@sentry/effect

Important Changes

• core: Warn on repeated Sentry.init() and unbind the client on close()

  Sentry.init() now warns when it runs while a client is still active. For now, the new client still replaces the active client, but the active client is not closed, so state from both can mix. Call Sentry.init() once, or call await Sentry.close() before you call it again. Sentry.close() now unbinds the client it closes, so after close(), getClient() returns undefined, isInitialized() returns false, and a later init() sets up a new client. In @sentry/cloudflare, later requests no longer reuse the isolate's cached client once it is closed or closing. On the server, @sentry/nextjs and @sentry/remix now return the active client from a repeated init() call, not undefined.

Other Changes

• bun: Keep server spans open until streaming responses finish

• langchain: Derive gen_ai.conversation.id from the invoke config

• nextjs: Add code.file.path to use cache fill spans

• react-router: Export createSentryServerInstrumentation from the Cloudflare entry

• remix: Parameterize Remix 3 navigation spans

• remix: Parameterize Remix 3 page load spans

• remix: Report the matched route on Remix 3 HTML responses

• ai: Normalize token usage and preserve cache breakdowns

• browser-utils: Handle NS_ERROR_NOT_INITIALIZED from setTimeout in Firefox

• bun: Add conversationIdIntegration to the default integrations

• cloudflare: Detect TypeSafe Jev calls in Workers AI by model ID

• cloudflare: Keep an installed OpenTelemetry async context strategy

• feedback: Correct overlapping screenshot annotations and drag dimming

• nextjs: Don't capture forbidden() and unauthorized() as errors

• nextjs: Include basePath in request url of Pages Router errors

• nextjs: Skip init inside a request of @sentry/cloudflare

• node: Flush telemetry before Vercel functions are suspended

• nuxt: Do not inject debug IDs into the SSR build

• react-router: Pass the project when creating a release

• remix: Pass the project when creating a release

• server-utils: Derive the Vercel AI conversation id from the OpenAI conversation option

• server-utils: Instrument the Vercel AI experimental_decide channel event

• server-utils: Map Flue messages and token usage to the gen_ai conventions

• server-utils: Record Vercel AI experimental_telemetry.metadata on gen_ai spans

• sveltekit: Export consoleLoggingIntegration from worker entry

• tanstackstart-react: Fix server route parametrization gaps

• deprecated

  core: Deprecate SEMANTIC_ATTRIBUTE_SENTRY_ORIGIN

Internal Changes

Work in this release was contributed by @minwookshin. Thank you for your contribution!