[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-sematic-ai--sematic":3,"tool-sematic-ai--sematic":61},[4,18,26,36,44,53],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":17},4358,"openclaw","openclaw\u002Fopenclaw","OpenClaw 是一款专为个人打造的本地化 AI 助手，旨在让你在自己的设备上拥有完全可控的智能伙伴。它打破了传统 AI 助手局限于特定网页或应用的束缚，能够直接接入你日常使用的各类通讯渠道，包括微信、WhatsApp、Telegram、Discord、iMessage 等数十种平台。无论你在哪个聊天软件中发送消息，OpenClaw 都能即时响应，甚至支持在 macOS、iOS 和 Android 设备上进行语音交互，并提供实时的画布渲染功能供你操控。\n\n这款工具主要解决了用户对数据隐私、响应速度以及“始终在线”体验的需求。通过将 AI 部署在本地，用户无需依赖云端服务即可享受快速、私密的智能辅助，真正实现了“你的数据，你做主”。其独特的技术亮点在于强大的网关架构，将控制平面与核心助手分离，确保跨平台通信的流畅性与扩展性。\n\nOpenClaw 非常适合希望构建个性化工作流的技术爱好者、开发者，以及注重隐私保护且不愿被单一生态绑定的普通用户。只要具备基础的终端操作能力（支持 macOS、Linux 及 Windows WSL2），即可通过简单的命令行引导完成部署。如果你渴望拥有一个懂你",349277,3,"2026-04-06T06:32:30",[13,14,15,16],"Agent","开发框架","图像","数据工具","ready",{"id":19,"name":20,"github_repo":21,"description_zh":22,"stars":23,"difficulty_score":10,"last_commit_at":24,"category_tags":25,"status":17},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,"2026-04-05T11:01:52",[14,15,13],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":32,"last_commit_at":33,"category_tags":34,"status":17},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",150037,2,"2026-04-10T23:33:47",[14,13,35],"语言模型",{"id":37,"name":38,"github_repo":39,"description_zh":40,"stars":41,"difficulty_score":32,"last_commit_at":42,"category_tags":43,"status":17},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。",108322,"2026-04-10T11:39:34",[14,15,13],{"id":45,"name":46,"github_repo":47,"description_zh":48,"stars":49,"difficulty_score":32,"last_commit_at":50,"category_tags":51,"status":17},6121,"gemini-cli","google-gemini\u002Fgemini-cli","gemini-cli 是一款由谷歌推出的开源 AI 命令行工具，它将强大的 Gemini 大模型能力直接集成到用户的终端环境中。对于习惯在命令行工作的开发者而言，它提供了一条从输入提示词到获取模型响应的最短路径，无需切换窗口即可享受智能辅助。\n\n这款工具主要解决了开发过程中频繁上下文切换的痛点，让用户能在熟悉的终端界面内直接完成代码理解、生成、调试以及自动化运维任务。无论是查询大型代码库、根据草图生成应用，还是执行复杂的 Git 操作，gemini-cli 都能通过自然语言指令高效处理。\n\n它特别适合广大软件工程师、DevOps 人员及技术研究人员使用。其核心亮点包括支持高达 100 万 token 的超长上下文窗口，具备出色的逻辑推理能力；内置 Google 搜索、文件操作及 Shell 命令执行等实用工具；更独特的是，它支持 MCP（模型上下文协议），允许用户灵活扩展自定义集成，连接如图像生成等外部能力。此外，个人谷歌账号即可享受免费的额度支持，且项目基于 Apache 2.0 协议完全开源，是提升终端工作效率的理想助手。",100752,"2026-04-10T01:20:03",[52,13,15,14],"插件",{"id":54,"name":55,"github_repo":56,"description_zh":57,"stars":58,"difficulty_score":32,"last_commit_at":59,"category_tags":60,"status":17},4721,"markitdown","microsoft\u002Fmarkitdown","MarkItDown 是一款由微软 AutoGen 团队打造的轻量级 Python 工具，专为将各类文件高效转换为 Markdown 格式而设计。它支持 PDF、Word、Excel、PPT、图片（含 OCR）、音频（含语音转录）、HTML 乃至 YouTube 链接等多种格式的解析，能够精准提取文档中的标题、列表、表格和链接等关键结构信息。\n\n在人工智能应用日益普及的今天，大语言模型（LLM）虽擅长处理文本，却难以直接读取复杂的二进制办公文档。MarkItDown 恰好解决了这一痛点，它将非结构化或半结构化的文件转化为模型“原生理解”且 Token 效率极高的 Markdown 格式，成为连接本地文件与 AI 分析 pipeline 的理想桥梁。此外，它还提供了 MCP（模型上下文协议）服务器，可无缝集成到 Claude Desktop 等 LLM 应用中。\n\n这款工具特别适合开发者、数据科学家及 AI 研究人员使用，尤其是那些需要构建文档检索增强生成（RAG）系统、进行批量文本分析或希望让 AI 助手直接“阅读”本地文件的用户。虽然生成的内容也具备一定可读性，但其核心优势在于为机器",93400,"2026-04-06T19:52:38",[52,14],{"id":62,"github_repo":63,"name":64,"description_en":65,"description_zh":66,"ai_summary_zh":67,"readme_en":68,"readme_zh":69,"quickstart_zh":70,"use_case_zh":71,"hero_image_url":72,"owner_login":73,"owner_name":74,"owner_avatar_url":75,"owner_bio":76,"owner_company":77,"owner_location":77,"owner_email":77,"owner_twitter":78,"owner_website":79,"owner_url":80,"languages":81,"stars":118,"forks":119,"last_commit_at":120,"license":121,"difficulty_score":122,"env_os":123,"env_gpu":124,"env_ram":125,"env_deps":126,"category_tags":138,"github_topics":140,"view_count":32,"oss_zip_url":77,"oss_zip_packed_at":77,"status":17,"created_at":152,"updated_at":153,"faqs":154,"releases":185},6409,"sematic-ai\u002Fsematic","sematic","An open-source ML pipeline development platform","Sematic 是一款开源的机器学习流水线开发平台，旨在帮助工程师和数据科学家轻松构建从数据处理到模型训练的端到端工作流。它解决了传统 ML 开发中本地环境与云端部署不一致、流程难以追踪复现以及资源调度复杂等痛点。\n\n无论是需要在笔记本上快速验证想法的研究人员，还是需要在 Kubernetes 集群上管理大规模生产任务的资深工程师，Sematic 都能满足需求。其核心理念是“本地与云端一致”，用户只需使用纯 Python 编写代码，即可无缝在本地机器或云环境中运行，无需关心底层基础设施的差异。\n\nSematic 的技术亮点在于其轻量级的 Python SDK 和强大的动态图能力。它支持将 Spark 数据处理、PyTorch\u002FTensorFlow 模型训练等任意 Python 逻辑串联起来，形成类型安全且可嵌套的流水线。所有执行过程、中间产物和数据血缘都会自动记录，并通过现代化的 Web 仪表盘进行可视化监控。此外，它还允许为每个步骤灵活配置计算资源（如 CPU、GPU 或 Spark 集群），确保在优化性能的同时控制成本。通过 Sematic，团队可以更专注于算法创新，而非繁琐的工程","Sematic 是一款开源的机器学习流水线开发平台，旨在帮助工程师和数据科学家轻松构建从数据处理到模型训练的端到端工作流。它解决了传统 ML 开发中本地环境与云端部署不一致、流程难以追踪复现以及资源调度复杂等痛点。\n\n无论是需要在笔记本上快速验证想法的研究人员，还是需要在 Kubernetes 集群上管理大规模生产任务的资深工程师，Sematic 都能满足需求。其核心理念是“本地与云端一致”，用户只需使用纯 Python 编写代码，即可无缝在本地机器或云环境中运行，无需关心底层基础设施的差异。\n\nSematic 的技术亮点在于其轻量级的 Python SDK 和强大的动态图能力。它支持将 Spark 数据处理、PyTorch\u002FTensorFlow 模型训练等任意 Python 逻辑串联起来，形成类型安全且可嵌套的流水线。所有执行过程、中间产物和数据血缘都会自动记录，并通过现代化的 Web 仪表盘进行可视化监控。此外，它还允许为每个步骤灵活配置计算资源（如 CPU、GPU 或 Spark 集群），确保在优化性能的同时控制成本。通过 Sematic，团队可以更专注于算法创新，而非繁琐的工程运维。","![Sematic Logo](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsematic-ai_sematic_readme_2bc0e3d9b38f.png)\n\n\u003Ch2 align=\"center\">The open-source Continuous Machine Learning Platform\u003C\u002Fh2>\n\n\u003Ch3 align=\"center\">Build ML pipelines with only Python, run on your laptop, or in the cloud.