (可选集成)","scikit-learn (用户代码依赖)","PyTorch (用户代码依赖)","LlamaIndex (用户代码依赖)",[13,51,14,26,15,54],[140,141,142,143,144,67,145,146,147,148,149,150,151,152,153,154,155,156,157,158],"mlops","machine-learning","data-science","production-ready","devops-tools","pipelines","metadata-tracking","deep-learning","pytorch","tensorflow","ml","ai","automl","workflow","llm","llmops","agentops","agents","genai","2026-03-27T02:49:30.150509","2026-04-06T05:27:28.983530",[],[163,168,173,178,183,188,193,198,203,208,213,218,223,228,233,238,243,248,253,258],{"id":164,"version":165,"summary_zh":166,"released_at":167},71543,"0.94.1","\u003C!-- ZENML_GITBOOK_RELEASE_NOTES_START tag=0.94.1 -->\n#### 🎯 管道执行控制\n\n- **暂停和恢复管道运行**：引入了 `zenml.wait(...)`，用于在等待外部输入时暂停动态管道，自动释放资源，直到提供输入为止。运行可以自动恢复（使用支持快照的远程编排器时）或通过 `zenml pipeline runs resume \u003CID>` 手动恢复。[PR #4588](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4588)\n- **重放时覆盖步骤输入**：现在可以在重放管道运行时覆盖步骤输入，从而更灵活地使用不同数据重新运行管道。[PR #4590](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4590)\n\n#### 🔧 材料化器与数据处理\n\n- **数据类材料化器**：新增了一个用于可 JSON 序列化数据类的内置材料化器，使结构化数据在步骤之间传递更加便捷。[PR #4600](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4600)\n- **LakeFS 数据版本控制示例**：新增示例，演示针对 TB 级别数据集的“传递引用而非数据”模式。ZenML 步骤之间交换轻量级的 LakeFS 指针，而实际数据则保留在 LakeFS 中，通过其兼容 S3 的网关进行访问。[PR #4559](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4559)\n\n#### ☁️ 基础设施与部署\n\n- **Helm 环境变量覆盖**：现在可以在 `zenml.environment`、`zenml.secretEnvironment` 以及工作节点部署配置中指定的环境变量，覆盖 Helm Chart 中计算出的设置。[PR #4595](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4595)\n- **Helm 中的机密环境变量**：增加了通过 Helm 将机密环境变量注入 ZenML 服务器部署的支持，而无需将机密写入 `values.yaml` 文件，从而实现更好的 GitOps 流程。[PR #4606](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4606)\n- **Docker 构建参数**：现在在自动生成的 Dockerfile 中，`DockerSettings` 中定义的构建参数会正确地以 `ARG` 指令声明。[PR #4612](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4612)\n\n#### 🔐 身份验证与凭据\n\n- **改进的 GCP 凭据刷新机制**：实现了基于服务连接器逻辑的原生 GCP 凭据刷新，取代了定期检查过期时间的方式，从而提供更可靠的 OAuth2 凭据管理。[PR #4527](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4527)\n\n#### 📊 管道配置\n\n- **步骤参数模式存储**：现在会存储步骤参数规范，以便在触发管道快照时进行更好的模式校验。[PR #4591](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4591)\n\n\u003Cdetails>\u003Csummary>修复\u003C\u002Fsummary>\n\n- **隔离步骤的异常处理**：动态管道中的隔离步骤现在会抛出正确的异常类型，而不是总是将其包裹在 `RuntimeError` 中，从而使错误处理更加直观和一致。[PR #4589](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4589)\n- **Kubernetes API 请求超时**：修复了 Kubernetes 编排器中的类型问题，并 ","2026-03-19T17:11:55",{"id":169,"version":170,"summary_zh":171,"released_at":172},71544,"0.94.0","\u003C!-- ZENML_GITBOOK_RELEASE_NOTES_START tag=0.94.0 -->\r\n### 重大变更\r\n\r\n* 已移除用于旧版触发器、操作和事件源的旧端点及客户端方法。**除非您在代码中显式使用了这些端点或方法，否则这不会对您造成影响。**\r\n* 自定义步骤运算符类型必须实现新的 `submit_step` 和 `get_step_status` 方法，才能与动态流水线配合使用。旧的 `launch` 方法仅作为回退方案在静态流水线中有效。Spark 步骤运算符目前尚不兼容动态流水线。[PR #4515](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4515)\r\n* 由于客户端与服务器之间的不兼容性，使用 `\u003C=0.92.0` 版本的客户端并启用步骤日志时，无法正常运行流水线。\n\n#### 🚀 新集成\r\n\r\n- **Run:AI 步骤运算符**：ZenML 现在支持在 Run:AI 集群上以分数级 GPU 分配方式运行单个流水线步骤，从而更高效地利用资源来处理机器学习工作负载。[PR #4439](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4439)\r\n\r\n#### ✨ 新特性\r\n\r\n- **步骤与流水线重放**：现在您可以使用相同的输入和配置重放现有的步骤或流水线运行。在重放流水线运行时，您可以指定跳过哪些步骤，并从原始运行中复用结果。此外，还提供调试模式，允许您在本地编排器上使用当前堆栈执行重放操作。[PR #4456](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4456)\r\n\r\n- **触发器与原生调度（PRO 版）**：引入了用于自动化流水线执行的 `Trigger` 概念。首个支持的触发器类型是调度，它提供了生命周期管理、与编排器的自动同步以及跨堆栈的集中化管理功能。[PR #4482](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4482)\r\n\r\n- **按版本筛选步骤运行**：新增了按版本筛选步骤运行的功能，便于跟踪和管理特定版本的流水线步骤。[PR #4518](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4518)\r\n\r\n#### 🔧 改进\r\n\r\n- **增强的动态流水线监控**：改进了动态流水线中隔离步骤的执行与监控机制。步骤提交现已与监控分离，避免了步骤执行期间的线程阻塞。[PR #4369](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4369)\r\n\r\n- **SkyPilot 集成更新**：将 SkyPilot 集成更新至 0.11.x 版本，包括迁移到新的异步 API 并支持新的资源设置。[PR #4462](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4462)\r\n\r\n- **Kubernetes 重试配置**：为 Kubernetes 编排器和步骤运算符的 API 调用添加了可配置的超时选项，确保正确的重试行为并防止不必要的挂起。[PR #4525](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4525)\r\n\r\n- **Git 子模块支持**：现在在上传代码归档到制品存储时，会包含 Git 子模块中的文件，从而确保具有子模块的代码库能够完整地进行代码追踪。[PR #4496](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenm","2026-03-04T18:22:00",{"id":174,"version":175,"summary_zh":176,"released_at":177},71545,"0.93.3","\u003C!-- ZENML_GITBOOK_RELEASE_NOTES_START tag=0.93.3 -->\n#### 性能改进\n\n此版本对 ZenML 服务器进行了显著的性能优化，尤其是在处理大规模部署时：\n\n- **数据库查询效率提升**：重写了过滤查询，消除了计数过程中不必要的排序；移除了多列上的低效 DISTINCT 语句，并优化了 OR 子查询，以提升大规模场景下的数据库性能。[PR #4449](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4449)\n\n- **API 事务管理增强**：将过期事务的清理工作移至一个独立的后台线程，该线程会定期运行，从而显著提升了 API 响应时间，特别是在处理包含大量步骤的管道快照等大负载请求时。[PR #4453](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4453)\n\n#### 日志功能增强\n\n日志功能得到了扩展，新增了多项特性和改进：\n\n- 为日志新增了 `create` 和 `update` 端点，并支持在 `StepRunRequest` 和 `PipelineRunRequest` 中使用 UUID。\n- 在管道运行的日志元数据中引入了工作空间 ID 和名称（保持向后兼容性）。\n- 在错误信息中添加了 `zenml.event.type` 字段，以便更好地追踪上下文。\n- 引入了一个环境变量来管理每次请求的最大日志条目数。\n- 修复了元数据键格式不一致的问题（统一使用 `zenml.` 前缀）。[PR #4405](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4405)、[PR #4467](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4467)\n\n#### 仪表板更新\n\n- 在 DAG 可视化中为步骤节点增加了已用时间显示，以更好地监控管道执行情况。[PR #994](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml-dashboard\u002Fpull\u002F994)\n\n\u003Cdetails>\u003Csummary>修复\u003C\u002Fsummary>\n\n- **严重数据丢失漏洞**：修复了 `download_artifact_files_from_response` 中的一个关键问题，该问题会导致下载大于 8KB 的工件时出现无声的数据损坏。此漏洞会使大工件最多损失 98% 以上的数据，仅保留最后一块数据。[PR #4422](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4422)\n\n- **ZenML Pro 迁移问题**：修复了一个问题，即当用户通过加入组织的方式将 ZenML OSS 服务器迁移到 ZenML Pro 后，本地用户账户的 Cookie 仍会残留，从而导致无法在 UI 中访问已迁移的资源。现在服务器会正确拒绝这些过时的 Cookie。[PR #4473](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4473)\n\n- **仅使用 uv 的环境**：修复了在仅使用 `uv` 而未安装 `pip` 的环境中管道运行崩溃的问题。ZenML 现在会在无法使用 `pip freeze` 收集环境元数据时，回退到 `uv pip freeze`。此外，还为 Docker 构建添加了 `UV_FREEZE` 导出方式。[PR #4484](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4484)\n\n- **Kubernetes 凭证过期问题**：修复了一个问题，即由服务连接器颁发的 Kubernetes 凭证在监控长时间运行的任务时会过期，从而导致监控失败。现在，在任务监控期间会正确刷新凭证。[PR #4493](https:\u002F\u002Fgit","2026-02-19T09:59:12",{"id":179,"version":180,"summary_zh":181,"released_at":182},71546,"0.93.2","\u003C!-- ZENML_GITBOOK_RELEASE_NOTES_START tag=0.93.2 -->\r\n#### 🎨 仪表板增强功能\r\n\r\nZenML 仪表板现在为您提供了更清晰的管道和基础设施视图：\r\n\r\n- **下载管道代码**：您现在可以直接从仪表板下载用于某个管道快照的代码。在“管道运行”详情页和“步骤”详情表中，“代码路径”部分都会显示一个新的“下载”按钮，方便您获取并审查实际执行的代码。[PR #4401](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4401)、[PR #989](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml-dashboard\u002Fpull\u002F989)\r\n\r\n- **异常信息展示**：当动态管道运行失败时，仪表板现在会显示详细的异常信息，帮助您快速诊断和排查问题。[PR #4395](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4395)、[PR #990](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml-dashboard\u002Fpull\u002F990)\r\n\r\n- **堆栈与组件标签**：附加到堆栈和组件上的标签现在可以在仪表板中查看，从而更轻松地组织和识别您的基础设施资源。[PR #992](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml-dashboard\u002Fpull\u002F992)\r\n\r\n#### 🔄 动态管道改进\r\n\r\n动态管道现在更加健壮且易于使用：\r\n\r\n- **正确的环境配置**：在运行动态管道的入口函数时，管道环境现在会被正确设置，从而确保在不同执行环境中的一致性行为。[PR #4420](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4420)\r\n\r\n#### 🤖 开发者体验\r\n\r\n- **Claude Code 插件**：ZenML 推出了一项针对 Claude Code 的 Quick Wins 技能，可帮助您在 AI 辅助编码工作流中直接实施 MLOps 最佳实践。该插件可通过 Claude Code 插件市场获得，并包含针对多种 AI 编码工具的完整文档。[PR #4426](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4426)\r\n\r\n\u003Cdetails>\u003Csummary>修复内容\u003C\u002Fsummary>\r\n\r\n#### 🚀 性能与可扩展性\r\n\r\n- **工件下载修复**：解决了由于下载端点上的 RBAC 检查不正确而导致工件版本下载失败的问题。[PR #4401](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4401)\r\n\r\n\u003C\u002Fdetails>\r\n\u003C!-- ZENML_GITBOOK_RELEASE_NOTES_END tag=0.93.2 -->\r\n\r\n## 变更内容\r\n* 在旧版文档中添加版本 0.93.0，由 @github-actions[bot] 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4391 中完成\r\n* 将 0.93.1 添加到迁移测试中，由 @github-actions[bot] 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4393 中完成\r\n* 添加关于调度激活\u002F停用及归档的文档，由 @strickvl 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4396 中完成\r\n* 修复发布流程，由 @schustmi 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4394 中完成\r\n* 禁用 HTTP 下使用安全 Cookie 的功能，由 @stefannica 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4399 中完成\r\n* 修正有关 ZenML 不支持调度更新的错误说法，由 @strickvl 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4400 中完成\r\n* 修复损坏的 Neptune 文档链接","2026-01-29T22:30:18",{"id":184,"version":185,"summary_zh":186,"released_at":187},71547,"0.93.1","\u003C!