Developer Week 2023 现已正式结束。上周，我们发布了 34 个帖子， 介绍我们开发人员平台的最新动态，并展望未来发展方向 —— 包括新产品和功能，帮助您入门的详尽教程，以及激励您的客户故事。
我们非常喜欢你们对发布内容给出的反馈：
🤯 Serverless machine learning deployments - OMG! We used to need a team of devops to deploy a todo list MVP, it’s now 3 clicks with workers. They’re gonna do the same with ML workloads. Just like that. Boom. https://t.co/AcKUQ79fv0
— Thibaut Tiberghien (@tibotiber) May 16, 2023
.@Cloudflare is totally rocking it with these AI releases 🤘 https://t.co/7dQb6TUWg3
— Philipp Tsipman (@ptsi) May 17, 2023
Love this direction, for open source, for demos and running things on edge.
It's still in development, but as someone who've built an AI product on top of Cloudflare and has been asking for something like this, I'm really excited! https://t.co/AnywRDqecb
— Altryne - targum.video (@altryne) May 15, 2023
I've been waiting for this. Next on my list is a @remix_run app using @CockroachDB deployed to Workers. The new multi-region capabilities of #CockroachDB #serverless should pair wonderfully with @Cloudflare deployments. https://t.co/RwnJ2K0ToY
— Aydrian Howard (@itsaydrian) May 16, 2023
Yes! Loving this! I definitely this is the right direction and will help the general DevX and onboarding to the platform a lot ❤️! https://t.co/rEQWreeS96
— Adrien Baron (@BaronAdri) May 18, 2023
接下来的几周内，希望你们能用我们上周发布的新工具编写代码并进行试验。在构建的同时，欢迎加入我们的开发人员 Discord并分享您的想法。
如果您错过了任何公告，请查看这里的总结：
AI 相关公告
公告
| Announcement | Summary |
|---|---|
| Batteries included: how AI will transform the who and how of programming | The emergence of large language models (LLMs) is going to change the way developers write, debug, and modify code. Developer Platforms need to evolve to integrate AI capabilities to assist developers in their journeys. |
| Introducing Constellation, bringing AI to the Cloudflare stack | Run pre-trained machine learning models and inference tasks on Cloudflare’s global network with Constellation AI. We’ll maintain a catalog of verified and ready-to-use models, or you can upload and train your own. |
| Introducing Cursor: the Cloudflare AI Assistant | When getting started with a new technology comes a lot of questions on how to get started. Finding answers quickly is a time-saver. To help developers build in the fastest way possible we’ve introduced Cursor, an experimental AI assistant, to answer questions you may have about the Developer Platform. The assistant responds with both text and relevant links to our documentation to help you go further. |
| Query Cloudflare Radar and our docs using ChatGPT plugins | ChatGPT, recently allowed the ability for developers to create custom extensions to make ChatGPT even more powerful. It’s now possible to provide guidance to the conversational workflows within ChatGPT such as up-to-date statistics and product information. We’ve published plugins for Radar and our Developer Documentation and a tutorial showing how you can build your own plugin using Workers. |
| A complete suite of Zero Trust security tools to help get the most from AI | With any new technology comes concerns about risk and AI is no different. If you want to build with AI and maintain a Zero Trust security posture, Cloudflare One offers a collection of features to build with AI without increased risk. We’ve also compiled some best practices around securing your LLM. |
| Cloudflare R2 and MosaicML enable training LLMs on any compute, anywhere in the world, with zero switching costs | Training large language models requires massive amount of compute which has led AI companies to look at multi-cloud architectures, with R2 and MosaicML companies can build these infrastructures at a fraction of the cost. |
| The S3 to R2 Super Slurper is now Generally Available | After partnering with hundreds of early adopters to migrate objects to R2 during the beta, the Super Slurper is now generally available. |
| A raft of Cloudflare services for AI startups | AI startups no longer need affiliation with an accelerator or an employee referral to gain access to the Startup Program. Bootstrapped AI startups can apply today to get free access to Cloudflare services including R2, Workers, Pages, and a host of other security and developer services. |
| How to secure Generative AI applications | 11 tips for securing your generative AI application. |
| Using LangChain JS and Cloudflare Workers together | A tutorial on building your first LangChainJS and Workers application to build more sophisticated applications by switching between LLMs or chaining prompts together. |
摘要
| Announcement | Summary |
|---|---|
| Announcing database integrations: a few clicks to connect to Neon, PlanetScale, and Supabase on Workers | We’ve partnered with other database providers, including Neon, PlanetScale, and Supabase, to make authenticating and connecting back to your databases there just work, without having to copy-paste credentials and connection strings back and forth. |