\u003C\u002Fh3>\n\n![PyPI](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fsematic\u002F0.41.0?style=for-the-badge)\n[![CircleCI](https:\u002F\u002Fimg.shields.io\u002Fcircleci\u002Fbuild\u002Fgithub\u002Fsematic-ai\u002Fsematic\u002Fmain?label=CircleCI&style=for-the-badge&token=60d1953bfee5b6bf8201f8e84a10eaa5bf5622fe)](https:\u002F\u002Fapp.circleci.com\u002Fpipelines\u002Fgithub\u002Fsematic-ai\u002Fsematic?branch=main&filter=all)\n![PyPI - License](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fl\u002Fsematic?style=for-the-badge)\n[![Python 3.9](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPython-3.9-blue?style=for-the-badge&logo=none)](https:\u002F\u002Fpython.org)\n[![Python 3.10](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPython-3.10-blue?style=for-the-badge&logo=none)](https:\u002F\u002Fpython.org)\n[![Python 3.11](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPython-3.11-blue?style=for-the-badge&logo=none)](https:\u002F\u002Fpython.org)\n[![Python 3.12](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPython-3.12-blue?style=for-the-badge&logo=none)](https:\u002F\u002Fpython.org)\n[![Python 3.13](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPython-3.13-blue?style=for-the-badge&logo=none)](https:\u002F\u002Fpython.org)\n![Discord](https:\u002F\u002Fimg.shields.io\u002Fdiscord\u002F983789877927747714?label=DISCORD&style=for-the-badge)\n[![Made By Sematic](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMade_by-Sematic_🦊-E19632?style=for-the-badge&logo=none)](https:\u002F\u002Fsematic.dev)\n![PyPI - Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fsematic?style=for-the-badge)\n\n![Sematic Screenshot](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsematic-ai_sematic_readme_ed533da4c05b.png)\n\n[Sematic](https:\u002F\u002Fsematic.dev) is an open-source ML development platform. It\nlets ML Engineers and Data Scientists write arbitrarily complex end-to-end\npipelines with simple Python and execute them on their local machine, in a cloud\nVM, or on a Kubernetes cluster to leverage cloud resources.\n\nSematic is based on learnings gathered at top self-driving car companies. It\nenables chaining data processing jobs (e.g. Apache Spark) with model training\n(e.g. PyTorch, Tensorflow), or any other arbitrary Python business logic into\ntype-safe, traceable, reproducible end-to-end pipelines that can be monitored\nand visualized in a modern web dashboard.\n\nRead our [documentation](https:\u002F\u002Fdocs.sematic.dev) and join our [Discord\nchannel](https:\u002F\u002Fdiscord.gg\u002F4KZJ6kYVax).\n\n## Why Sematic\n\n- **Easy onboarding** – no deployment or infrastructure needed to get started,\n  simply install Sematic locally and start exploring.\n- **Local-to-cloud parity** – run the same code on your local laptop and on your\n  Kubernetes cluster.\n- **End-to-end traceability** – all pipeline artifacts are persisted, tracked,\n  and visualizable in a web dashboard.\n- **Access heterogeneous compute** – customize required resources for each\n  pipeline step to optimize your performance and cloud footprint (CPUs, memory,\n  GPUs, Spark cluster, etc.)\n- **Reproducibility** – rerun your pipelines from the UI with guaranteed\n  reproducibility of results\n\n## Getting Started\n\nTo get started locally, simply install Sematic in your Python environment:\n\n```shell\n$ pip install sematic\n```\n\nStart the local web dashboard:\n\n```shell\n$ sematic start\n```\n\nRun an example pipeline:\n\n```shell\n$ sematic run examples\u002Fmnist\u002Fpytorch\n```\n\nCreate a new boilerplate project:\n\n```shell\n$ sematic new my_new_project\n```\n\nOr from an existing example:\n\n```shell\n$ sematic new my_new_project --from examples\u002Fmnist\u002Fpytorch\n```\n\nThen run it with:\n\n```shell\n$ python3 -m my_new_project\n```\n\nTo deploy Sematic to Kubernetes and leverage cloud resources, see our\n[documentation](https:\u002F\u002Fdocs.sematic.dev).\n\n## Features\n\n- **Lightweight Python SDK** – define arbitrarily complex end-to-end pipelines\n- **Pipeline nesting** – arbitrarily nest pipelines into larger pipelines\n- **Dynamic graphs** – Python-defined graphs allow for iterations, conditional\n  branching, etc.\n- **Lineage tracking** – all inputs and outputs of all steps are persisted and\n  tracked\n- **Runtime type-checking** – fail early with run-time type checking\n- **Web dashboard** – Monitor, track, and visualize pipelines in a modern web UI\n- **Artifact visualization** – visualize all inputs and outputs of all steps in\n  the web dashboard\n- **Local execution** – run pipelines on your local machine without any\n  deployment necessary\n- **Cloud orchestration** – run pipelines on Kubernetes to access GPUs and other\n  cloud resources\n- **Heterogeneous compute resources** – run different steps on different\n  machines (e.g. CPUs, memory, GPU, Spark, etc.)