-- ZENML_GITBOOK_RELEASE_NOTES_START tag=0.93.1 -->\r\n\r\n#### 🎛️ 调度管理增强\r\n\r\n现在您可以直接通过 CLI **暂停和恢复管道调度**，从而更好地控制自动化管道执行。使用新命令可按需启用或禁用调度：\n\n```bash\nzenml pipeline schedule deactivate \u003Cschedule_id>\nzenml pipeline schedule activate \u003Cschedule_id>\n``` \n\n此功能目前适用于 Kubernetes 编排器。[PR #4328](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4328)\n\n调度现支持作为软删除操作的**归档**功能。当您删除某个调度时，它将被归档而非永久移除，从而保留历史引用，确保您的管道运行仍能维持其调度关联。[PR #4339](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4339)\n\n#### 🖥️ 仪表板改进\r\n\r\n**堆栈管理**：您现在可以直接从 UI 更新现有堆栈，而无需先删除再重新创建。新增的专用堆栈更新页面使您可以高效地添加或替换堆栈组件（编排器、制品存储、容器注册表等）。[PR #978](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml-dashboard\u002Fpull\u002F978)\n\n**步骤缓存管理**：您可直接从步骤详情面板查看并管理步骤缓存的过期时间。缓存过期字段会显示步骤缓存何时到期（若未设置过期时间则显示“永不”），已过期的缓存也会清晰标记。此外，您还可通过单击操作手动使步骤缓存失效。[PR #976](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml-dashboard\u002Fpull\u002F976)\n\n**日志体验增强**：管道运行现设有专门的日志页面，并配有侧边栏，方便在运行级别和步骤级别日志之间切换。新的日志查看器采用虚拟化渲染技术，以提升对大型输出的性能；同时提供搜索和筛选功能，并显示步骤执行时长。[PR #985](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml-dashboard\u002Fpull\u002F985)\n\n#### ⚡ 性能与可靠性\r\n\r\n**Kubernetes 编排器改进**：Kubernetes 编排器现已通过可配置的 DAG 运行器工作进程、优化的缓存候选获取以及针对失败步骤 Pod 的更佳错误处理机制，实现更高效率的运行。[PR #4368](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4368)\n\n**数据库备份速度**：全新的 mydumper\u002Fmyloader 备份策略显著提升了操作效率：\n- 数据库备份速度提升至原来的 **30 倍**\n- 数据库恢复速度提升至原来的 **2.5 倍**\n- 存储空间需求降低至原来的 **10 分之一**\n\n[PR #4358](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4358)\n\n#### 🚀 编排器功能\r\n\r\n**AzureML 动态管道**：动态管道现已完全支持 AzureML 编排器，进一步扩展了您灵活执行管道的选择范围。[PR #4363](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4363)\n\n**Kubernetes 初始化容器模板化**：在为 Kubernetes 编排器配置初始化容器时，您现在可以使用 `\"{{ image }}\"` 占位符，该占位符将自动替换为实际的编排","2026-01-14T09:20:00",{"id":189,"version":190,"summary_zh":191,"released_at":192},71548,"0.93.0","## ⚠️ 注意\n如果您的数据库中包含没有任何运行的管道，那么此版本中的一个数据库迁移将会失败。请确保在将服务器升级到此版本之前，删除所有此类管道。\n```python\nfrom zenml.client import Client\nfrom zenml.utils import pagination_utils\n\nfor pipeline in pagination_utils.depaginate(Client().list_pipelines):\n  if pipeline.latest_run_status is None:\n    Client().delete_pipeline(pipeline.id)\n```\n\n## ⚠️ 破坏性变更\n\n- `PipelineRunResponse.logs` 字段已被移除。请改用 `PipelineRunResponse.log_collection`。\n- `StepRunResponse.logs` 字段已被移除。请改用 `StepRunResponse.log_collection`。\n- `\u002Fapi\u002Fv1\u002Fpipelines\u002F\u003CID>\u002Fruns` 端点已被移除。请改用 `\u002Fapi\u002Fv1\u002Fruns?pipeline=\u003CID>`。\n\n## 新特性\n\n### 日志存储\n日志存储是新的堆栈组件，用于控制管道和步骤日志的持久化位置。目前提供三种类型：Artifact Store（将日志写入堆栈的 Artifact Store）、通用 OpenTelemetry (OTEL) 和 Datadog。如果堆栈未配置日志存储，则默认使用 Artifact Store 类型。\n\n### 其他\n- CLI 表格已得到改进。现在可以指定列和输出格式（例如 CSV 或 JSON）。例如：`zenml pipeline runs list --columns id,index,name --output json`。\n- 每个管道运行现在都会公开一个在其所属管道内唯一的递增索引，从而更易于区分和引用运行。\n- 调度器现在支持所有类型的优雅停止。您可以通过 UI 或 CLI 调用：`zenml pipeline runs stop \u003CID> --graceful`。\n- AzureML 调度器和步骤运算符现在支持可配置的共享内存大小。\n- 在按相关实体过滤对象时，现在可以使用 `oneof` 过滤运算符：`zenml pipeline runs list --pipeline='oneof:[\"some_pipeline\", \"other_pipeline\"]'`。\n- UI：现在可以在 DAG 可视化工具中查看管道代码。\n\n## 错误修复\n- 在 AzureML 上恢复 MLflow 运行现在可以可靠地工作。\n- Kubernetes 调度器中的调度状态不一致问题已解决。\n- 运行管道的最新快照现在会选择正确的快照。\n- 从步骤内部创建快照的功能现已恢复正常。\n- 在获取日志以及下载工件数据或可视化内容时，已添加缺失的 RBAC 检查。这些端点现在会执行适当的日志和工件存储 RBAC 检查。\n\n## 变更内容\n* 将版本 0.91.2 添加到旧版文档中，由 @github-actions[bot] 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4283 中完成。\n* 将 0.92.0 添加到迁移测试中，由 @github-actions[bot] 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4285 中完成。\n* 添加工作流以要求 PR 上带有发布标签，由 @strickvl 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4281 中完成。\n* 改进 CLI 表格和 ID 处理，由 @safoinme 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4241 中完成。\n* 将静态\u002F动态管道纳入分析中，由 @schustmi 在 https:\u002F\u002Fgithub.com\u002Fz","2025-12-16T09:19:32",{"id":194,"version":195,"summary_zh":196,"released_at":197},71549,"0.92.0","## ⚠️ 重大变更\n\n* `StepRunResponse.regular_inputs` 和 `StepContext.inputs` 现在的类型为 `Dict[str, List[ArtifactVersionResponse]]`，而非 `Dict[str, ArtifactVersionResponse]`。\n\n## 新特性\n\n* 本地镜像构建器现在可以使用 Docker CLI 而不是 Python SDK，从而启用 Python SDK 不支持的 BuildKit 选项。\n* Evidently 集成已更新，现支持 NumPy 2.0。\n* 登录时可通过 `--project` 标志指定当前项目（例如，`zenml login ... --project ...`）。\n* GCP 镜像构建器现支持配置构建发生的区域。\n\n## 问题修复\n* 运行元数据现在可以存储集合和元组。\n* 已修复在未安装 `local` extra 的情况下连接到 ZenML 服务器时出现的错误。\n* 已修复当为 Weights & Biases 实验跟踪器配置 `enable_weave=False` 时，Weave 导入产生的副作用问题。\n\n## 变更内容\n* @github-actions[bot] 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4213 中将版本 0.91.1 添加到旧版文档中。\n* @github-actions[bot] 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4215 中将 0.91.2 添加到迁移测试中。\n* @bcdurak 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4208 中修复了 `mlflow` 测试。\n* @schustmi 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4217 中增加了 CI 构建机的磁盘空间。\n* @htahir1 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4226 中修复了 Azure ML 计算集群验证中的拼写错误。\n* @stefannica 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4235 中固定了 Sagemaker 版本，以避免引入庞大的 3.0 依赖项。\n* @htahir1 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4228 中修复了 AzureMLOrchestratorSettings 类中可选参数导致的 AttributeError。\n* @htahir1 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4229 中修复了带有已弃用标志的 CLI 登录错误信息。\n* @strickvl 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4244 中将 `@` 命令调用替换为 `\u002F` 调用。\n* @strickvl 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4245 中将 DeepSeek Packt 书籍的缩略图和链接添加到 `README` 文件中。\n* @Json-Andriopoulos 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4216 中实现了 Feature:4151 心跳激活逻辑。\n* @stefannica 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4232 中实现了 AWS\u002FSagemaker 动态流水线。\n* @htahir1 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4233 中为 flavors 添加了 display_name 字段，以改善 UI 展示效果。\n* @htahir1 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4231 中将流水线部署功能加入到入职流程中。\n* @schustmi 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4159 中实现了动态流水线的 Map\u002FReduce 实现。\n* @schustmi 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4250 中修复了 alembic 的顺序问题。\n* [进行中] @htahir1 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4249 中修复了元数据系统中集合类型的序列化问题。\n* @htahir1 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4243 中更新了 Evidently 集成，以支持 NumPy 2.0。\n* @adam6878 添加了登录时设置项目的选项。","2025-12-02T06:42:19",{"id":199,"version":200,"summary_zh":201,"released_at":202},71550,"0.91.2","## 新特性\n\n### Kubernetes 部署器\n- 新的部署器实现，允许您在 Kubernetes 上部署管道\n\n### 其他\n- 支持 Mlflow 3.0\n\n## 错误修复\n\n- S3 资产存储现在再次支持自定义后端\n- 为 RestZenStore 传递内联 SSL 证书的功能现已正常工作\n- Weights & Biases 实验跟踪器现在不会因管道运行名称超过标签最大长度而失败\n\n\n## 变更内容\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4164 中，由 @github-actions[bot] 将版本 0.91.0 添加到旧版文档中\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4167 中，由 @stefannica 修复了远程部署中的快照缺失问题\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4166 中，由 @github-actions[bot] 将 0.91.1 添加到迁移测试中\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4169 中，由 @strickvl 向变更日志验证工作流添加成功注释\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4171 中，由 @schustmi 修复了默认 Kubernetes 资源的应用问题\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4176 中，由 @htahir1 修复了部署暂存工作区工作流中的错误\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4168 中，由 @schustmi 修复了发布 Docker 构建工作流\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4143 中，由 @Json-Andriopoulos 修复了破坏 CI 检查（代码风格检查和集成测试）的问题\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4147 中，由 @strickvl 对 FiftyOne 的 CV 进行改进（谱系\u002F数据集版本控制）\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4172 中，由 @strickvl 添加自定义命令、评审子代理和 Claude 工作流\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4187 中，由 @strickvl 修复了容器化和 Vertex 编排器文档中的错误\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4181 中，由 @schustmi 修复了无步骤模板的动态管道的代码上传逻辑\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4188 中，由 @htahir1 修复了 SSL 证书验证中的 UnboundLocalError 错误\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4178 中，由 @schustmi 引入缓存键索引，以加快缓存查找速度\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4173 中，由 @htahir1 为动态管道实验性功能添加文档\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4160 中，由 @htahir1 添加对 MLflow 3.0 的支持\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4195 中，由 @strickvl 授予 Claude 运行格式化脚本的权限\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4186 中，由 @strickvl 修复了自定义 S3 后端的 S3 资产存储端点配置\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4200 中，由 @strickvl 修复了 Claude GitHub Actions 工作流配置\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4199 中，由 @bcdurak 对问题模板进行了更改\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4194 中，由 @schustmi 改进了多步构建过程中本地镜像的检测\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4189 中，由 @schustmi 对 wandb 标签进行了清理\n* 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4173 中，由 @htahir1 添加了一个工作流，用于在 PR 闲置 4 周后自动关闭 stale PR","2025-11-19T06:31:25",{"id":204,"version":205,"summary_zh":206,"released_at":207},71551,"0.91.1","## 新特性\n\n### Huggingface 部署器\n- 新的部署器实现，允许您在 Huggingface 上部署您的流水线。