fix(site): only report Sentry errors from deployed function runtime (#1393)

Co-authored-by: Claude <noreply@anthropic.com>

site/sentry.server.config.ts

@@ -1,9 +1,21 @@
+import process from 'node:process';
+
 import * as Sentry from '@sentry/astro';
 
+// Only the deployed Netlify Function runtime sets AWS_LAMBDA_FUNCTION_NAME;
+// it's absent during `astro build`. Gating on it keeps errors from pre-render
+// (e.g. a broken PR) out of Sentry — we only want live request failures.
+const isLambdaRuntime = Boolean(process.env.AWS_LAMBDA_FUNCTION_NAME);
+
+// Alert for production (videojs.org) and branch-deploy (next.videojs.org) only.
+// Deploy-preview errors are the PR author's to triage before merge.
+const context = import.meta.env.CONTEXT;
+const isAlertingContext = context === 'production' || context === 'branch-deploy';
+
 Sentry.init({
   dsn: 'https://6bcdfa6b82da6dd4d7753618a9a69c7c@o43841.ingest.us.sentry.io/4510671167160320',
-  environment: import.meta.env.CONTEXT || 'development',
-  enabled: import.meta.env.PROD,
+  environment: context || 'development',
+  enabled: import.meta.env.PROD && isLambdaRuntime && isAlertingContext,
   release: import.meta.env.COMMIT_REF || import.meta.env.DEPLOY_ID || undefined,
   // Adds request headers and IP for users, for more info visit:
   // https://docs.sentry.io/platforms/javascript/guides/astro/configuration/options/#sendDefaultPii