\n- **Helm chart deployment** – install Sematic on your Kubernetes cluster\n- **Pipeline reruns** – rerun pipelines from the UI from an arbitrary point in\n  the graph\n- **Step caching** – cache expensive pipeline steps for faster iteration\n- **Step retry** – recover from transient failures with step retries\n- **Metadata and collaboration** – Tags, source code visualization, docstrings,\n  notes, etc.\n- **Numerous integrations** – See below\n\n## Integrations\n\n- **Apache Spark** – on-demand in-cluster Spark cluster\n- **Ray** – on-demand Ray in-cluster Ray resources\n- **Snowflake** – easily query your data warehouse (other warehouses supported\n  too)\n- **Plotly, Matplotlib** – visualize plot artifacts in the web dashboard\n- **Pandas** – visualize dataframe artifacts in the dashboard\n- **Grafana** – embed Grafana panels in the web dashboard\n- **Bazel** – integrate with your Bazel build system\n- **Helm chart** – deploy to Kubernetes with our Helm chart\n- **Git** – track git information in the web dashboard\n\n## Community and resources\n\nLearn more about Sematic and get in touch with the following resources:\n\n- [Sematic landing page](https:\u002F\u002Fsematic.dev)\n- [Documentation](https:\u002F\u002Fdocs.sematic.dev)\n- [Discord channel](https:\u002F\u002Fdiscord.gg\u002F4KZJ6kYVax)\n- [YouTube channel](https:\u002F\u002Fwww.youtube.com\u002F@sematic-ai)\n- [Our Blog](https:\u002F\u002Fsematic.dev\u002Fblog)\n\n## Contribute!\n\nTo contribute to Sematic, check out [open issues tagged \"good first\nissue\"](https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fissues?q=is%3Aopen+is%3Aissue+label%3A%22good+first+issue%22),\nand get in touch with us on [Discord](https:\u002F\u002Fdiscord.gg\u002F4KZJ6kYVax).\nYou can find instructions on how to get your development environment set up\nin our [developer docs](.\u002Fdeveloper-docs\u002FREADME.md). If you'd like to add\nan example, you may also find\n[this guide](https:\u002F\u002Fdocs.sematic.dev\u002Fproject\u002Fcontributor-guide\u002Fcontribute-example)\nhelpful.\n\n![scarf pixel](https:\u002F\u002Fstatic.scarf.sh\u002Fa.png?x-pxid=80c3593f-25a0-4b06-90a1-0b670a6567d4)\n","![Sematic Logo](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsematic-ai_sematic_readme_2bc0e3d9b38f.png)\n\n\u003Ch2 align=\"center\">开源的持续机器学习平台\u003C\u002Fh2>\n\n\u003Ch3 align=\"center\">仅用 Python 即可构建机器学习流水线，可在您的笔记本电脑或云端运行。\u003C\u002Fh3>\n\n![PyPI](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fsematic\u002F0.41.0?style=for-the-badge)\n[![CircleCI](https:\u002F\u002Fimg.shields.io\u002Fcircleci\u002Fbuild\u002Fgithub\u002Fsematic-ai\u002Fsematic\u002Fmain?label=CircleCI&style=for-the-badge&token=60d1953bfee5b6bf8201f8e84a10eaa5bf5622fe)](https:\u002F\u002Fapp.circleci.com\u002Fpipelines\u002Fgithub\u002Fsematic-ai\u002Fsematic?branch=main&filter=all)\n![PyPI - License](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fl\u002Fsematic?style=for-the-badge)\n[![Python 3.9](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPython-3.9-blue?style=for-the-badge&logo=none)](https:\u002F\u002Fpython.org)\n[![Python 3.10](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPython-3.10-blue?style=for-the-badge&logo=none)](https:\u002F\u002Fpython.org)\n[![Python 3.11](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPython-3.11-blue?style=for-the-badge&logo=none)](https:\u002F\u002Fpython.org)\n[![Python 3.12](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPython-3.12-blue?style=for-the-badge&logo=none)](https:\u002F\u002Fpython.org)\n[![Python 3.13](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPython-3.13-blue?style=for-the-badge&logo=none)](https:\u002F\u002Fpython.org)\n![Discord](https:\u002F\u002Fimg.shields.io\u002Fdiscord\u002F983789877927747714?label=DISCORD&style=for-the-badge)\n[![Made By Sematic](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMade_by-Sematic_🦊-E19632?style=for-the-badge&logo=none)](https:\u002F\u002Fsematic.dev)\n![PyPI - Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fsematic?style=for-the-badge)\n\n![Sematic 截图](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsematic-ai_sematic_readme_ed533da4c05b.png)\n\n[Sematic](https:\u002F\u002Fsematic.dev) 是一个开源的机器学习开发平台。它使机器学习工程师和数据科学家能够使用简单的 Python 编写任意复杂的端到端流水线，并在本地机器、云虚拟机或 Kubernetes 集群上执行，以充分利用云资源。\n\nSematic 基于顶级自动驾驶公司积累的经验。它支持将数据处理任务（如 Apache Spark）与模型训练（如 PyTorch、TensorFlow）或其他任意 Python 业务逻辑串联成类型安全、可追踪且可复现的端到端流水线，这些流水线可以在现代 Web 仪表板中进行监控和可视化。\n\n请阅读我们的 [文档](https:\u002F\u002Fdocs.sematic.dev)，并加入我们的 [Discord 社区](https:\u002F\u002Fdiscord.gg\u002F4KZJ6kYVax)。\n\n## 为什么选择 Sematic\n\n- **轻松上手**：无需部署或基础设施即可开始使用，只需在本地安装 Sematic 并开始探索。\n- **本地与云端一致**：在本地笔记本电脑和 Kubernetes 集群上运行相同的代码。\n- **端到端可追溯性**：所有流水线工件都会被持久化、跟踪，并可在 Web 仪表板中可视化。\n- **访问异构计算资源**：为每个流水线步骤自定义所需的资源，以优化性能和云资源占用（CPU、内存、GPU、Spark 集群等）。\n- **可复现性**：从 UI 重新运行流水线，确保结果的可复现性。\n\n## 快速入门\n\n要在本地开始使用，只需在您的 Python 环境中安装 Sematic：\n\n```shell\n$ pip install sematic\n```\n\n启动本地 Web 仪表板：\n\n```shell\n$ sematic start\n```\n\n运行示例流水线：\n\n```shell\n$ sematic run examples\u002Fmnist\u002Fpytorch\n```\n\n创建一个新的样板项目：\n\n```shell\n$ sematic new my_new_project\n```\n\n或者基于现有示例：\n\n```shell\n$ sematic new my_new_project --from examples\u002Fmnist\u002Fpytorch\n```\n\n然后通过以下命令运行：\n\n```shell\n$ python3 -m my_new_project\n```\n\n要将 Sematic 部署到 Kubernetes 并利用云资源，请参阅我们的 [文档](https:\u002F\u002Fdocs.sematic.dev)。