\n\n### Kubernetes 编排器\n- 新增配置选项，用于指定容器的安全上下文。\n- 新增配置选项，可用于跳过创建所有者引用。\n\n### 其他\n- HashiCorp Vault 秘密存储新增认证选项。\n- 支持较新的 Databricks 版本。\n- 引入动态流水线 v1。此功能目前仍处于高度实验阶段，不建议用于生产环境，我们诚邀您提供早期反馈。更多详情请参阅[这些文档](https:\u002F\u002Fzenml-io.gitbook.io\u002Fhamzsterdam-docs\u002FNTXnvHIs4p2Trs9Bz26G\u002Fconcepts\u002Fsteps_and_pipelines\u002Fdynamic_pipelines)。\n\n## Bug 修复\n\n- 修复本地部署中的端口复用问题。\n- 修复并行部署调用的问题。\n- 修复监控同步运行时的键盘中断处理问题。\n- 修复更新堆栈组件等实体名称时的大小写敏感性问题。\n\n## 已知问题\n\n- 当 Kubernetes 编排器配置为 `pass_zenml_token_as_secret=True`，且 `orchestrator_pod_settings.env` 中未包含任何值时，编排器将无法正确传递 Secret。作为此版本的临时解决方案，您可以：\n  - 将 `pass_zenml_token_as_secret` 设置为 `False`；\n  - 或配置一个虚拟环境变量（例如使用以下 CLI 命令：`zenml orchestrator update --orchestrator_pod_settings='{\"env\": [{\"name\": \"DUMMY\", \"value\": \"\"}]}'`）。\n\n## 变更内容\n* @github-actions[bot] 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4094 中将版本 0.90.0 添加到旧版文档中。\n* @htahir1 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4077 中添加了经典机器学习示例。\n* @strickvl 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4086 中向代理框架集成中添加了 `Qwen-Agent` 示例。\n* @github-actions[bot] 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4096 中将 0.91.0 添加到迁移测试中。\n* @htahir1 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4090 中更新了代理框架集成。\n* @htahir1 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4100 中进行了大规模的消息变更。\n* @htahir1 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4103 中更新了 hello-world.md 中的流水线执行说明。\n* @stefannica 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4097 中进行了多项部署修复和改进。\n* @stefannica 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4106 中向流水线运行元数据中添加了触发信息。\n* @htahir1 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4109 中为天气代理示例添加了核心依赖项。\n* @stefannica 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4110 中增加了更多 Hashicorp Vault 秘密存储的认证方法。\n* @schustmi 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4098 中修复了发布 Cloud Build 的问题。\n* @strickvl 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4115 中更新了代理文档，以避免不必要的代码注释。\n* @htahir1 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fz 中提供了结合 FiftyOne 和 Ultralytics 的计算机视觉示例。","2025-11-11T12:43:05",{"id":209,"version":210,"summary_zh":211,"released_at":212},71552,"0.91.0","\n此版本带来了重大的部署增强功能（本地部署工具、完全可自定义的部署服务器和部署可视化）、强大的缓存控制机制（基于文件\u002F对象的缓存失效、缓存过期以及自定义缓存函数）、对 Python 3.13 的支持、MLX 数组实例化器，以及在构建过程中自定义镜像标签的新选项。同时，该版本弃用了旧版模型部署工具，并停止对 Python 3.9 的支持。\n\n## ⚠️ 重大变更\n\n- 停止对 Python 3.9 的支持\n\n## 弃用内容\n\n- 模型部署工具已被弃用；请改用部署工具来部署整个流水线\n\n## 新特性\n\n### 部署\n- 本地部署工具，用于在本地部署流水线\n- 部署服务器可通过设置进行完全自定义\n- 可为部署附加自定义可视化组件\n\n### 缓存\n- 可指定使某步骤缓存失效的文件或 Python 对象\n- 缓存过期功能，用于限制缓存的有效期\n- 自定义缓存函数，用于实现高级的缓存失效逻辑\n\n### 其他\n- MLX 数组实例化器\n- 支持 Python 3.13\n- 可自定义构建的 Docker 镜像标签\n\n## 错误修复\n\n- 修复了使用 `numba` 时打印捕获的不兼容问题\n- 修复了 Hashicorp Vault 秘密存储中的挂载点配置问题\n\n## 变更内容\n* 由 @github-actions[bot] 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4029 中将版本 0.85.0 添加到旧版文档\n* 由 @github-actions[bot] 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4031 中将 0.90.0 添加到迁移测试中\n* @stefannica 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4017 中记录了流水线部署和部署工具的相关文档\n* @schustmi 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4023 中进行了文档快照\n* @safoinme 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3960 中更新了已部署流水线的示例\n* @schustmi 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4032 中引用了 Helm Chart 的描述\n* @nicholasjng 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4027 中添加了 MLX 数组实例化器\n* @schustmi 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4035 中改进了 CLI 错误信息\n* @strickvl 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4039 中阻止了在 Intel Mac 上安装 MLX 集成\n* @bcdurak 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4052 中修复了日志中的时间戳问题\n* @Json-Andriopoulos 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4051 中修复了破坏 `mypy` 检查的问题\n* @Json-Andriopoulos 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3990 中扩展了列表流水线过滤器\n* @Json-Andriopoulos 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4025 中添加了 Docker 镜像标签设置\n* @schustmi 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4041 中对文档进行了小幅改进\n* @stefannica 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4056 中修复了 AWS 部署工具的秘钥标签和日志问题\n* @schustmi 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4034 中修复了快照独占标签的问题\n* @bcdurak 在 https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F4053 中移除了对 Python 3.9 的支持，并添加了对 Python 3.13 的支持\n* @bcdurak 在 https:\u002F\u002Fgithub.co","2025-10-25T03:08:07",{"id":214,"version":215,"summary_zh":216,"released_at":217},71553,"0.90.0","This release introduces pipeline snapshots and pipeline deployments, refactors the base package dependencies, and adds runtime environment variable support.\r\n\r\n## ⚠️ Breaking Changes\r\n\r\n- **Client-Server Compatibility**: Not compatible with previous versions. Client and server must be upgraded simultaneously.\r\n- **Run Templates**: Existing run templates need to be recreated as they contain previous Client versions in their container images.\r\n- **Base Package**: The `zenml` package no longer includes local database dependencies. Install `zenml[local]` if you want to use ZenML without a local\u002Fremote server.\r\n\r\n## Deprecations\r\n\r\n- **Run Templates**: Deprecated in favor of pipeline snapshots. Snapshots can be triggered from server\u002Fdashboard and deployed as endpoints.\r\n- **ZenML Pro On-Prem deployments only**: The `ZENML_SERVER_MAX_CONCURRENT_TEMPLATE_RUNS` server environment variable is deprecated. Use `ZENML_SERVER_MAX_CONCURRENT_SNAPSHOT_RUNS` instead.\r\n- **Stack Component API**: The `StackComponent.prepare_pipeline_deployment(...)` method has been removed and will not be called anymore. This might be relevant if you've implemented custom stack components that rely on this method being called.\r\n\r\n## New Features\r\n\r\n### Pipeline Snapshots\r\n- Capture immutable snapshots of pipeline code, configuration, and container images\r\n- Execute snapshots from server\u002Fdashboard without local code environment\r\n- Can be deployed\r\n\r\n### Pipeline Deployments\r\n- Add a new `Deployer` stack component and Docker, AWS and GCP implementations\r\n- Deploy pipelines as HTTP endpoints for online inference or agentic usecases\r\n- Add pipeline init\u002Fcleanup hooks for quicker deployment runs\r\n\r\n### Runtime Environment Variables\r\n- Configure environment variables when running pipelines\r\n- Environment variables can either be defined directly or from ZenML secrets\r\n\r\n### Dependency Management\r\n- Move local database dependencies to `local` package extra\r\n- Add support for latest Pydantic and CloudPickle versions\r\n- Reduce base package dependencies\r\n\r\n### Jax Support\r\n- Add materializer for Jax arrays\r\n\r\n## What's Changed\r\n* Add 0.85.0 to the migration tests by @github-actions[bot] in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3951\r\n* Add version 0.84.3 to legacy docs by @github-actions[bot] in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3949\r\n* Add documentation for timeline view by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3948\r\n* Add documentation for the `StepContext` object by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3953\r\n* Runtime environment variables by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3336\r\n* Fix legacy merge conflict in docs by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3955\r\n* Docs\u002Fdocumentation for execution modes by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3956\r\n* Pipeline snapshots by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3856\r\n* Upgrade pydantic by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3944\r\n* Dedent exception info traceback