sanity-live-cache-components

作者: sanity-io

将next-sanity应用迁移至cacheComponents——严格模式、三层组件模式、显式perspective/stega/includeDrafts、属性透传约定

npx skills add https://github.com/sanity-io/next-sanity --skill sanity-live-cache-components

Sanity Live + Cache Components

Wires next-sanity into a Next.js 16+ app with cacheComponents: true. Data is fetched with sanityFetch (which calls cacheTag/cacheLife internally), and <SanityLive> in the root layout revalidates cached content over an EventSource connection to Sanity Content Lake. Visual Editing and Presentation Tool are fully supported when draft mode is enabled.

Read the relevant guide in node_modules/next/dist/docs/ (when available) before writing code. If a guide conflicts with this skill, follow this skill.

This skill assumes familiarity with the next-cache-components skill — it covers 'use cache', cacheLife, cacheTag, and the cookies/headers/params rule. The only Sanity-relevant exception: await draftMode() is allowed inside 'use cache' (Next.js bypasses caching when draft mode is enabled — see the use cache reference).

Prerequisites

  • Next.js 16.2+ installed in the project (check package.json or run pnpm list next / npm ls next — don't use pnpm view next version, that reports the registry's latest, not what's installed).
  • AGENTS.md exists, or follow the guide.
  • These environment variables are set:
    • NEXT_PUBLIC_SANITY_PROJECT_ID
    • NEXT_PUBLIC_SANITY_DATASET
    • SANITY_API_READ_TOKEN
  • Embedded Sanity Studio configuration (sanity.config.ts, sanity.cli.ts, anything under sanity/) needs no changes — this skill only touches the Next.js app surface.

Reference files

FileWhen to read
reference/live-helpers.mdFull client.ts / live.ts, sanityFetch* and getDynamicFetchOptions details
reference/three-layer-pattern.mdThe Page → Dynamic → Cached pattern for page.tsx, including the searchParams variant
reference/layouts.mdNon-blocking data fetching inside layout.tsx with a shared 'use cache' helper
reference/dynamic-segments.mdHigh-performance [slug] routes: loading.tsx + partial generateStaticParams, or non-blocking dynamic params in a layout

1. Install next-sanity@^13

npm install next-sanity@^13 --save-exact

Migrating an existing Sanity Live setup

If the app is already using defineLive, this skill is a refactor, not a rewrite. The 5-step sequence below still applies, but watch for these specific differences:

  • Don't overwrite client.ts or live.ts if they exist. Append missing options. Preserve any existing token and stega.* settings — see reference/live-helpers.md.
  • Search the codebase for hardcoded perspective: 'published' and stega: false in sanityFetch callsites and refactor them to source perspective/stega via getDynamicFetchOptions and the three-layer pattern.
  • Search for sanityFetch calls inside generateStaticParams → swap for sanityFetchStaticParams.
  • Search for sanityFetch calls inside generateMetadata / sitemap.ts / opengraph-image.tsx / etc. → swap for sanityFetchMetadata.
  • Search for sanityFetch calls directly inside a 'use server' function → split into a separate 'use cache' helper.
  • Verify there is exactly one <SanityLive> and one <VisualEditing> in the tree. Multiple renders are undefined behavior.

The "Anti-patterns to grep for" section at the bottom of this file lists the search patterns.


2. Configure next.config.ts

Enable cacheComponents and set cacheLife.default to sanity so default revalidation is 1 year (instead of 15 minutes). sanityFetch is optimized for on-demand revalidation and doesn't need time-based revalidation.

// next.config.ts
import type {NextConfig} from 'next'
import {sanity} from 'next-sanity/live/cache-life'

const nextConfig: NextConfig = {
  cacheComponents: true,
  cacheLife: {default: sanity},
}

export default nextConfig

3. Configure defineLive and export helpers

Create src/sanity/lib/client.ts and src/sanity/lib/live.ts. The minimal defineLive call:

// src/sanity/lib/live.ts (excerpt)
export const {SanityLive, sanityFetch} = defineLive({
  client,
  serverToken: token,
  browserToken: token,
  strict: true,
})

Full file contents (including client.ts, getDynamicFetchOptions, sanityFetchMetadata, sanityFetchStaticParams) and per-helper guidance: reference/live-helpers.md.

The helpers exported from live.ts:

HelperUsed in
sanityFetch'use cache' components rendered from page.tsx / layout.tsx
sanityFetchMetadatagenerateMetadata, generateViewport, sitemap.ts, robots.ts, opengraph-image.tsx, etc.
sanityFetchStaticParamsgenerateStaticParams only
getDynamicFetchOptionsResolving perspective/stega outside any 'use cache' boundary
SanityLiveRendered once in a root layout

4. Render <SanityLive> in a root layout

<SanityLive> and <VisualEditing> both belong in a layout.tsx, never a page.tsx. Both must be rendered at most once across the whole tree — duplicate renders are undefined behavior.