\n\n## 功能特性\n\n- **轻量级 Python SDK**：定义任意复杂的端到端流水线。\n- **流水线嵌套**：可以将流水线任意嵌套到更大的流水线中。\n- **动态图**：Python 定义的图允许迭代、条件分支等操作。\n- **血缘追踪**：所有步骤的输入和输出都会被持久化并追踪。\n- **运行时类型检查**：通过运行时类型检查尽早发现问题。\n- **Web 仪表板**：在现代 Web 界面中监控、跟踪和可视化流水线。\n- **工件可视化**：在 Web 仪表板中可视化所有步骤的输入和输出。\n- **本地执行**：无需任何部署即可在本地机器上运行流水线。\n- **云编排**：在 Kubernetes 上运行流水线以访问 GPU 和其他云资源。\n- **异构计算资源**：在不同设备上运行不同的步骤（如 CPU、内存、GPU、Spark 等）。\n- **Helm Chart 部署**：将 Sematic 安装到您的 Kubernetes 集群。\n- **流水线重跑**：可以从图中的任意点通过 UI 重新运行流水线。\n- **步骤缓存**：缓存昂贵的流水线步骤以加快迭代速度。\n- **步骤重试**：通过步骤重试恢复因临时故障而中断的任务。\n- **元数据与协作**：标签、源代码可视化、文档字符串、备注等。\n- **丰富的集成**：详见下文。\n\n## 集成\n\n- **Apache Spark**：按需在集群内启动 Spark 集群。\n- **Ray**：按需在集群内分配 Ray 资源。\n- **Snowflake**：轻松查询您的数据仓库（也支持其他数据仓库）。\n- **Plotly、Matplotlib**：在 Web 仪表板中可视化图表工件。\n- **Pandas**：在仪表板中可视化 DataFrame 工件。\n- **Grafana**：将 Grafana 面板嵌入 Web 仪表板。\n- **Bazel**：与您的 Bazel 构建系统集成。\n- **Helm Chart**：使用我们的 Helm Chart 部署到 Kubernetes。\n- **Git**：在 Web 仪表板中跟踪 Git 信息。\n\n## 社区与资源\n\n欲了解更多关于 Sematic 的信息并取得联系，请访问以下资源：\n\n- [Sematic 首页](https:\u002F\u002Fsematic.dev)\n- [文档](https:\u002F\u002Fdocs.sematic.dev)\n- [Discord 社区](https:\u002F\u002Fdiscord.gg\u002F4KZJ6kYVax)\n- [YouTube 频道](https:\u002F\u002Fwww.youtube.com\u002F@sematic-ai)\n- [我们的博客](https:\u002F\u002Fsematic.dev\u002Fblog)\n\n## 参与贡献！\n\n如您想为 Sematic 做出贡献，请查看 [标记为“good first issue”的未解决问题](https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fissues?q=is%3Aopen+is%3Aissue+label%3A%22good+first+issue%22)，并在 [Discord](https:\u002F\u002Fdiscord.gg\u002F4KZJ6kYVax) 上与我们联系。您可以在我们的 [开发者文档](.\u002Fdeveloper-docs\u002FREADME.md) 中找到设置开发环境的说明。如果您想添加一个示例，也可以参考 [此指南](https:\u002F\u002Fdocs.sematic.dev\u002Fproject\u002Fcontributor-guide\u002Fcontribute-example)。\n\n![scarf pixel](https:\u002F\u002Fstatic.scarf.sh\u002Fa.png?x-pxid=80c3593f-25a0-4b06-90a1-0b670a6567d4)","# Sematic 快速上手指南\n\nSematic 是一个开源的持续机器学习平台，允许你仅使用 Python 构建复杂的端到端 ML 流水线，并支持在本地笔记本、云端虚拟机或 Kubernetes 集群上无缝运行。\n\n## 环境准备\n\n- **操作系统**：Linux、macOS 或 Windows（WSL2 推荐）\n- **Python 版本**：3.9 - 3.13\n- **前置依赖**：\n  - `pip` 包管理工具\n  - 可选：Docker（用于本地模拟云环境）或 Kubernetes 集群（用于生产部署）\n\n> **提示**：国内开发者建议使用国内镜像源加速安装，例如清华源或阿里源。\n\n## 安装步骤\n\n1. **安装 Sematic**\n\n   使用默认源：\n   ```shell\n   pip install sematic\n   ```\n\n   或使用国内镜像源（推荐）：\n   ```shell\n   pip install sematic -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n   ```\n\n2. **启动本地 Web 仪表盘**\n\n   安装完成后，启动本地开发服务器以查看流水线可视化界面：\n   ```shell\n   sematic start\n   ```\n   启动后，浏览器会自动打开或访问 `http:\u002F\u002Flocalhost:8080`。\n\n## 基本使用\n\n### 1. 运行示例流水线\n\n体验 Sematic 的最快方式是运行内置的 MNIST 手写数字识别示例（基于 PyTorch）：\n\n```shell\nsematic run examples\u002Fmnist\u002Fpytorch\n```\n\n运行期间，你可以在 Web 仪表盘中实时查看每一步的执行状态、输入输出数据及可视化图表。\n\n### 2. 创建新项目\n\n基于模板创建一个全新的项目骨架：\n\n```shell\nsematic new my_new_project\n```\n\n或者基于现有示例（如 MNIST）创建：\n\n```shell\nsematic new my_new_project --from examples\u002Fmnist\u002Fpytorch\n```\n\n### 3. 运行自定义项目\n\n进入项目目录并执行：\n\n```shell\ncd my_new_project\npython3 -m my_new_project\n```\n\n此时，你的代码将在本地执行，结果自动记录并在 Web 仪表盘中展示。无需任何额外的基础设施配置即可实现从本地开发到云端部署的代码一致性。","某电商公司的数据科学团队需要每周构建一个从原始日志清洗、特征工程到推荐模型训练及评估的端到端机器学习流程。\n\n### 没有 sematic 时\n- **环境割裂严重**：数据预处理脚本在本地能跑，但移到云端 GPU 集群训练时，常因依赖冲突或路径问题导致失败，反复调试耗费大量时间。\n- **过程黑盒难追溯**：中间产生的特征数据和模型版本散落在不同服务器目录中，一旦线上效果波动，难以快速定位是哪次代码变更或数据偏差导致的。\n- **资源调度僵化**：整个流程只能运行在单一配置的计算节点上，无法让轻量的数据处理步骤使用廉价 CPU，而让繁重的模型训练独占昂贵 GPU，造成成本浪费。\n- **复现成本极高**：重新运行实验需要手动整理参数和脚本顺序，稍有不慎就会导致结果不一致，团队协作时经常互相覆盖文件或产生歧义。\n\n### 使用 sematic 后\n- **本地云端无缝切换**：团队仅用纯 Python 定义一次流水线逻辑，即可先在笔记本上调试，随后一键部署到 Kubernetes 集群，彻底消除了“在我机器上是好的”这类问题。\n- **全链路可视化追踪**：sematic 自动记录并展示每一步的输入输出、代码版本及运行指标，通过 Web 仪表盘可直观下钻分析任意中间产物，问题定位从小时级缩短至分钟级。\n- **细粒度资源定制**：利用 sematic 的动态图能力，为数据清洗步骤分配低配 CPU，仅为模型训练步骤申请高性能 GPU，显著降低了云资源账单。\n- **一键可复现实验**：任何历史运行记录均可在界面中直接重放，系统保证环境与参数完全一致，让团队成员能放心地基于彼此的结果进行迭代优化。\n\nsematic 通过统一的 Python 接口打通了从本地开发到云端生产的全链路，让复杂的机器学习工程变得透明、高效且低成本。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsematic-ai_sematic_2bc0e3d9.png","sematic-ai","Sematic","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fsematic-ai_14d6a6c6.png","Prototype-to-production ML in days not weeks.",null,"SematicAI","https:\u002F\u002Fsematic.dev","https:\u002F\u002Fgithub.com\u002Fsematic-ai",[82,86,90,94,98,102,105,108,112,115],{"name":83,"color":84,"percentage":85},"Python","#3572A5",72.5,{"name":87,"color":88,"percentage":89},"TypeScript","#3178c6",26.5,{"name":91,"color":92,"percentage":93},"Starlark","#76d275",0.4,{"name":95,"color":96,"percentage":97},"JavaScript","#f1e05a",0.2,{"name":99,"color":100,"percentage":101},"Shell","#89e051",0.1,{"name":103,"color":104,"percentage":101},"HTML","#e34c26",{"name":106,"color":107,"percentage":101},"Makefile","#427819",{"name":109,"color":110,"percentage":111},"MDX","#fcb32c",0,{"name":113,"color":114,"percentage":111},"Smarty","#f0c040",{"name":116,"color":117,"percentage":111},"CSS","#663399",995,60,"2026-04-07T13:17:30","Apache-2.0",1,"Linux, macOS, Windows","非必需。