by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3957\r\n* Update (broken) metadata logging example for pipeline runs by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3961\r\n* Fix cache policy docs formatting by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3965\r\n* Bugfix for duplicated log records for exceptions by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3964\r\n* Bugfix for handling old logs with timestamp by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3968\r\n* Snapshot UX improvements by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3959\r\n* Track pipeline output spec and schema by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3972\r\n* Misc snapshot improvements by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3980\r\n* Update self-hosted ZenML Pro deployment docs by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3981\r\n* Slimmer base package by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3916\r\n* Cleanup base package workflow by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3983\r\n* Rename some legacy attributes by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3934\r\n* Add JAX array materializer by @nicholasjng in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3712\r\n* Fix requirements for local agentic examples tests by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3986\r\n* Fix setting a snapshot name to None by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3985\r\n* Improve error message for duplicate pipeline run names by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3701\r\n* Fix Terraform documentation URLs and improve link checker robustness by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3988\r\n* Fix formatting errors on embeddings finetuning docs by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3979\r\n* Remove remaining RBAC check when fetching tags by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3989\r\n* Remove dependency on pydantic_se","2025-10-02T07:02:40",{"id":219,"version":220,"summary_zh":221,"released_at":222},71554,"0.85.0","The `0.85.0` release delivers powerful **pipeline execution enhancements** and **caching improvements** that provide users with greater control over pipeline behavior.\r\n\r\n## ⚠️ Breaking Changes\r\n\r\n- **Local Orchestrator Behavior**: The local orchestrator will now continue executing steps after some steps fail (instead of stopping execution immediately)\r\n- **Docker Package Installer**: Default Python package installer in Docker settings switched from `pip` to `uv` \r\n- **Log Endpoint Format**: Log endpoints return a different format - affects manual API access but not regular pipeline operations\r\n- **Fetching Runs with old Client**: Using a client of a previous release (`\u003C0.85.0`) to fetch pipeline runs created by a client with version `>=0.85.0` can cause an error if the run is in\r\nthe new `provisioning` status.\r\n\r\n## 📢 Upcoming Breaking Change\r\n\r\n**Important Notice for Future Release**: In the next release, the base `zenml` package will no longer include dependencies for running ZenML connected to a local database.. Users will only be able to connect to deployed ZenML servers with the base package.\r\nTo continue using ZenML locally with a SQLite database, install with the `local` extra: `pip install 'zenml[local]'`. If you're using ZenML with a local server, you're already installing `zenml[server]` and this change will not affect you.\r\n\r\n## 🚀 New Features\r\n\r\n### Pipeline Execution Modes\r\n- **Flexible Failure Handling**: Configure what happens to a pipeline run when any step fails, providing fine-grained control over pipeline execution behavior\r\n\r\n### Advanced Caching System\r\n- **Value-Based Caching**: Materializers now support caching artifacts based on their actual content\u002Fvalue rather than just artifact ID, enabling more intelligent cache reuse\r\n- **Cache Policies**: New cache policy system allows users to specify precisely when a step should be cached, providing granular control over caching behavior\r\n\r\n### Airflow 3.0 support\r\n- **Airflow 3.0 Compatibility**: Support for Apache Airflow 3.0, ensuring compatibility with the latest Airflow features and improvements\r\n\r\n\r\n## What's Changed\r\n* Add version 0.84.2 to legacy docs by @github-actions[bot] in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3910\r\n* Made connection docs easier to read by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3896\r\n* Add Google ADK framework example and `DockerSettings` for all examples by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3912\r\n* Delete run if model version doesn't exist during creation by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3915\r\n* Caching by value by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3900\r\n* Add start time to step node metadata by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3921\r\n* Prevent stopping runs without orchestrator run id by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3923\r\n* Airflow 3 support by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3922\r\n* fix: typo in core-concepts.md by @mhmunem in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3926\r\n* Run template config improvements by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3918\r\n* Add provisioning execution status by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3924\r\n* Fix unnecessary code upload by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3932\r\n* Allow cloudpickle>3.x by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3914\r\n* Switch to uv as default package installer by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3935\r\n* Fix unit test by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3945\r\n* Efficient queries for execution mode changes by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3942\r\n* Different pipeline execution modes by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3874\r\n* Changes to the fetch logs endpoints by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3845\r\n* Adding verbosity levels to log messages in storage by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3812\r\n\r\n## New Contributors\r\n* @mhmunem made their first contribution in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3926\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fcompare\u002F0.84.3...0.85.0\r\n","2025-09-12T11:24:47",{"id":224,"version":225,"summary_zh":226,"released_at":227},71555,"0.84.3","The `0.84.3` release introduces **ZenML Pro service account authentication** support and includes important **Kubernetes integration fixes**. This release enhances authentication flexibility for automated workflows and improves the reliability of Kubernetes-based orchestration.\r\n\r\n## 🚀 New Features\r\n\r\n### ZenML Pro Service Account Authentication\r\n- **CLI Login Support**: Added ability to authenticate with ZenML Pro using service account API keys via `zenml login --api-key`\r\n- **Programmatic Access**: Service account API keys can now be used for programmatic access to ZenML Pro workspaces\r\n- **Organization-Level Access**: Service accounts provide access to all workspaces within an organization for automated workflows\r\n\r\n## 🛠️ Fixes\r\n\r\n### Kubernetes Integration\r\n- **Name Sanitization**: Fixed Kubernetes resource name sanitization to properly handle special characters and ensure compliance with Kubernetes naming requirements\r\n\r\n### Dependencies\r\n- **Click Version**: Relaxed Click dependency version constraints to improve compatibility with other packages\r\n\r\n## What's Changed\r\n* Add version 0.84.1 to legacy docs by @github-actions[bot] in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3885\r\n* Fix PyPI download stats badge by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3882\r\n* Don't include decorator code in step cache key computation by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3880\r\n* Add an agent in production example by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3890\r\n* Fix kubernetes name sanitization by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3887\r\n* Fix SyntaxWarning in `exception_utils` regex pattern by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3892\r\n* Add agent framework integration examples by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3898\r\n* Add weekly agent examples test workflow by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3902\r\n* Add ability to use a ZenML Pro API key with `zenml login` by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3895\r\n* Add docs for workload manager env vars by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3907\r\n* Bump allowed versions of click by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3905\r\n* Allow workspace service accounts and API keys to be created by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3908\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fcompare\u002F0.84.2...0.84.3","2025-08-27T15:10:20",{"id":229,"version":230,"summary_zh":231,"released_at":232},71556,"0.84.2","The `0.84.2` release brings significant **performance improvements** and **reliability enhancements** that strengthen ZenML's orchestration capabilities. This release features a major refactor of the Kubernetes orchestrator with enhanced restart capabilities, faster pipeline compilation for large pipelines, critical security fixes, and improved logging performance.