POST /v1/sites/{siteId}/vitals/checks

認証

ベアラー トークンとして API キーを送信します。このエンドポイントでは仕様に特定の権限が記載されていないため、キーに必要な最小限の権限を付与し、推測するのではなくレスポンスを確認してください。

このエンドポイントは組織IDを受け付けません。お使いのキーによって所属する組織がすでに特定されており、レスポンスはその組織にスコープされます。

試してみる

アングルブラケット内のすべてをご自身の値に置き換え、キーのプレースホルダーをご利用中のダッシュボードのキーに置き換えてください。

curl -X POST https://api.zinndigital.com/v1/sites/{siteId}/vitals/checks \
  -H "Authorization: Bearer zdk_live_…" \
  -H "Content-Type: application/json" \
  -d '{ "url": <string> }'

ログインしていますか？ダッシュボード内のAPIコンソールでは、実際の組織IDやお客様ご自身のキーが自動入力され、ライブAPIに対してリクエストが実行されるため、実際のレスポンスを確認することができます。 API コンソールでこのエンドポイントを開く

詳細

`getSiteVitals` measures the **homepage** on a schedule, which makes the product descriptive. This makes it diagnostic: *"the homepage is fine, it is `/shop/category/boots` that is slow"*. Gated on `sites.view` — asking how fast your own page is changes nothing about it. **Returns `202`, never the measurement.** A Lighthouse run takes 10-30 seconds; holding the request open is the uncached hot-path work CLAUDE.md 2.16 forbids and it times out behind Cloudflare. Poll `getSiteVitalsCheck` with the returned `id` until `status` is `done` or `failed`. **A URL outside this site is a `404`, not a `403`.** A 403 would confirm that some other organisation owns that hostname, turning this endpoint into an ownership oracle over the estate. The submitted address may be a path (`/shop/boots`), a bare hostname, or an absolute URL; it is normalised — lowercased, port and trailing dot stripped, leading `www.` folded — before being compared against the site's `primary_domain` and its aliases. The URL that comes back is rebuilt on whichever of the site's own hostnames matched, **not** on the folded form and not on what was typed: a site that lives on `www.` is measured on `www.`, because the apex may 301 or not resolve at all. Clients should display the returned `url`. **A recent identical check is returned instead of starting a new one.** Within the cache window the stored measurement comes back with `cached: true` and `status` already `done`, and it costs no quota. Clients should render `measured_at` rather than implying the page was just re-measured. **A check of the same URL that is still running is returned too** — same `202`, `cached: false`, `status` `queued` or `running`. A double-click is therefore one measurement and one quota unit, not two. Clients get an `id` to poll either way, so no branch on this is needed; it is documented because the `id` may be one the caller has already seen. **A check that has not finished within the staleness window is reported `failed`** with `error_code: timed_out`, by the read path rather than by a writer. Nothing writes a terminal status when a workflow is cancelled or its task queue has no worker, so a client that waits for one would poll a `running` check for ever. **`429` when the organisation's daily allowance is spent**, with `Retry-After` and a body naming the limit. The allowance exists because one customer in a refresh loop would otherwise burn the whole estate's shared PageSpeed quota and take a free feature down for everyone. Cached results never count against it.

パラメータ

リクエスト本文

返信

このエンドポイントが返すエラー

401 · 403 · 404 · 422 · 429 · 503