支持按需配置 GPU 资源（特别是在 Kubernetes 云环境中），具体型号和显存取决于用户运行的任务（如 PyTorch\u002FTensorFlow 训练）。本地运行无需特定 GPU。","未说明。取决于具体流水线任务的负载，本地运行需满足操作系统及 Python 进程的基本需求。",{"notes":127,"python":128,"dependencies":129},"1. 该工具主打‘轻量级’，本地启动无需部署复杂基础设施，仅需 pip install sematic 即可运行。2. 支持‘本地到云端’的一致性，代码可在笔记本直接运行，也可无缝迁移至 Kubernetes 集群以调用异构计算资源（CPU\u002FGPU\u002FSpark）。3. 核心功能包括类型安全、血缘追踪、动态图执行及 Web 可视化仪表盘。4. 若需云部署，需具备 Kubernetes 集群及 Helm 环境。","3.9, 3.10, 3.11, 3.12, 3.13",[64,130,131,132,133,134,135,136,137],"PyTorch (可选，用于示例)","TensorFlow (可选)","Apache Spark (可选集成)","Ray (可选集成)","Pandas (可选集成)","Plotly\u002FMatplotlib (可选集成)","Snowflake (可选集成)","Kubernetes\u002FHelm (用于云部署)",[13,14,15,16,139],"其他",[141,142,143,144,145,146,147,148,149,150,151],"ai","data-science","machine-learning","ml","ml-ops","ml-pipeline","ml-pipelines","mlops","pipeline","python","python3","2026-03-27T02:49:30.150509","2026-04-11T08:12:53.582671",[155,160,165,170,175,180],{"id":156,"question_zh":157,"answer_zh":158,"source_url":159},29021,"如何配置 GitHub 集成以在 Sematic 流水线通过之前阻止合并？","可以通过配置 API Token 和插件设置来实现。状态更新将包含四个关键信息：1. 状态检查名称（例如 `sematic-pipelines-pass`）；2. 状态（`failure`, `success`, `error`, `pending`）；3. 指向详细信息的链接；4. 人类可读的状态消息。维护者指出需要在设置中配置 API Token，布尔标志是不够的，后续会通过 PR #839 修复并实现该功能。","https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fissues\u002F639",{"id":161,"question_zh":162,"answer_zh":163,"source_url":164},29022,"为什么多行函数文档字符串（docstrings）在 UI 中渲染不正确？","这是一个已知的 UI 渲染问题，特别是在处理包含参数说明的多行文档字符串时。维护者已在 PR #325 中修复了此问题。之前的讨论建议，如果检测到未使用 Markdown，可以直接原样显示文档字符串而不尝试特殊格式渲染，最终采用了一个简单的解决方案来纠正渲染逻辑。","https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fissues\u002F310",{"id":166,"question_zh":167,"answer_zh":168,"source_url":169},29023,"如何缓存昂贵的函数输出以避免重复计算？","Sematic 支持通过标记特定函数的输出来实现缓存。你可以在 Resolver 级别指定一个 `cache_namespace`（缓存命名空间），所有嵌套的 Future 将继承该命名空间。缓存值会在以下情况失效：更改缓存命名空间、更改输入值，或（如果明确声明）更改源代码内容。注意：`cache` 和 `cache_namespace` 参数应同时暴露给 `@sematic.func` 装饰器和 `Future.set` 方法。如果设置了 `cache=True` 但未生效，系统会记录警告日志。","https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fissues\u002F338",{"id":171,"question_zh":172,"answer_zh":173,"source_url":174},29024,"如何在 UI 中直接链接到特定的嵌套运行（nested run）或面板？","Sematic 已实现深度链接（deep linking）功能，允许通过 URL 唯一标识每个面板而不仅仅是分辨率（resolution）。URL 格式类似于 `\u002Fpipelines\u002F:calculatorPath\u002F:rootId?deeplink=params`。当用户访问此类链接时，之前选择的嵌套运行或打开的标签页会被自动高亮显示。这解决了之前 URL 只能标识分辨率，导致面板状态无法通过链接共享的问题。","https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fissues\u002F278",{"id":176,"question_zh":177,"answer_zh":178,"source_url":179},29025,"如果 Resolver 死亡或启动失败，如何清理孤儿资源？","系统引入了一个“垃圾回收”cron job（通常通过 Helm 安装在 Kubernetes 上）来定期检查并清理孤儿资源、无法启动的 Resolver 以及其他需要清理的对象。此外，如果作业被取消但 Resolver 错过了取消事件（因为事件交付最多保证一次），或者 Resolver Pod 发生错误（如 OOM），该机制也会停止相应的 Resolver Pod 及其子运行和相关资源。此功能已通过 PR #758 修复。","https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fissues\u002F427",{"id":181,"question_zh":182,"answer_zh":183,"source_url":184},29026,"在流水线执行期间，DAG 视图的缩放级别为什么会重置？","这是一个 UI Bug，之前在流水线执行且图表每次更新时，用户的缩放级别会被强制重置，导致体验不佳。该问题已被标记为技术债务并进行了重构（Refactor DAG view），现在的实现会在图表更新时保持用户设定的缩放级别，确保持续的视觉一致性。","https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fissues\u002F307",[186,191,196,201,206,211,216,221,226,231,236,241,246,251,256,261,266,271,276,281],{"id":187,"version":188,"summary_zh":189,"released_at":190},197870,"v0.41.0","## 变更内容\n\n* [改进] 移除企业版许可，将所有代码（包括此前仅限于企业版的功能）恢复为采用 Apache 2.0 许可证。\n* [改进] 增加对 Python 3.13 的支持，停止对 3.8 的支持。\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.40.0...v0.41.0\n\n## Helm Chart 版本\n1.2.1\n\n## 兼容性\n此版本的 Sematic 服务器可以兼容从 `v0.30.0` 开始的 pip 包。\n\n## 升级说明\n\n请参阅[这些说明](https:\u002F\u002Fdocs.sematic.dev\u002Fproject\u002Fupgrades)，了解如何确保正确使用新版本！","2025-01-09T18:06:29",{"id":192,"version":193,"summary_zh":194,"released_at":195},197871,"v0.40.0","## 变更内容\n\n* [功能] 允许为 Ray 集成自定义标签和注解\n* [修复] 修复了元组类型强制转换检查中的一个问题\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.39.1...v0.40.0\n\n## Helm Chart 版本\n1.2.0\n\n## 兼容性\n该版本的 Sematic 服务器可以支持追溯到 `v0.30.0` 的 pip 包。\n\n## 升级说明\n\n请参阅[这些说明](https:\u002F\u002Fdocs.sematic.dev\u002Fproject\u002Fupgrades)，以确保您能够正确使用新版本！","2024-10-16T18:05:10",{"id":197,"version":198,"summary_zh":199,"released_at":200},197872,"v0.39.1","## 变更内容\n\n  * [bugfix] 修复在 SQLAlchemy 升级后从全新安装进行数据库迁移的问题\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.39.0...v0.39.1\n\n## Helm Chart 版本\n1.1.23\n\n## 兼容性\n该版本的 Sematic 服务器可以兼容低至 `v0.30.0` 的 pip 包。\n\n## 升级说明\n\n请参阅[这些说明](https:\u002F\u002Fdocs.sematic.dev\u002Fproject\u002Fupgrades)，以确保您能够正确使用新版本！请注意，如果您的项目除了 Sematic 之外还以某种方式依赖于 SQLAlchemy，您可能需要采取额外措施才能顺利升级到此版本。\n\n感谢您的首次贡献，@bcalvert-graft！","2024-09-25T15:49:42",{"id":202,"version":203,"summary_zh":204,"released_at":205},197873,"v0.39.0","## 变更内容\n\n  * [改进] 重大变更：将 SQLAlchemy 升级至 >=2.0.0 版本\n  * [改进] 增加对 Python 3.11 和 3.12 的支持\n\n**完整变更日志**：https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.38.1...v0.39.0\n\n## Helm Chart 版本\n1.1.22\n\n## 兼容性\n此版本的 Sematic 服务器可兼容至 `v0.30.0` 的 pip 包。