Update to Vercel

{
  "name": "runtime-cache",
  "description": "Vercel Runtime Cache API guidance — ephemeral per-region key-value cache with tag-based invalidation. Shared across Functions, Routing Middleware, and Builds. Use when implementing caching strategies beyond framework-level caching.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 169
    }
  ],
  "skill_md_contents": "---\nname: runtime-cache\ndescription: Vercel Runtime Cache API guidance — ephemeral per-region key-value cache with tag-based invalidation. Shared across Functions, Routing Middleware, and Builds. Use when implementing caching strategies beyond framework-level caching.\nmetadata:\n  priority: 6\n  docs:\n    - \"https://nextjs.org/docs/app/building-your-application/caching\"\n  sitemap: \"https://nextjs.org/sitemap.xml\"\n  pathPatterns: \n    - 'lib/cache/**'\n    - 'src/lib/cache/**'\n    - 'lib/cache.*'\n    - 'src/lib/cache.*'\n  bashPatterns:\n    - '\\bnpm\\s+(install|i|add)\\s+[^\\n]*@vercel/functions\\b'\n    - '\\bpnpm\\s+(install|i|add)\\s+[^\\n]*@vercel/functions\\b'\n    - '\\bbun\\s+(install|i|add)\\s+[^\\n]*@vercel/functions\\b'\n    - '\\byarn\\s+add\\s+[^\\n]*@vercel/functions\\b'\n---\n\n# Vercel Runtime Cache API\n\nYou are an expert in the Vercel Runtime Cache — an ephemeral caching layer for serverless compute.\n\n## What It Is\n\nThe Runtime Cache is a **per-region key-value store** accessible from Vercel Functions, Routing Middleware, and Builds. It supports **tag-based invalidation** for granular cache control.\n\n- **Regional**: Each Vercel region has its own isolated cache\n- **Isolated**: Scoped per project AND per deployment environment (`preview` vs `production`)\n- **Persistent across deployments**: Cached data survives new deploys; invalidation via TTL or `expireTag`\n- **Ephemeral**: Fixed storage limit per project; LRU eviction when full\n- **Framework-agnostic**: Works with any framework via `@vercel/functions`\n\n## Key APIs\n\nAll APIs from `@vercel/functions`:\n\n### Basic Cache Operations\n\n```ts\nimport { getCache } from '@vercel/functions';\n\nconst cache = getCache();\n\n// Store data with TTL and tags\nawait cache.set('user:123', userData, {\n  ttl: 3600,                      // seconds\n  tags: ['users', 'user:123'],    // for bulk invalidation\n  name: 'user-profile',           // human-readable label for observability\n});\n\n// Retrieve cached data (returns value or undefined)\nconst data = await cache.get('user:123');\n\n// Delete a specific key\nawait cache.delete('user:123');\n\n// Expire all entries with a tag (propagates globally within 300ms)\nawait cache.expireTag('users');\nawait cache.expireTag(['users', 'user:123']); // multiple tags\n```\n\n### Cache Options\n\n```ts\nconst cache = getCache({\n  namespace: 'api',                    // prefix for keys\n  namespaceSeparator: ':',             // separator (default)\n  keyHashFunction: (key) => sha256(key), // custom key hashing\n});\n```\n\n### Full Example (Framework-Agnostic)\n\n```ts\nimport { getCache } from '@vercel/functions';\n\nexport default {\n  async fetch(request: Request) {\n    const cache = getCache();\n    const cached = await cache.get('blog-posts');\n\n    if (cached) {\n      return Response.json(cached);\n    }\n\n    const posts = await fetch('https://api.example.com/posts').then(r => r.json());\n\n    await cache.set('blog-posts', posts, {\n      ttl: 3600,\n      tags: ['blog'],\n    });\n\n    return Response.json(posts);\n  },\n};\n```\n\n### Tag Expiration from Server Action\n\n```ts\n'use server';\nimport { getCache } from '@vercel/functions';\n\nexport async function invalidateBlog() {\n  await getCache().expireTag('blog');\n}\n```\n\n## CDN Cache Purging Functions\n\nThese purge across **all three cache layers** (CDN + Runtime Cache + Data Cache):\n\n```ts\nimport { invalidateByTag, dangerouslyDeleteByTag } from '@vercel/functions';\n\n// Stale-while-revalidate: serves stale, revalidates in background\nawait invalidateByTag('blog-posts');\n\n// Hard delete: next request blocks while fetching from origin (cache stampede risk)\nawait dangerouslyDeleteByTag('blog-posts', {\n  revalidationDeadlineSeconds: 3600,\n});\n```\n\n**Important distinction**:\n- `cache.expireTag()` — operates on Runtime Cache