| Announcing connect() - a new API for creating TCP sockets from Cloudflare Workers | Connect back to existing PostgreSQL and MySQL databases directly from Workers with outbound TCP sockets allowing you to connect to any database when building with Workers. |
| D1: We turned it up to 11 | D1 is now not only significantly faster, but has a raft of new features, including the ability to time travel: restore your database to any minute within the last 30 days, without having to make a manual backup. |
| Smart Placement speed up applications by moving code close to your backend - no config needed | Bringing compute closer to the end user isn’t always the right answer to improve performance. Smart Placement for Workers and Pages Functions moves compute to the optimal location whether that is closer to the end user or closer to backend services and data. |
| Use Snowflake with R2 to extend your global data lake | Get valuable insights from your data when you use Snowflake to query data stored in your R2 data lake and load data from R2 into Snowflake’s Data Cloud. |
| Developer Week Performance Update: Spotlight on R2 | Retrieving objects from storage needs to be fast. R2 is 20-40% faster than Amazon S3 when serving media content via public access. |
| Announcement | Summary |
|---|---|
| Making Cloudflare the best place for your web applications | Create Cloudflare CLI (C3) is a companion CLI to Wrangler giving you a single entry-point to configure Cloudflare via CLI. Pick your framework, all npm dependencies are installed, and you’ll receive a URL for where your application was deployed. |
| A whole new Quick Edit in Cloudflare Workers | QuickEdit for Workers powered by VSCode giving you a familiar environment to edit Workers directly in the dash. |
| Bringing a unified developer experience to Cloudflare Workers and Pages | Manage all your Workers scripts and Pages projects from a single place in the Cloudflare dashboard. Over the next year we’ll be working to converge these two separate experiences into one eliminating friction when building. |
| Modernizing the toolbox for Cloudflare Pages builds | Now in beta, the build system for Pages includes the latest versions of Node.js, Python, Hugo, and more. You can opt in to use this for existing projects or stay on the existing system, so your builds won’t break. |
| Improved local development with Wrangler and workerd | Having a local development environment that mimics production as closely as possible helps to ensure everything runs as expected in production. You can test every aspect prior to deployment. Wrangler 3 now leverages Miniflare3 based on workerd with local-by-default development. |
| Goodbye, section 2.8 and hello to Cloudflare’s new terms of service | Our terms of service were not clear about serving content hosted on the Developer Platform via our CDN. We’ve made it clearer that customers can use the CDN to serve video and other large files stored on the Developer Platform including Images, Pages, R2, and Stream. |
| More Node.js APIs in Cloudflare Workers-Streams, Pat, StringDecoder | We’ve expanded support Node.js APIs to increase compatibility with the existing ecosystem of open source npm packages. |
大型语言模型（LLM）的出现将改变开发人员编写、调试和修改代码的方式。开发人员平台需要与时驱进，整合 AI 能力来助力开发。
| Announcement | Summary |
|---|---|
| How Cloudflare is powering the next generation of platforms with Workers | A retrospective on the first year of Workers for Platform, what’s coming next, and featuring how customers like Shopify and Grafbase are building with it. |
| Building Cloudflare on Cloudflare | A technical deep dive into how we are rearchitecting internal services to use Workers. |
| Announcing Cloudflare Secrets Store | A centralized repository to store sensitive data for use across all of Cloudflare’s products. |
| Cloudflare Queues: messages at your speed with consumer concurrency and explicit acknowledgement | Announcing new features for Queues to ensure queues don’t fall behind, and processing time doesn’t slow down. |
| Workers Browser Rendering API enters open beta | Deploy a Worker script that requires Browser Rendering capabilities through Wrangler. |
隆重推出 Constellation，将 AI 加入 Cloudflare 堆栈
使用 Constellation AI 在 Cloudflare 的全球网络上运行预训练机器学习模型和推理任务。我们将维护一个经过验证且可立即使用的模型目录，您也可以上传并训练自己的模型。
当开始使用新技术时，往往会有很多关于如何入门的问题。快速找到答案是节省时间的好方法。为了帮助开发人员以最快的方式构建，我们推出了一个实验性的 AI 助手 Cursor，用于回答您有关开发人员平台的任何问题。这个助手会以文本和相关文档链接的形式给出回应，以帮助您进行下一步工作。
使用 ChatGPT 插件查询 Cloudflare Radar 和我们的文档
最近，ChatGPT 开始允许开发人员创建自定义扩展，使其更加强大。现在可以对 ChatGPT 中的对话工作流程提供指导，例如最新的统计数据和产品信息。我们已经发布了 Radar 和开发人员文档的插件，以及一个教程，展示如何使用 Workers 构建自己的插件.
