[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-9876":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":18,"compositeScore":20,"rankGlobal":10,"rankLanguage":10,"license":21,"archived":22,"fork":22,"defaultBranch":23,"hasWiki":22,"hasPages":22,"topics":24,"createdAt":10,"pushedAt":10,"updatedAt":45,"readmeContent":46,"aiSummary":47,"trendingCount":16,"starSnapshotCount":16,"syncStatus":48,"lastSyncTime":49,"discoverSource":50},9876,"metaflow","Netflix\u002Fmetaflow","Netflix","Build, Manage and Deploy AI\u002FML Systems","https:\u002F\u002Fmetaflow.org",null,"Python",10128,1294,286,294,0,4,16,47,44.34,"Apache License 2.0",false,"master",[25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44],"agents","ai","aws","azure","cost-optimization","datascience","distributed-training","gcp","generative-ai","high-performance-computing","kubernetes","llm","llmops","machine-learning","ml","ml-infrastructure","ml-platform","mlops","model-management","python","2026-06-12 02:02:13","![Metaflow_Logo_Horizontal_FullColor_Ribbon_Dark_RGB](https:\u002F\u002Fuser-images.githubusercontent.com\u002F763451\u002F89453116-96a57e00-d713-11ea-9fa6-82b29d4d6eff.png)\n\n# Metaflow\n\n[Metaflow](https:\u002F\u002Fmetaflow.org) is a human-centric framework designed to help scientists and engineers **build and manage real-life AI and ML systems**. Serving teams of all sizes and scale, Metaflow streamlines the entire development lifecycle—from rapid prototyping in notebooks to reliable, maintainable production deployments—enabling teams to iterate quickly and deliver robust systems efficiently.\n\nOriginally developed at [Netflix](https:\u002F\u002Fnetflixtechblog.com\u002Fopen-sourcing-metaflow-a-human-centric-framework-for-data-science-fa72e04a5d9) and now supported by [Outerbounds](https:\u002F\u002Fouterbounds.com), Metaflow is designed to boost the productivity for research and engineering teams working on [a wide variety of projects](https:\u002F\u002Fnetflixtechblog.com\u002Fsupporting-diverse-ml-systems-at-netflix-2d2e6b6d205d), from classical statistics to state-of-the-art deep learning and foundation models. By unifying code, data, and compute at every stage, Metaflow ensures seamless, end-to-end management of real-world AI and ML systems.\n\nToday, Metaflow powers thousands of AI and ML experiences across a diverse array of companies, large and small, including Amazon, Doordash, Dyson, Goldman Sachs, Ramp, and [many others](ADOPTERS.md). At Netflix alone, Metaflow supports over 3000 AI and ML projects, executes hundreds of millions of data-intensive high-performance compute jobs processing petabytes of data and manages tens of petabytes of models and artifacts for hundreds of users across its AI, ML, data science, and engineering teams.\n\n## From prototype to production (and back)\n\nMetaflow provides a simple and friendly pythonic [API](https:\u002F\u002Fdocs.metaflow.org) that covers foundational needs of AI and ML systems:\n\u003Cimg src=\".\u002Fdocs\u002Fprototype-to-prod.png\" width=\"800px\">\n\n1. [Rapid local prototyping](https:\u002F\u002Fdocs.metaflow.org\u002Fmetaflow\u002Fbasics), [support for notebooks](https:\u002F\u002Fdocs.metaflow.org\u002Fmetaflow\u002Fmanaging-flows\u002Fnotebook-runs), and built-in support for [experiment tracking, versioning](https:\u002F\u002Fdocs.metaflow.org\u002Fmetaflow\u002Fclient) and [visualization](https:\u002F\u002Fdocs.metaflow.org\u002Fmetaflow\u002Fvisualizing-results).