\r\n\r\n## Kubernetes Orchestrator\r\n\r\n- **Enhanced Robustness**: Complete rework to use Jobs instead of raw pods for orchestrator execution with state reconstruction support, enabling automatic restarts for a much more robust orchestration experience\r\n- **Deprecated Settings**: Several settings attributes are now deprecated due to the shift from pods to Jobs, streamlining the configuration interface\r\n- **Updated Logging Behavior**: The orchestrator pod no longer streams logs from step pods directly, improving performance and resource usage\r\n\r\n## 🚀 Improvements\r\n- **Faster Pipeline Compilation**: Significantly improved pipeline compilation performance for large pipelines\r\n\r\n## 🛠️ Fixes\r\n- **Run Creation Deadlock**: Fixed deadlock that could happen if two steps try to create a run at the same time\r\n- **Security Enhancement**: Enhanced path materializer to properly validate symlinks and hardlinks to prevent path traversal attacks\r\n- **Logging Performance**: Improved logging thread performance by avoiding unnecessary sleep when shutdown is requested\r\n- **WandB Integration**: Fixed WandB experiment tracker flavor initialization when the integration is not installed\r\n- **Type Annotations**: Fixed type annotations for `experiment_tracker` and `step_operator` parameters to accept both string and boolean values\r\n\r\n\r\n## What's Changed\r\n* Add version 0.84.0 to legacy docs by @github-actions[bot] in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3863\r\n* Add 0.84.1 to the migration tests by @github-actions[bot] in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3865\r\n* Fix wandb flavor needing integration to be installed by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3871\r\n* Verify symlinks and hardlinks in the path materializer by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3870\r\n* Don't sleep in logging thread if shutdown was requested by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3872\r\n* Fix run creation deadlock by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3876\r\n* Fix nonlinear scaling of pipeline compilation time by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3873\r\n* Fix type annotations by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3878\r\n* Orchestrator pod restarts by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3869\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fcompare\u002F0.84.1...0.84.2","2025-08-06T19:18:12",{"id":234,"version":235,"summary_zh":236,"released_at":237},71557,"0.84.1","The `0.84.1` release delivers important **stability improvements** and **feature enhancements** that strengthen ZenML's orchestration capabilities and developer experience. This release focuses on enhanced Kubernetes orchestrator management with schedule support and better error handling, improved step exception collection for debugging, external service account support, dynamic fan-out\u002Ffan-in patterns for run templates, and critical fixes for Vertex step operators and logging reliability.\r\n\r\n## 🚀 Improvements\r\n\r\n- **Step Exception Handling**: Improved collection of step run exception information for better debugging\r\n- **External Service Accounts**: Added support for external service accounts for improved flexibility\r\n\r\n## Kubernetes Orchestrator\r\n- **Enhanced Schedule Management**: Added schedule management capabilities (updating and deleting) for the Kubernetes orchestrator\r\n- **Better Error Handling**: Added explicit Kubernetes failure reasons to logs and improved pod monitoring\r\n- **Secret Cleanup**: Fixed cleanup of Kubernetes secrets when orchestrator pods fail to start\r\n\r\n## 🛠️ Fixes\r\n- **Vertex step operator credential refresh**: Fixed retry logic and credential refresh for the Vertex step operator\r\n- **Logging Fixes**: Resolved race conditions in logging for better reliability\r\n\r\n## 🛠️ Documentation & Examples\r\n- **Agent Examples**: Updated README and added comprehensive agent examples\r\n- **Template Updates**: Bumped LLM template version for latest improvements\r\n- **Deprecation Cleanup**: Removed deprecation warnings in quickstart for cleaner user experience\r\n- **Dynamic Fan-out\u002FFan-in**: Added support for dynamic fan-out\u002Ffan-in patterns with run templates for more flexible pipeline architectures\r\n\r\n## What's Changed\r\n* Add version 0.83.1 to legacy docs by @github-actions[bot] in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3828\r\n* Add dynamic fan-out\u002Ffan-in with run templates by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3826\r\n* Add version 0.84.0 to DB migration script by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3833\r\n* Bump LLM template by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3837\r\n* Update README and add agent example by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3815\r\n* Remove deprecation warning in quickstart by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3834\r\n* Fix appending to DB migrations in release action by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3836\r\n* Add intermediate cleanup in CLI profiling CI by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3835\r\n* Purge unused\u002Finactive local docker services by @AlexejPenner in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3768\r\n* Fix the retry logic and credential refresh for the Vertex step operator by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3853\r\n* Schedule management for the Kubernetes orchestrator by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3847\r\n* Cleanup Kubernetes secret if orchestrator pod fails to start by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3846\r\n* Enable schedule RBAC by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3848\r\n* Add explicit k8s failure reasons to logs by @avishniakov in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3854\r\n* Collect step run exception info by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3838\r\n* Block bot traffic from Segment analytics to prevent MTU quota exhaustion by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3858\r\n* Fail on certain container waiting reasons during job monitoring by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3851\r\n* Add support for external service accounts by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3793\r\n* Solving the race conditions for logging by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3855\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fcompare\u002F0.84.0...0.84.1","2025-07-30T10:09:30",{"id":239,"version":240,"summary_zh":241,"released_at":242},71558,"0.84.0","The `0.84.0` release delivers significant **architectural improvements** and **orchestration enhancements** that boost ZenML's reliability, performance, and developer experience. This release focuses on pipeline execution resilience with step retries and early stopping capabilities, enhanced Kubernetes orchestrator features, and improved server-side processing for better scalability.