Shopify推出的通用商务协议（UCP）为AI代理与商家交易提供开放标准，支持灵活的能力发现与谈判。UCP分层设计，核心交易功能与扩展能力分离，商家可自定义功能，代理仅处理支持的操作。协议无需中央审批，支持动态支付谈判与无缝交接，适应多样化的商务需求，推动电商生态持续进化。

Shopify利用SkyPilot框架实现多云GPU资源智能调度，通过YAML声明式配置自动路由任务——H200需求分发给Nebius集群，L4/CPU任务定向到GCP。定制插件集成Kueue实现配额管理与优先级调度，开发环境一键启停。系统自动处理InfiniBand配置和缓存挂载，工程师只需专注资源需求描述，底层跨云复杂性由平台消化。（139字）

Shopify构建了一套基于用户行为序列的生成式推荐系统，通过时序编码捕捉购物场景的季节性变化，利用负采样技术优化商品边界区分，并采用增量召回策略增强推荐多样性。该系统将买家旅程转化为连续事件预测任务，结合高效CUDA内核实现低延迟推理，在BFCM大促期间成功提升5%高价值点击率及0.71%转化率。当前正探索语义ID技术以突破商品ID空间的算力限制。

Shopify开发了GraphQL Cardinal，一款广度优先执行引擎，解决深度优先遍历在高基数查询中的性能瓶颈。通过一次性解析所有对象的字段，Cardinal显著提升了大型列表查询的速度和内存效率，减少了CPU开销和垃圾回收压力。迁移过程中，Shopify逐步更新了字段解析器和追踪器，进一步优化性能。Cardinal的成功为GraphQL社区提供了新的执行模型思路，展示了广度优先策略在处理大规模数据时的潜力。

线上购物愈发便捷，实体店的价值转向体验化。AI魔镜应运而生，通过视觉识别和互动，为顾客提供个性化信息、产品推荐等，提升购物趣味性。魔镜核心由全尺寸镜子、摄像头和计算机组成，运行HTTP协议，实现快速互动。借助Remix框架，服务器和客户端逻辑统一，简化开发流程。魔镜可自定义，适用于多种场景，如美妆试色、互动挑战等，为品牌创造独特购物体验。

面对电商产品分类系统的快速扩展，传统手动维护分类法难以应对新兴产品和市场变化。为此，开发了一套基于AI的多智能体系统，通过结构分析和产品驱动分析，自动发现并优化分类标签，确保分类法的灵活性和前瞻性。该系统结合质量保障机制，提升分类一致性和效率，显著缩短了新兴类别的更新周期，为商家和顾客提供更精准的分类支持。

Shopify的AI助手Sidekick通过“代理循环”架构处理商户需求，从简单工具调用发展为复杂代理平台。面对工具复杂度问题，采用“即时指令”策略提升系统维护性和性能。构建稳健的LLM评估系统，采用真实数据集和LLM作为评判者，确保评估质量。通过GRPO训练和奖励破解应对，不断优化模型表现。生产级代理系统需注重架构设计、评估方法和训练技巧，以实现可靠AI助手。

Shopify成功将Shopify Mobile和POS应用迁移至React Native新架构，保持每周发布并服务数百万商户。迁移涉及复杂代码库、数百个屏幕和原生模块，通过最小化代码更改、保持双架构兼容性及维护性能稳定，确保了开发速度和功能不中断。过程中解决了状态批处理、空白屏幕等常见问题，最终提升了应用启动时间和渲染性能。这一经验为其他团队提供了宝贵的迁移策略参考。

Shopify通过结合视觉语言模型和标准化产品分类体系，显著提升了产品理解能力。从基础分类到多模态融合，再到全面理解产品属性，系统不断进化。当前架构采用FP8量化和动态批处理优化推理效率，支持实时处理数百万产品预测。未来将引入DAG结构，扩展多语言支持和更细粒度属性提取，持续优化平台搜索、推荐和税收计算等功能。

Shopify 面对移动应用中大量非核心页面的开发挑战，创新性地推出了 Mobile Bridge 框架，优化了 WebViews 的性能、外观和体验。通过预加载、缓存和重用 WebViews，加载时间从 6 秒降至 1.4 秒。同时，禁用缩放、调整 UI 元素及引入 Polaris 设计，让 WebViews 更接近原生体验。Mobile Bridge 还支持双向通信，解决了导航、模态框等原生交互问题，并整合了原生功能如日期选择器和条码扫描器，极大提升了用户体验。

Every year for Black Friday Cyber Monday (BFCM), we put together a real-time visualization of purchases made through Shopify-powered merchants worldwide. This year we're cooking up something big, and…

In the dynamic landscape of commerce, Shopify merchants rely on our platform's ability to seamlessly and reliably deliver highly relevant products to potential customers. Therefore, a rich and…

Insight into how Shopify created a seamless one-click manual testing experience (known as "tophatting") for mobile applications by creating a macOS app.

One line of code can 5x your Rails application performance. What if you could make a small change to your database design that would unlock massively more efficient data access? At Shopify, we dusted off some old database principles and did exactly that with the primary Rails application.

As an integral part of Shopify's ecosystem, our mobile app serves millions of merchants around the world every single day. It allows them to run their business from anywhere and offers vital insights…

In September, 2020, our team at Shopify released a Ruby gem named Packwerk, a tool to enforce boundaries and modularize Rails applications. Since its release, Packwerk has taken on a life of its own, inspiring blog posts, conference talks, and even an entire gem ecosystem. Its popularity is an indication that Packwerk clearly filled a void in the Rails community.

Packwerk is a static analysis tool, similar to tools like Rubocop and Sorbet. Applied to a codebase, it analyzes constant references to help decouple code and organize it into well-defined packages.

But Packwerk is more than just a tool. Over the years, Packwerk’s approach to modularity has come to embody distinct and sometimes conflicting perspectives on code organization and its evolution. Packwerk’s feedback can change the entire trajectory of a codebase to a degree that distinguishes it from other tools of its kind.

This retrospective is our effort, as the team that developed Packwerk at Shopify, to shine a light on our learnings working with the tool, concerns about its use, and hopes for its future.