\n\n## 升级说明\n\n请参阅[这些说明](https:\u002F\u002Fdocs.sematic.dev\u002Fproject\u002Fupgrades)，以确保您能够正确使用新版本！请注意，如果您的项目除了 Sematic 之外还以某种方式依赖 SQLAlchemy，您可能需要采取措施来升级到此版本。\n","2024-08-27T16:07:34",{"id":207,"version":208,"summary_zh":209,"released_at":210},197874,"v0.38.1","## 变更内容\n\n  * [功能] 允许管道作者指定已批准的 Kubernetes 注解和标签\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.38.0...v0.38.1\n\n## Helm Chart 版本\n1.1.21\n\n## 兼容性\n此版本的 Sematic 服务器可以支持追溯到 `v0.30.0` 的 pip 包。\n\n## 升级说明\n请参阅[这些说明](https:\u002F\u002Fdocs.sematic.dev\u002Fproject\u002Fupgrades)，了解如何确保正确使用新版本！","2024-06-17T15:37:06",{"id":212,"version":213,"summary_zh":214,"released_at":215},197875,"v0.38.0","## 变更内容\n\n  * [改进] 在删除 Kubernetes Job 之前，添加更多 Pod 信息\n  * [修复] 当 `CloudRunner` 的 Pod 被驱逐时，阻止其将自身标记为已取消\n  * [修复] 提升 `CloudRunner` 利用并行性的能力\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.37.0...v0.38.0\n\n## Helm Chart 版本\n1.1.20\n\n## 兼容性\n此版本的 Sematic 服务器可兼容至 `v0.30.0` 的 pip 包。\n\n## 升级说明\n请参阅[这些说明](https:\u002F\u002Fdocs.sematic.dev\u002Fproject\u002Fupgrades)，以确保您能够正确使用新版本！","2024-04-15T21:22:59",{"id":217,"version":218,"summary_zh":219,"released_at":220},197876,"v0.37.0","## 变更内容\n\n  * [改进] 提升日志读取的响应速度\n  * [修复] 修复在集合中渲染数据类子类时出现的问题\n  * [修复] 修复完成日期渲染相关的问题\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.36.0...v0.37.0\n\n## Helm Chart 版本\n1.1.19\n\n## 兼容性\n该版本的 Sematic 服务器支持回退至 `v0.30.0` 的 pip 包。\n\n## 升级说明\n请参阅[这些说明](https:\u002F\u002Fdocs.sematic.dev\u002Fproject\u002Fupgrades)，以确保您能够正确使用新版本！","2024-03-08T21:36:54",{"id":222,"version":223,"summary_zh":224,"released_at":225},197877,"v0.36.0","## 变更内容\n\n  * [功能] 允许为 `CloudRunner` 指定自定义的 Kubernetes 资源。\n  * [功能] 添加了通过运行 ID 阻塞等待运行完成并获取其输出的 API。\n  * [改进] 在管道和运行搜索结果中，按住 Command\u002FCTRL 键单击某一行时，将在新标签页中打开该行对应的页面。\n  * [改进] 当 Runner Pod 消失时，清理程序会同步清理管道运行的元数据。\n  * [改进] 在使用 Future 进行比较时，提供更具信息量的错误消息。\n  * [改进] 确保在 Helm ConfigMap 发生变更后，服务器 Pod 会自动重启。\n  * [修复] 当写入外部存储失败时，以更明确的错误信息提示失败原因。\n  * [修复] 为 Google OAuth 添加了电子邮件域名回退机制。\n  * [修复] 修复了多个运行在同一时刻发生超时的 bug。\n  * [修复] 使 Kuberay 版本解析对 Ray 集成更加稳健。\n\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.35.0...v0.36.0\n\n## Helm Chart 版本\n1.1.18\n\n## 兼容性\n此版本的 Sematic 服务器可以支持追溯到 `v0.30.0` 的 pip 包版本。\n\n## 升级说明\n请参阅[这些说明](https:\u002F\u002Fdocs.sematic.dev\u002Fproject\u002Fupgrades)，了解如何确保正确使用新版本！\n\n感谢您的贡献，@kaushil24！","2023-11-27T18:58:55",{"id":227,"version":228,"summary_zh":229,"released_at":230},197878,"v0.35.0","## 变更内容\n  * [功能] 添加对在 Google Cloud Platform 上的 GKE 集群进行部署的支持\n  * [功能] 将新版 Dashboard 设置为默认版本。此前该版本一直处于可选的 Beta 测试阶段。您仍可通过用户图标切换回旧版\n  * [功能] 允许从那些无法基于服务账号自动访问的私有镜像仓库拉取镜像\n  * [改进] 对文档、Dashboard、日志记录以及内部 API 验证进行了小幅优化\n  * [改进] 优化了 Dashboard 首页布局，提升易用性\n  * [改进] 提高了开发文档的可见性\n  * [修复] 破坏性变更：通过始终要求使用专用的 Socket.io 微服务实例，避免了极端情况下的部署错误\n  * [修复] 修复了一个 CLI 流水线取消信号被发送到错误服务器地址的 bug\n  * [修复] 修复了传递性依赖错误\n  * [修复] 修复了 Union 类型的序列化错误\n  * [修复] 当服务器不可达时，避免为 CLI 版本命令记录无益的堆栈跟踪信息\n  * [修复] 修复了 Dashboard 在渲染缓存运行的信息提示时显示不正确的问题\n  * [修复] 将概念重命名同步更新到 Dashboard 中面向用户的提示信息\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.34.1...v0.35.0\n\n## Helm Chart 版本\n1.1.17\n\n## 兼容性\n此版本的 Sematic 服务器支持追溯至 `v0.30.0` 的 pip 包。\n\n## 升级说明\n请参阅[这些说明](https:\u002F\u002Fdocs.sematic.dev\u002Fproject\u002Fupgrades#vx.x.x-to-v0.35.0)，以确保您能够正确使用新版本！","2023-10-11T22:58:47",{"id":232,"version":233,"summary_zh":234,"released_at":235},197879,"v0.34.1","## 变更内容\n* [功能] 使 Bazel 镜像生成宏可配置\n* [改进] 对语义版本号进行了多项改进\n* [改进] 加快了某些流水线取消或失败时的清理速度\n* [修复] 修复了新仪表板中实时日志显示的 bug\n* [修复] 在 Python >=3.10 中启用了对列表和集合的正确类型检查\n* [修复] 移除了 MNIST 示例中“负载过大”的可能性\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.34.0...v0.34.1\n\n## 兼容性\n此版本的 Sematic 服务器可以支持回退至 `v0.30.0` 的 pip 包。\n\n## 相应的 Helm Chart 版本\n\n1.1.16\n","2023-08-29T19:56:38",{"id":237,"version":238,"summary_zh":239,"released_at":240},197880,"v0.34.0","## What's Changed\r\n* [feature] New API to trigger a Pipeline rerun with Artifact ID overrides for the root run function's input parameters.\r\n* [improvement] Adjust header menu order to better align with user habits\r\n* [improvement] Restyle the state icons to make it more obvious\r\n* [improvement] Better error message for Ray cluster from non-standalone function\r\n* [improvement] Make runner reentrant\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.33.0...v0.34.0\r\n\r\n## Compatibility\r\nSematic Servers on this version can support pip packages back to `v0.30.0`.