only\n- `invalidateByTag()` / `dangerouslyDeleteByTag()` — purges CDN + Runtime + Data caches\n\n## Next.js Integration\n\n### Next.js 16+ (`use cache: remote`)\n\n```ts\n// next.config.ts\nconst nextConfig: NextConfig = { cacheComponents: true };\n```\n\n```ts\nimport { cacheLife, cacheTag } from 'next/cache';\n\nasync function getData() {\n  'use cache: remote'     // stores in Vercel Runtime Cache\n  cacheTag('example-tag')\n  cacheLife({ expire: 3600 })\n  return fetch('https://api.example.com/data').then(r => r.json());\n}\n```\n\n- `'use cache'` (no `: remote`) — in-memory only, ephemeral per instance\n- `'use cache: remote'` — stores in Vercel Runtime Cache\n\n### Next.js 16 Invalidation APIs\n\n| Function | Context | Behavior |\n|----------|---------|----------|\n| `updateTag(tag)` | Server Actions only | Immediate expiration, read-your-own-writes |\n| `revalidateTag(tag, 'max')` | Server Actions + Route Handlers | Stale-while-revalidate (recommended) |\n| `revalidateTag(tag, { expire: 0 })` | Route Handlers (webhooks) | Immediate expiration from external triggers |\n\n**Important**: Single-argument `revalidateTag(tag)` is deprecated in Next.js 16. Always pass a `cacheLife` profile as the second argument.\n\n### Runtime Cache vs ISR Isolation\n\n- Runtime Cache tags do **NOT** apply to ISR pages\n- `cache.expireTag` does **NOT** invalidate ISR cache\n- Next.js `revalidatePath` / `revalidateTag` does **NOT** invalidate Runtime Cache\n- To manage both, use same tag and purge via `invalidateByTag` (hits all cache layers)\n\n## CLI Cache Commands\n\n```bash\n# Purge all cached data\nvercel cache purge                    # CDN + Data cache\nvercel cache purge --type cdn         # CDN only\nvercel cache purge --type data        # Data cache only\nvercel cache purge --yes              # skip confirmation\n\n# Invalidate by tag (stale-while-revalidate)\nvercel cache invalidate --tag blog-posts,user-profiles\n\n# Hard delete by tag (blocks until revalidated)\nvercel cache dangerously-delete --tag blog-posts\nvercel cache dangerously-delete --tag blog-posts --revalidation-deadline-seconds 3600\n\n# Image invalidation\nvercel cache invalidate --srcimg /images/hero.jpg\n```\n\nNote: `--tag` and `--srcimg` cannot be used together.\n\n## CDN Cache Tags\n\nAdd tags to CDN cached responses for later invalidation:\n\n```ts\nimport { addCacheTag } from '@vercel/functions';\n\n// Via helper\naddCacheTag('product-123');\n\n// Via response header\nreturn Response.json(product, {\n  headers: {\n    'Vercel-CDN-Cache-Control': 'public, max-age=86400',\n    'Vercel-Cache-Tag': 'product-123,products',\n  },\n});\n```\n\n## Limits\n\n| Property | Limit |\n|----------|-------|\n| Item size | 2 MB |\n| Tags per Runtime Cache item | 64 |\n| Tags per CDN item | 128 |\n| Max tag length | 256 bytes |\n| Tags per bulk REST API call | 16 |\n\nTags are **case-sensitive** and **cannot contain commas**.\n\n## Observability\n\nMonitor hit rates, invalidation patterns, and storage usage in the Vercel Dashboard under **Observability → Runtime Cache**. The CDN dashboard (March 5, 2026) provides a unified view of global traffic distribution, cache performance metrics, a redesigned purging interface, and **project-level routing** — update response headers or rewrite to external APIs without triggering a new deployment. Project-level routes are available on all plans and take effect instantly.\n\n## When to Use\n\n- Caching API responses or computed data across functions in a region\n- Tag-based invalidation when content changes (CMS webhook → expire tag)\n- Reducing database load for frequently accessed data\n- Cross-function data sharing within a region\n\n## When NOT to Use\n\n- Framework-level page caching → use Next.js Cache Components (`'use cache'`)\n- Persistent storage → use a database (Neon, Upstash)\n- CDN-level full response caching → use `Cache-Control` / `Vercel-CDN-Cache-Control` headers\n- Cross-region shared state → use a database\n- User-specific data that differs per request\n\n## References\n\n- 📖 docs: https://vercel.com/docs/runtime-cache\n- 📖 changelog: https://vercel.com/changelog/introducing-the-runtime-cache-api\n- 📖 CLI cache: https://vercel.com/docs/cli/cache\n- 📖 CDN cache purging: https://vercel.com/docs/cdn-cache/purge\n"
}