\n2. [Effortlessly scale horizontally and vertically in your cloud](https:\u002F\u002Fdocs.metaflow.org\u002Fscaling\u002Fremote-tasks\u002Fintroduction), utilizing both CPUs and GPUs, with [fast data access](https:\u002F\u002Fdocs.metaflow.org\u002Fscaling\u002Fdata) for running [massive embarrassingly parallel](https:\u002F\u002Fdocs.metaflow.org\u002Fmetaflow\u002Fbasics#foreach) as well as [gang-scheduled](https:\u002F\u002Fdocs.metaflow.org\u002Fscaling\u002Fremote-tasks\u002Fdistributed-computing) compute workloads [reliably](https:\u002F\u002Fdocs.metaflow.org\u002Fscaling\u002Ffailures) and [efficiently](https:\u002F\u002Fdocs.metaflow.org\u002Fscaling\u002Fcheckpoint\u002Fintroduction).\n3. [Easily manage dependencies](https:\u002F\u002Fdocs.metaflow.org\u002Fscaling\u002Fdependencies) and [deploy with one-click](https:\u002F\u002Fdocs.metaflow.org\u002Fproduction\u002Fintroduction) to highly available production orchestrators with built in support for [reactive orchestration](https:\u002F\u002Fdocs.metaflow.org\u002Fproduction\u002Fevent-triggering).\n\nFor full documentation, check out our [API Reference](https:\u002F\u002Fdocs.metaflow.org\u002Fapi) or see our [Release Notes](https:\u002F\u002Fgithub.com\u002FNetflix\u002Fmetaflow\u002Freleases) for the latest features and improvements.\n\n\n## Getting started\n\nGetting up and running is easy. If you don't know where to start, [Metaflow sandbox](https:\u002F\u002Fouterbounds.com\u002Fsandbox) will have you running and exploring in seconds.\n\n### Installing Metaflow\n\nTo install Metaflow in your Python environment from [PyPI](https:\u002F\u002Fpypi.org\u002Fproject\u002Fmetaflow\u002F):\n\n```sh\npip install metaflow\n```\nAlternatively, using [conda-forge](https:\u002F\u002Fanaconda.org\u002Fconda-forge\u002Fmetaflow):\n\n```sh\nconda install -c conda-forge metaflow\n```\n\nOnce installed, a great way to get started is by following our [tutorial](https:\u002F\u002Fdocs.metaflow.org\u002Fgetting-started\u002Ftutorials). It walks you through creating and running your first Metaflow flow step by step.\n\nFor more details on Metaflow’s features and best practices, check out:\n- [How Metaflow works](https:\u002F\u002Fdocs.metaflow.org\u002Fmetaflow\u002Fbasics)\n- [Additional resources](https:\u002F\u002Fdocs.metaflow.org\u002Fintroduction\u002Fmetaflow-resources)\n\nIf you need help, don’t hesitate to reach out on our [Slack community](http:\u002F\u002Fslack.outerbounds.co\u002F)!\n\n\n### Deploying infrastructure for Metaflow in your cloud\n\u003Cimg src=\".\u002Fdocs\u002Fmulticloud.png\" width=\"800px\">\n\n\nWhile you can get started with Metaflow easily on your laptop, the main benefits of Metaflow lie in its ability to [scale out to external compute clusters](https:\u002F\u002Fdocs.metaflow.org\u002Fscaling\u002Fremote-tasks\u002Fintroduction)\nand to [deploy to production-grade workflow orchestrators](https:\u002F\u002Fdocs.metaflow.org\u002Fproduction\u002Fintroduction). To benefit from these features, follow this [guide](https:\u002F\u002Fouterbounds.com\u002Fengineering\u002Fwelcome\u002F) to\nconfigure Metaflow and the infrastructure behind it appropriately.\n\n\n## Get in touch\nWe'd love to hear from you. Join our community [Slack workspace](http:\u002F\u002Fslack.outerbounds.co\u002F)!\n\n## Contributing\nWe welcome contributions to Metaflow. Please see our [contribution guide](https:\u002F\u002Fdocs.metaflow.org\u002Fintroduction\u002Fcontributing-to-metaflow) for more details.\n","Metaflow 是一个专为科学家和工程师设计的框架，用于构建和管理实际的AI和ML系统。其核心功能包括从快速原型设计到生产部署的全流程支持，具备实验跟踪、版本控制及可视化等特性，并能无缝扩展至云端进行大规模并行计算。该工具特别适合需要高效迭代开发和维护复杂机器学习模型的应用场景，如统计分析、深度学习以及大型基础模型训练等。基于Python语言开发，Metaflow 旨在通过统一代码、数据与计算资源来简化AI\u002FML项目的生命周期管理，提高团队生产力。",2,"2026-06-11 03:25:09","top_topic"]