一整套 Zero Trust 安全工具，帮助您充分利用 AI
随着任何新技术的出现都会引发对风险的担忧，AI 也不例外。如果您想使用 AI 构建并保持 Zero Trust 态势，Cloudflare One 提供一套功能集合，可在不增加风险的情况下使用 AI 进行构建 。我们还编制了一些有关确保 LLM 安全的最佳实践。
Cloudflare R2 与 MosaicML 合作，支持用户在世界各地的任意计算机上训练 LLM，转换成本为零
训练大型语言模型需要大量计算资源，这促使 AI 公司考虑多云架构。借助 R2 和 MosaicML，AI 公司能够以低得多的成本构建这些基础设施。
Super Slurper 现已正式发布，可将数据从 S3 迁移到 R2
在 beta 测试期间与数百个早期采用者合作将对象迁移到 R2 后，Super Slurper 现已普遍可用。
AI 初创公司不再需要与加速器的关联或员工推荐就能参加 Startup Program。AI 初创公司今天就可以申请免费使用 Cloudflare 服务的，包括 R2、Workers、Pages 以及其他多项安全和开发人员服务。
保护生成式 AI 应用程序的 11 个提示
搭配使用 LangChainJS 与 Cloudflare Workers
一个有关构建您第一个 LangChainJS 和 Workers 应用程序的教程，可通过在 LLM 之间切换或链接提示来构建更复杂的应用程序。
数据相关公告
公告
摘要
宣布推出 Database Integrations，单击几下鼠标即可连接到 Workers 上的 Neon、PlanetScale 和 Supabase 数据库
我们与其他数据库提供商进行合作，包括 Neon、PlanetScale 和 Supabase，简化了身份验证和连接数据库的操作，无需来回复制/粘贴凭据和连接字符串。
宣布推出新的 API：connect() ，用于从 Cloudflare Workers 创建 TCP 套接字
使用出站 TCP 套接字，您可以直接从 Workers 连接到现有的 PostgreSQL 和 MySQL 数据库，让您在使用 Workers 构建时可以连接到任何数据库。
D1 现在不仅显著提高了速度，还增加了大量新功能，包括 Time Travel：可将数据库恢复到过去 30 天内的任一分钟，无需进行手动备份。
Smart Placement 通过将代码移动到接近后端的地方来加快应用程序运行速度，无需配置
将计算资源更接近终端用户并不总是改善性能的正确答案。Smart Placement 将计算资源移动到最佳位置，无论是更接近终端用户还是更接近后端服务和数据。
当您使用 Snowflake 查询存储在 R2 数据湖中的数据，并将数据从 R2 加载到 Snowflake 的数据云中时，可以从数据中获得有用的洞察。
从存储中检索对象需要快速完成。使用公共访问方式提供媒体内容时，R2 比 Amazon S3 快20-40%。
开发人员体验相关公告
公告
摘要
让 Cloudflare 成为托管 Web 应用程序的不二之选
Create Cloudflare CLI (C3) 是 Wrangler 的辅助 CLI，为您提供通过 CLI 配置 Cloudflare 的单一入口点。选择您的框架，安装所有 npm 依赖项，您将收到应用部署位置的 URL。
QuickEdit for Workers 由 VSCode 驱动，为您提供一个熟悉的环境，在仪表板中直接编辑 Workers。
为使用 Cloudflare Workers 和 Pages 的开发人员提供统一体验
在 Cloudflare 仪表板中的单一位置管理所有 Workers 脚本和 Pages 项目。在未来一年中，我们将致力于将这两个独立的体验融为一体，消除构建过程中的摩擦。
实现 Cloudflare Pages 内部版本工具箱的现代化
目前处于 beta 测试状态的 Pages 构建系统包括最新版的 Node.js、Python、Hugo 和其他语言。您可以选择为现有项目使用这个构建系统，或者继续使用现有系统，以确保构建不会出现问题。
拥有尽可能接近生产环境的本地开发环境，有助于确保在生产环境中一切都按预期运行。您可以在部署之前对每个方面进行测试。Wrangler 3 现在利用基于 workerd 的 Miniflare3 进行本地默认开发。
删除第 2.8 条并推出更新版 Cloudflare 服务条款
对于通过我们的 CDN 提供托管在开发人员平台上的内容，我们的服务条款没有明确规定。我们已经进一步明确，客户可以使用 CDN 来提供存储在开发人员平台上的视频和其他大文件，包括 Images、Pages、R2 和 Stream。
Cloudflare Workers-Streams、Path、StringDecoder 中提供更多 Node.js API
我们扩展了对 Node.js API 的支持，以增加与现有开源 npm 包生态系统的兼容性。
等等，还有更多内容
公告
摘要
Cloudflare 如何使用 Workers 驱动下一代平台
回顾 Workers for Platform 的第一年，介绍接下来的计划，并展示 Shopify 和 Grafbase 等客户如何使用它进行构建。
以 Cloudflare 为基础，构建 Cloudflare 解决方案
深度技术剖析：我们如何重构内部服务以使用 Workers。
Cloudflare 推出集中式存储库，用于存储所有 Cloudflare 产品的敏感数据。
Cloudflare Queues：借助消费者并发请求处理和显式应答功能，快速发送消息
宣布 Queues 新功能，确保队列不会滞后，处理时间不会减慢。
Workers Browser Rendering API 进入公测
通过 Wrangler 部署需要浏览器渲染能力的 Worker 脚本。
在 Cloudflare TV 上观看
如果您错过了任何公告，或想观看博客作者详细介绍每一个公告的相关 Cloudflare TV 节目， 可在 Cloudflare TV 上观看全部 Developer Week 视频。