  • includeDrafts is required when defineLive is configured with strict: true (the recommended setup). TypeScript will surface the error if it's missing; pass includeDrafts={isDraftMode} so live revalidation includes drafts only in draft mode.
  • Preserve any existing optional callback props on <SanityLive> when migrating: onError, onWelcome, onReconnect. They are commonly wired to a toast/notification helper and silently dropping them regresses UX.
// src/app/layout.tsx
import {SanityLive} from '@/sanity/lib/live'
import {VisualEditing} from 'next-sanity/visual-editing'
import {draftMode} from 'next/headers'

export default async function RootLayout({children}: LayoutProps<'/'>) {
  const {isEnabled: isDraftMode} = await draftMode()
  return (
    <html lang="en">
      <body>
        {children}
        <SanityLive includeDrafts={isDraftMode} />
        {isDraftMode && <VisualEditing />}
      </body>
    </html>
  )
}

With an embedded Sanity Studio

If a route mounts NextStudio from next-sanity/studio (e.g. app/studio/[[...index]]/page.tsx), <SanityLive> must live in a layout the embedded studio doesn't share. Use route groups: put <SanityLive> in src/app/(website)/layout.tsx and keep the rest of the app under src/app/(website).


5. Apply the three-layer pattern to pages and layouts

Every route that should be statically prerendered uses the same shape:

Page/Layout (Layer 1: draftMode branch)
  ├── NOT draft mode → <CachedX perspective="published" stega={false} />  (no Suspense)
  └── draft mode → <Suspense fallback={...}>
                      <DynamicX params={params} />  (Layer 2: awaits dynamic APIs)
                        └── <CachedX perspective={p} stega={s} />  (Layer 3: 'use cache')

Critical rule: Only Layer 3 carries 'use cache'. The top-level Page / Layout must not have 'use cache' — it awaits params, searchParams, or cookies() (via getDynamicFetchOptions), and those dynamic APIs are forbidden inside 'use cache'. Layer 3 carrying 'use cache' is enough for the whole route to prerender into the static shell. Adding 'use cache' to the top-level function is the most common failure mode — TypeScript and the runtime will both complain.

Pick the right reference for the file you're editing:


Anti-patterns to grep for

When auditing an app, search for these and refactor:

  • perspective: 'published' and stega: false hardcoded together in a sanityFetch call → use the three-layer pattern, source perspective/stega via getDynamicFetchOptions.
  • sanityFetch( directly inside a function whose body begins with 'use server' → split into a separate 'use cache' helper.
  • sanityFetch( inside generateStaticParams → swap for sanityFetchStaticParams.
  • sanityFetch( inside generateMetadata / generateViewport / sitemap.ts / robots.ts / opengraph-image.tsx etc. → swap for sanityFetchMetadata and resolve perspective via getDynamicFetchOptions.
  • await draftMode() immediately followed by await getDynamicFetchOptions() at the top of a page.tsx or layout.tsx without a sibling loading.tsx → move those dynamic-API calls into a child component wrapped in <Suspense> so the static shell can prerender.
  • More than one <SanityLive> or <VisualEditing> rendered in the tree → consolidate to a single render in the right layout.

来自 sanity-io 的更多技能

sanity-migration
sanity-io
规划、实施并审查从其他CMS和内容系统迁移至Sanity的过程。适用于从AEM、Adobe Experience Manager、Contentful、Strapi、Webflow、WordPress、Payload、Drupal、Markdown/MDX/frontmatter文件、WXR/XML导出、CMS API、数据库转储、静态HTML进行迁移或平台重构,或设计数据提取、转换、Portable Text转换、资产迁移、重定向、验证及切换工作流时使用。
officialdevelopmentdatabase
create-agent-with-sanity-context
sanity-io
通过Agent Context构建对Sanity内容的结构化访问的AI代理。用于设置由Sanity驱动的聊天机器人、将AI助手连接到Sanity…
official
dial-your-context
sanity-io
交互式会话,用于为Sanity Agent Context MCP创建指令字段内容。当用户提到调整代理上下文、改进……时,使用此技能。
official
optimize-agent-prompt
sanity-io
通过引导式对话调整你的Sanity Agent Context代理。将探索数据转化为可用于生产的指令,并构建系统提示词……
official
shape-your-agent
sanity-io
交互式会话,用于为基于Sanity Agent Context MCP的AI代理编写系统提示。当用户希望定义代理个性时使用此技能…
official
content-experimentation-best-practices
sanity-io
结构化指导,用于设计、执行和分析内容实验,以提升转化率和参与度。涵盖假设框架、指标选择、样本量计算以及A/B和多变量实验中的统计显著性检验。包含关于p值、置信区间、功效分析和贝叶斯方法的详细资源,用于解读结果。提供CMS集成模式,用于在字段级别管理变体并连接外部...
official
content-modeling-best-practices
sanity-io
结构化内容建模指南,涵盖模式设计、可复用性及多渠道交付。核心原则包括:将内容视为数据而非页面、维护单一事实来源、面向未来渠道设计、优化编辑工作流。提供引用与嵌入对象的选择框架、关注点分离及内容复用模式。包含扁平化、层级化及分面分类法的分类学指导。适用于...
official
portable-text-conversion
sanity-io
将HTML和Markdown内容转换为适用于Sanity的Portable Text块。在从旧版CMS迁移内容、将HTML或Markdown导入Sanity时使用。
official