\r\n\r\n## 🚨 Breaking Changes\r\n\r\n### Kubernetes Orchestrator Compatibility\r\n- **Client-Orchestrator Version Compatibility**: This release introduces changes to the Kubernetes orchestrator that make the new client version incompatible with previous versions. When using the Kubernetes orchestrator, **the ZenML version in both the client and orchestrator pod must match exactly**. This means that if you upgrade your ZenML client to 0.84.0, you must also ensure that the orchestrator pod images are also using ZenML 0.84.0 to avoid compatibility issues.\r\n\r\n## 🚀 Orchestration Enhancements\r\n\r\n### Pipeline Control & Reliability\r\n- **Early Pipeline Stopping**: Added ability to stop pipelines early with Kubernetes orchestrator implementation for better resource management\r\n- **Step Retries**: Implemented configurable step retry mechanisms for improved pipeline resilience\r\n- **Step Status Refresh**: Enhanced step status monitoring with real-time refresh capabilities\r\n\r\n### Kubernetes Orchestrator Improvements\r\n- **Run Steps using Jobs**: Leverage Kubernetes jobs for better retry behavior when running steps\r\n- **Enhanced Pod Management**: Always build pipeline images for Kubernetes orchestrator and improved pod caching\r\n- **Pod Logs Access**: Added orchestrator pod logs functionality for better debugging and monitoring\r\n- **Kubernetes Model Validation**: Added warnings for invalid Kubernetes model data\r\n\r\n## ⚡ Performance Improvements\r\n- **Threadsafe RestZenStore**: Implemented thread-safe operations for improved concurrent access\r\n- **Server-side Processing**: Moved parent step computation and cascading tags processing to server-side for better performance\r\n- **Improved Fetching**: Enhanced pipeline and step run fetching mechanisms for faster response times\r\n\r\n## 🔧 Developer Experience\r\n\r\n### CLI & Login Improvements\r\n- **New ZenML Login**: Redesigned login experience with improved CLI styling and user interface\r\n\r\n### Configuration & Compatibility\r\n- **Easier Configuration**: Simplified step operator and experiment tracker configuration for steps\r\n- **Unified Imports**: Standardized config imports to match documentation examples\r\n\r\n## 🔒 Infrastructure & Security\r\n- **Helm Chart Improvements**: Added PVC creation option for local configuration path\r\n- **Service Connector Security**: Hidden service connector secrets as internal implementation details\r\n\r\n## 🛠️ Fixes\r\n- **Model Version Fetching**: Fixed fetching model versions by UUID passed as string\r\n- **Visualization Handling**: Ensured visualizations are committed before potential transaction rollbacks\r\n- **SageMaker Configuration**: Made `SagemakerOrchestratorSettings.processor_tags` optional\r\n- **Data Artifact Access**: Fixed data artifact fetching issues\r\n- **Pydantic Validation**: Added support for step\u002Fparameter names with leading underscores in run templates\r\n- **Path Sanitization**: Added missing remote path and Docker tag sanitizations\r\n\r\n## 📚 Documentation\r\n- **Best Practices**: Updated documentation with minimum permissions for cloud stack components\r\n- **Concurrent Execution**: Added guidance for handling concurrent pipeline execution in separate containers\r\n- **User Guides**: Enhanced user guides with `llm-complete-guide` project and improved examples\r\n- **ZenML Pro**: Updated URLs in documentation to use zenml.io\u002Fpro\r\n- **VSCode Integration**: Added VSCode tutorial pipeline testing to CI\r\n\r\n## What's Changed\r\n* Update issue template configurations and URLs by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3756\r\n* Add new release to migration tests by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3764\r\n* Split run submission and monitoring by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3723\r\n* Made --no-verify-ssl work for pro tenants as well by @AlexejPenner in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3767\r\n* Threadsafe RestZenStore by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3758\r\n* Server-side parent step computation by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3762\r\n* Unified config imports to match docs by @AlexejPenner in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3743\r\n* Fixed misleading error message by @AlexejPenner in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3759\r\n* Adjusted settings to include submodules by @AlexejPenner in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3757\r\n* Switch from pkg_resources to importlib by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3722\r\n* Helm chart: Add option to create PVC for local configuration path by @jsuchome in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3761\r\n* Hide service connector secrets as internal im","2025-07-11T13:43:02",{"id":244,"version":245,"summary_zh":246,"released_at":247},71559,"0.83.1","\r\nThe `0.83.1` release delivers significant **performance improvements** and optimizations that enhance ZenML's efficiency and responsiveness. This release focuses on architectural improvements including step configuration separation from deployment, idempotent POST requests with request caching, and enhanced artifact loading performance, alongside important bug fixes and documentation updates.\r\n\r\n## ⚡ Performance Improvements\r\n\r\n- **Architectural Enhancement**: Separated step configurations from deployment for improved modularity and reduced overhead\r\n- **Request Optimization**: Implemented idempotent POST requests with intelligent request caching for faster API responses\r\n- **Artifact Loading**: Enhanced artifact loading performance by using the active artifact store as cache\r\n\r\n## Fixes\r\n\r\n- Fixed potential race condition during pipeline run status updates\r\n- Fixed missing artifact nodes in DAG visualization\r\n- Fixed run template fetching issues that could cause errors during pipeline execution\r\n- Fixed missing pipeline substitutions for legacy steps\r\n- Reset active project after deleting local database\r\n\r\n## Documentation\r\n\r\n- Updated Kubernetes configuration documentation with clearer guidance\r\n- Enhanced run template documentation with better examples and explanations\r\n- Added shared CSS guidance for HTML visualizations\r\n- Updated kaniko project status documentation\r\n- Fixed authentication confusion in ZenML Pro API documentation\r\n\r\n\r\n## What's Changed\r\n* More explicit handling of carriage return logs in steps by @AlexejPenner in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3714\r\n* Add `0.82.1` to the legacy docs by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3715\r\n* Reset active project after deleting local database by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3689\r\n* Keep visualizations when using existing tag by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3710\r\n* Use the active artifact store as cache when loading artifacts by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3721\r\n* Fix potential race condition during run status update by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3720\r\n* Random cleanup and bugfixes by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3709\r\n* Added documentation for a PowerShell quirk when referencing secrets in experiment tracker by @hala201 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3717\r\n* Fix some links by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3729\r\n* Fix run template fetching by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3726\r\n* Fix missing artifact nodes in DAG by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3727\r\n* Revert \"Keep visualizations when using existing tag\" by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3732\r\n* Tag improvements by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3724\r\n* Improve docs imports by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3731\r\n* Kubernetes configuration documentation update by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3736\r\n* Fix authentication confusion in ZenML Pro API documentation by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3734\r\n* Include missing pipeline substitutions for legacy steps by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3740\r\n* Better run template docs by @AlexejPenner in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3730\r\n* Add shared CSS guidance for HTML visualizations by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3742\r\n* Update kaniko project status in docs by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3744\r\n* Fix line breaks in llms.txt docs by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3746\r\n* Separate step configurations from deployment by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3739\r\n* Dynamically import entrypoint in root __init__ by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3728\r\n* Idempotent POST requests, request caching and other performance improvements by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3738\r\n\r\n## New Contributors\r\n* @hala201 made their first contribution in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3717\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fcompare\u002F0.83.0...0.83.1","2025-06-23T13:40:34",{"id":249,"version":250,"summary_zh":251,"released_at":252},71560,"0.83.0","🚀 **Major Performance Release** - ZenML 0.83.0 introduces significant performance improvements and response optimizations that dramatically reduce API response times and database query overhead. This release includes breaking changes and is **not compatible with earlier client\u002Fserver versions**.