\r\n\r\n## Corresponding Helm Chart Release\r\n\r\n1.1.15\r\n","2023-08-15T19:03:17",{"id":242,"version":243,"summary_zh":244,"released_at":245},197881,"v0.33.0","## What's Changed\r\n* [feature] Enable the Grafana plugin in the new UI\r\n* [feature] Show user privacy in the new UI\r\n* [feature] Support expand\u002Fcollapse all in the new UI artifact display\r\n* [feature] New landing page in the new UI\r\n* [feature] Add `--no-cache` option for Docker builds\r\n* [improvement] Display cloned state icon for cloned runs in the new UI\r\n* [improvement] Changes to backend data model in preparation for upcoming features\r\n* [improvement] Improve run cancellations for local runs\r\n* [improvement] Minor visual improvements to new DAG view UI\r\n* [improvement] Add Llama 2 to fine-tuning example\r\n* [bugfix] Make the recent DB upgrade script more robust\r\n* [bugfix] Allow changing back to old UI from unauthenicated local executions\r\n* [bugfix] Resolve an issue with node expand\u002Fcollapses in new DAG view\r\n* [bugfix] Miscellaneous bug fixes in the new UI\r\n* [bugfix] Fix prompt display in fine-tuning example\r\n* [bugfix] Avoid collisions with public Docker repos for Docker builds\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.32.0...v0.33.0\r\n\r\n## Compatibility\r\nSematic Servers on this version can support pip packages back to `v0.30.0`.\r\n\r\n## Corresponding Helm Chart Release\r\n\r\n1.1.14\r\n","2023-07-25T23:21:46",{"id":247,"version":248,"summary_zh":249,"released_at":250},197882,"v0.32.0","## What's Changed\r\n* [feature] Publish a new version of the Dashboard UI, which is currently in \"Beta\". You can\r\n  switch to the new version by clicking on the pop-up banner, and between the two versions\r\n  through your profile window pop-up.\r\n* [improvement] Add constraints to the DB schema in order to improve validations. In case you\r\n  get any errors during the upgrade, please contact us on\r\n  [Discord](https:\u002F\u002Fdiscord.gg\u002F4KZJ6kYVax) so that we can assist you.\r\n* [improvement] Added documentation for custom user metrics[^1]\r\n* [improvement] Improve error messaging for unschedulable pipeline runs\r\n* [improvement] When running the CLI via Bazel, use the current directory as the working\r\n  directory\r\n* [improvement] Add Hugging Face model types, visualizations, and documentation\r\n* [example] Add an example pipeline which fine tunes LLMs that summarize a text\r\n* [deprecation] Ended backwards-compatibility support for `Calculator`, which had been renamed\r\n  to `Function` in v0.30.0\r\n* [deprecation] Ended backwards-compatibility support for pre-v0.27.0 log message sourcing and\r\n  for an API response serialization\r\n* [bugfix] Fix a bug where long running jobs' durations were not correctly rendered\r\n* [bugfix] Fix a regression where the URL generated in Slack updates was incorrect\r\n* [bugfix] Redact the DB URL from the migration logs\r\n* [bugfix] Fix a bug where the Native Docker Build System did not support image URIs in quotes\r\n* [bugfix] Add missing configuration to deployment documentation\r\n* [bugfix] Fix clipping of metrics graphs in the Dashboard when new data points are received\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.31.2...v0.32.0\r\n\r\n## Compatibility\r\nSematic Servers on this version can support pip packages back to `v0.30.0`.\r\n\r\n## Upgrade Instructions\r\nPlease see [these instructions](https:\u002F\u002Fdocs.sematic.dev\u002Fproject\u002Fupgrades#vx.x.x-to-v0.32.0) on how to ensure you can use the new version correctly!\r\n\r\n## Corresponding Helm Chart Release\r\n\r\n1.1.13\r\n\r\n## New Contributors\r\n* @twitchax made their first contribution in https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fpull\u002F944\r\n\r\nThank you @twitchax!\r\n\r\n----------------\r\n\r\n[^1]: This feature is for Sematic's \"Enterprise Edition\" only. Please reach out if\r\nyou are interested in using Sematic EE.\r\n","2023-07-14T14:28:13",{"id":252,"version":253,"summary_zh":254,"released_at":255},197883,"v0.31.2","### What's changed\r\n\r\n* [improvement] Add ability to customize images for Ray workers\r\n* [improvement] Add image pull secrets and pull policy to migration pod\r\n\r\nFull Changelog https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.31.1...v0.31.2\r\n\r\n### Compatibility\r\nSematic Servers on this version can support pip packages back to `v0.30.0`.","2023-06-30T00:28:42",{"id":257,"version":258,"summary_zh":259,"released_at":260},197884,"v0.31.1","### What's changed\r\n\r\n* [improvement] Add index on edges for source\u002Fdest run ids\r\n\r\nFull Changelog https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.31.0...v0.31.1\r\n\r\n### Compatibility\r\nSematic Servers on this version can support pip packages back to `v0.30.0`.","2023-06-28T23:57:13",{"id":262,"version":263,"summary_zh":264,"released_at":265},197885,"v0.31.0","### What's changed\r\n\r\n\r\n* [feature] Enable remote execution using pure-Docker, without bazel\r\n* [feature] Support live-metrics during Sematic Function execution[^1]\r\n* [feature] Add visualization for Prompt\u002FResponse pairs\r\n* [example] Add Hacker News summarization example\r\n* [improvement] Expose all Kubernetes classes in the base sematic module\r\n* [improvement] Switch from a WSGI server & gevent to an ASGI server (uvicorn)\r\n* [bugfix] Ensure UI-reruns don't automatically \"rerun from here\" for the root run\r\n* [bugfix] Fix an issue that prevented rendering of matplotlib figures\r\n* [bugfix] Remove a password that could be shown in cleaner logs\r\n* [bugfix] Include missing information from a local storage error message\r\n\r\nFull Changelog https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.30.0...v0.31.0\r\n\r\nThank you @kaushil24 for your contribution!