\r\n\r\n## ⚡ Performance Improvements\r\n\r\nThis release delivers substantial performance enhancements across the ZenML server:\r\n\r\n- **Optimized API Responses**: Pipeline run responses no longer include unpaginated step lists, dramatically reducing response sizes for large pipelines\r\n- **Database Query Optimization**: Improved query patterns with strategic joined loads to minimize database roundtrips\r\n- **Reduced Response Payloads**: Many attributes moved from `body` to `resources` to avoid unnecessary data transfer when objects are embedded in other responses\r\n\r\n## 🔄 Breaking Changes\r\n\r\n⚠️ **Client\u002FServer Compatibility**: This version is **not compatible** with earlier ZenML client\u002Fserver versions. Please ensure both client and server are upgraded to 0.83.0.\r\n\r\n### API Response Changes\r\n- Pipeline run responses no longer include `metadata.steps` or `metadata.step_substitutions`\r\n- Many model attributes moved from `body` to `resources` for performance\r\n- Project metadata structure simplified across all responses\r\n- Model version responses no longer include comprehensive artifact and run ID lists\r\n\r\n### Method Deprecations\r\n- `Model.get_pipeline_run(...)` and `ModelVersionResponse.get_pipeline_run(...)` have been removed\r\n\r\nFor a comprehensive list of all response changes and migration details, see [PR #3675](https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3675).\r\n\r\n## 🚀 Orchestrator Enhancements\r\n\r\n### Kubernetes Orchestrator\r\n- **Enhanced Caching**: Added caching capabilities in Kubernetes orchestrator entrypoint to improve performance and reduce unnecessary pod creations\r\n\r\n### Skypilot Orchestrator  \r\n- **Updated Version and Settings**: Updated to new Skypilot version and added new setting options\r\n\r\n## What's Changed\r\n* Adding 0.82.0 to the legacy docs by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3671\r\n* New deployment scenarios by @AlexejPenner in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3666\r\n* Add instruction to `WANDB_DISABLED` to `True` for the quickstart by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3673\r\n* Fix service connector docs example by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3679\r\n* Fix YAML extension check by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3677\r\n* Fix typos, bugs, and improve test precision by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3678\r\n* Performance boost fix: don't fetch entire pipeline run to verify pipeline API token validity by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3684\r\n* Async wrapper for FastAPI endpoints to run serialization in event loop by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3685\r\n* Don't log parent image digest warning if build is skipped by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3676\r\n* Convert string to raw to avoid warnings in python 3.12+ by @jlopezpena in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3687\r\n* Use correct artifact store for nested materializers by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3670\r\n* Tiny Discord docs fix by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3691\r\n* Store validated config with converted types in DB by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3668\r\n* Added zenml codespace env detection by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3686\r\n* Fix setuptools vulnerabilities and deprecate pip as DockerSettings default by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3694\r\n* Improve alerter documentation with comprehensive ask step coverage by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3693\r\n* Add comprehensive agent guidelines with `AGENTS.md` and `CLAUDE.md` by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3680\r\n* Performance test utilities, stats and stress test pipeline updates by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3690\r\n* Fix GCP service connector expiry by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3697\r\n* Fix trivy image scanning GitHub actions by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3699\r\n* Add ability to strip the timestamps from the logs, on request by @avishniakov in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3683\r\n* Send POST request for RBAC permission checks to avoid URL length limits by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3702\r\n* Response and database improvements by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3675\r\n* Add Caching in Kubernetes orchestrator entrypoint by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3703\r\n* New Pro onboarding by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3704\r\n* Suppress repeated DockerSettings warnings by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3705\r\n* New OSS dashboa","2025-05-28T14:37:46",{"id":254,"version":255,"summary_zh":256,"released_at":257},71561,"0.82.1","The `0.82.1` release focuses on incremental improvements to [run template](https:\u002F\u002Fdocs.zenml.io\u002Fconcepts\u002Ftemplates) management, [Kubernetes orchestration](https:\u002F\u002Fdocs.zenml.io\u002Fstacks\u002Fstack-components\u002Forchestrators\u002Fkubernetes), [Docker build performance](https:\u002F\u002Fdocs.zenml.io\u002Fconcepts\u002Fcontainerization), and overall robustness, while shipping a wide range of documentation updates and quality-of-life enhancements. Key highlights include configurable Kubernetes job clean-up behavior, cascading tags for cached step runs, [`pyproject.toml` support for `DockerSettings`](https:\u002F\u002Fdocs.zenml.io\u002Fconcepts\u002Fcontainerization#python-dependencies), improved login stability, and faster Docker build caching through parent image digests.\r\n\r\n## Features\r\n\r\n- Added [`pyproject.toml` support for configuring `DockerSettings`](https:\u002F\u002Fdocs.zenml.io\u002Fconcepts\u002Fcontainerization#python-dependencies), making container builds easier to manage.\r\n- Added a unique instance label to the Helm chart to simplify the operation of multiple ZenML deployments in the same cluster.\r\n- Introduced a new stress-test example that showcases ZenML scalability and can be used to benchmark installations.\r\n\r\n## Improvements\r\n\r\n- Added cascading tags for cached step runs to improve cache reuse and pipeline run performance.\r\n- Added configurable Kubernetes job clean-up options for the Kubernetes orchestrator.\r\n- Added a limit to the maximum number of concurrent template runs and improved the overall run template UX.\r\n- Prevented unnecessary hydration in project-scoped API responses, reducing payload size and improving performance.\r\n- Optimized Docker build caching by using parent image digests and extending development Dockerfiles.\r\n- Pinned `setuptools` to a stable version and removed redundant script dependencies to avoid build failures.\r\n\r\n## Fixes\r\n\r\n- Fixed DockerHub repository digest detection when building images.\r\n- Fixed miscellaneous login issues and introduced an API login lock for added robustness.\r\n- Fixed dashboard bolt icon rendering.\r\n- Updated Alembic to address compatibility issues.\r\n\r\n## Documentation\r\n\r\n- Added a [\"5-minute quick wins\" guide](https:\u002F\u002Fdocs.zenml.io\u002Fuser-guides\u002Fbest-practices\u002Fquick-wins) and [a new dedicated docs section regarding orchestrator selection](https:\u002F\u002Fdocs.zenml.io\u002Fuser-guides\u002Fbest-practices\u002Fchoose-orchestration-environment).\r\n- Added [documentation for dashboard features](https:\u002F\u002Fdocs.zenml.io\u002Fconcepts\u002Fdashboard-features) and an accurate list of workload manager options.\r\n- Added `0.81.0` to legacy docs and fixed [artifact visualization guidance](https:\u002F\u002Fzenml-io.gitbook.io\u002Fbarisky\u002Fconcepts\u002Fartifacts\u002Fvisualizations).\r\n- Numerous minor documentation fixes and cleanup.\r\n\r\n\r\n## What's Changed\r\n* Adding 0.81.0 to the legacy docs by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3630\r\n* Extending the Dockerfiles  by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3632\r\n* Use parent image digest for cache invalidation by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3617\r\n* Pro API login lock and other robustness improvements by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3625\r\n* Fixing images for the Hello World and various other fixes by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3637\r\n* Pin setuptools and remove it from scripts by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3636\r\n* Update link validation to skip GitHub links and improve progress tracking by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3641\r\n* Added docs section by @AlexejPenner in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3640\r\n* Various improvements to the release flow by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3638\r\n* Fix Dockerhub repo digest detection by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3621\r\n* Add unique instance label to helm chart by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3639\r\n* Slight doc fix. Fixes #3645 by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3646\r\n* Add 5-min quick wins page to docs by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3633\r\n* Limit max concurrent template runs by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3627\r\n* Fix bolt icon by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3648\r\n* Update docs with accurate list of workload manager options by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3643\r\n* Workflow to deploy workspaces for PRs by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3618\r\n* Format link checker by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3656\r\n* Improve run template UX by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3602\r\n* Prevent unnecessary hydration in project-scoped responses by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3657\r\n* Update alembic version to \"^1.8.1\" in pyproject.toml by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3529\r\n* Fixing artifact visualization docs by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3661\r\n* Add stre","2025-05-14T16:34:51",{"id":259,"version":260,"summary_zh":261,"released_at":262},71562,"0.82.0","The 0.82.0 release delivers significant improvements to [Kubernetes orchestrator](https:\u002F\u002Fdocs.zenml.io\u002Fstacks\u002Fstack-components\u002Forchestrators\u002Fkubernetes), enhanced documentation, and numerous fixes to improve overall stability and performance. Key highlights include configurable max parallelism for Kubernetes orchestrator, customizable pod name prefixes and scheduler options, improved runner timeouts, and support for private service connections in Vertex AI. This release also includes comprehensive documentation updates, and library compatibility improvements for NumPy and Pandas.\r\n\r\n# Features\r\n\r\n- Added max parallelism option for [Kubernetes orchestrator](https:\u002F\u002Fdocs.zenml.io\u002Fstacks\u002Fstack-components\u002Forchestrators\u002Fkubernetes)\r\n- Added support for pod name prefixes and scheduler configuration\r\n- Added private service connect option for [Vertex AI orchestrator](https:\u002F\u002Fdocs.zenml.io\u002Fstacks\u002Fstack-components\u002Forchestrators\u002Fvertex)\r\n- Made runner timeout easily configurable\r\n- Added storage for list of Python packages\r\n- Added an ability to deep refresh the status of your run from the dashboard\r\n\r\n# Improvements\r\n\r\n- Adjusted GitHub code repo regex pattern for better compatibility\r\n- Improved build invalidation when parent Dockerfile changes\r\n- Enhanced directory handling during code download\r\n- Added ability to list model versions without models\r\n- Added support for extra attributes in all ZenML models\r\n- Disabled default project behavior for pro workspaces\r\n\r\n# Fixes\r\n\r\n- Fixed run templates listing\r\n- Eliminated premature active project warning logs\r\n- Updated scikit-learn requirement in SklearnIntegration\r\n- Updated NumPy integration to work with both 1.x and 2.x library versions\r\n- Added Pandas custom data type error handling and logging\r\n- Removed unnecessary and invalid settings\r\n- Various frontend bug fixes\r\n\r\n# Documentation\r\n\r\n- Completed comprehensive documentation revamp\r\n- Added documentation for [self-hosted run templates](https:\u002F\u002Fdocs.zenml.io\u002Fpro\u002Fdeployments\u002Fself-hosted#enabling-run-templates-support)\r\n- Removed outdated redirects and sections\r\n- Fixed broken absolute links and restored missing sections\r\n- Added documentation for run template TTL environment variable\r\n- Updated image paths for ZenML pipeline screenshots\r\n- Migrated starter guide to unified log_metadata method\r\n\r\n# Breaking Changes\r\n\r\n- Removed `scikit-image` as a requirement of the sklearn integration\r\n\r\n## What's Changed\r\n* Adding 0.80.2 to legacy docs by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3547\r\n* Add 0.81.0 to migration tests by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3546\r\n* Document self-hosted run templates by @stefannica in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3552\r\n* Adjusted github code repo regex pattern matching to work in more cases by @AlexejPenner in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3550\r\n* Documentation revamp by @bcdurak in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3524\r\n* Remove outdated redirects and sections by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3564\r\n* Removed orphan assets by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3567\r\n* Fixed broken absolute links and brought back some sections by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3585\r\n* Fix listing run templates by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3587\r\n* Don't log missing active project warnings preemtively  by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3586\r\n* Docs fixes by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3565\r\n* More docs changes  by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3592\r\n* Update scikit-learn requirement in SklearnIntegration by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3551\r\n* Small docs fixes by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3584\r\n* More small docs fixes by @strickvl in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3601\r\n* Docs: Migrate to unified log_metadata method in starter guide by @marwan37 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3603\r\n* Add docs for run template TTL env var by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3604\r\n* Make runner timeout easily configurable by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3589\r\n* Disable default project behavior for pro workspaces by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3605\r\n* Remove unnecessary and invalid settings by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3548\r\n* Update image paths for ZenML pipeline screenshots by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3611\r\n* Update NumPy integration to work with 1.x and 2.x versions of the library by @htahir1 in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3532\r\n* Invalidate the build when parent dockerfile changes by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3607\r\n* Ignore existing directories during code download by @schustmi in https:\u002F\u002Fgithub.com\u002Fzenml-io\u002Fzenml\u002Fpull\u002F3609\r\n* Max parallelism option for kubernetes orchestrator by @schustmi in https:\u002F\u002Fgit","2025-05-01T01:02:56"]