\r\n\r\n### Compatibility\r\nSematic Servers on this version can support pip packages back to `v0.30.0`.\r\n\r\n----------------\r\n\r\n[^1]: This feature is for Sematic's \"Enterprise Edition\" only. Please reach out if\r\nyou are interested in using Sematic EE.\r\n","2023-06-20T19:42:46",{"id":267,"version":268,"summary_zh":269,"released_at":270},197886,"v0.30.0","### What's changed\r\n\r\n* [feature] Grafana dashboards tailored for Sematic installable via Helm\r\n* [feature] User metrics SDK\r\n* [feature] Add horizontal pod autoscaling and pod disruption budget support\r\n* [feature] Health endpoint to display DB health in dashboard\r\n* [feature] Add support for set types\r\n* [feature] Add GitHub commit check support\r\n* [improvement] Have resolver continue in in a new resolution if resolver restarts\r\n* [improvement] Log request IDs in server\r\n* [improvement] Early resolution failure is more robust\r\n* [improvement] Make API retries more robust\r\n* [improvement] Enable run search deep links\r\n* [deprecation] `Calculator` renamed to `Function`\r\n* [bugfix] Fixed a DAG view display issue\r\n* [bugfix] Fixed application logs being duplicated\r\n* [bugfix] Adjust the Name column width distribution\r\n* [bugfix] Add missing call to init to Function\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.29.0...v0.30.0\r\n\r\nThank you @kaushil24 for your contribution!\r\n\r\n### Compatibility\r\nSematic Servers on this version can support pip packages back to `v0.24.1`.\r\n\r\n### Upgrade Instructions\r\n\r\nDefault Kubernetes deployments of the Sematic server will now run with 2 pods for the API server, in order to enable high availability. As such, the memory and CPU requests and limits for each pod has been halved.","2023-05-31T01:02:13",{"id":272,"version":273,"summary_zh":274,"released_at":275},197887,"v0.29.0","### What's changed\r\n\r\n* [feature] Garbage collection CRON job\r\n* [feature] Enable customization of local storage path\r\n* [feature] Add Sematic Grafana dashboard as helm package\r\n* [feature] Add support for function timeout\r\n* [improvement] Enable backward logs scrolling\r\n* [improvement] Rename `inline=False` to `standalone=True`\r\n* [deprecation] Remove direct support for matplotlib figures, use `Image` instead\r\n* [bugfix] Constraint plotly version for MNIST example\r\n* [bugfix] Fix Kuberay autoscale\r\n\r\nFull Changelog https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.28.1...v0.29.0\r\n\r\n### Compatibility\r\n\r\nSematic Servers on this version can support pip packages back to `v0.24.1`.\r\n\r\n### New contributors\r\n\r\nMany thanks to @kaushil24 and @v-pwais.","2023-05-04T04:27:18",{"id":277,"version":278,"summary_zh":279,"released_at":280},197888,"v0.28.1","## What's Changed\r\n\r\n* [improvement] Allow selecting S3 paths in UI\r\n* [improvement] Backend logging improvements\r\n* [bugfix] Ensure gevent import doesn't monkeypatch standard lib late when importing Sematic\r\n* [bugfix] Resolve issue with S3 links in the UI for S3 \"directories\"\r\n* [bugfix] Eliminate one situation that could lead to duplicated logs in the UI\r\n* [bugfix] Fix a casting issue with floats\r\n    \r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.28.0...v0.28.1\r\n\r\n## Compatibility\r\nSematic Servers on this version can support pip packages back to `v0.24.1`.\r\n","2023-04-12T22:43:36",{"id":282,"version":283,"summary_zh":284,"released_at":285},197889,"v0.28.0","## What's Changed\r\n\r\n* [feature] Display metrics for pipelines (success rate, run count, runtime)\r\n* [feature] Added support for setting and memorizing a dev debug flag for the Dashboard\r\n* [deprecation] Deprecate Kubernetes 1.22 support\r\n* [improvement] Backend improvements to syncing with Kubernetes job states\r\n* [improvement] Several minor logging improvements and fixes\r\n* [bugfix] Enable local server to run with python 3.10\r\n* [bugfix] Ensured different users can rerun a pipeline\r\n* [bugfix] Ensured pipeline reruns use the submitting user's credentials\r\n* [bugfix] Ensure canceled\u002Fterminated runs have proper runtime display\r\n* [bugfix] Properly display duration for cloned runs\r\n* [bugfix] Fix matplotlib figure serialization, use Sematic Image type for support\r\n* [bugfix] Make storage object URL redirects consistent\r\n* [bugfix] Remove possible infinite reconnect loop when canceling local runs\r\n* [bugfix] Fixed a bug where the Resolver Socket.io client would not be cleanly closed\r\n* [bugfix] Fix corner case in the comparison of sqlite versions\r\n* [bugfix] Wrap long pipeline import paths in pipeline\u002Frun display\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fsematic-ai\u002Fsematic\u002Fcompare\u002Fv0.27.0...v0.28.0\r\n\r\n## Compatibility\r\nSematic Servers on this version can support pip packages back to `v0.24.1`.\r\n","2023-04-07T17:50:03"]