github cloudflare/workers-sdk wrangler@4.163.0

Minor Changes

  • #16917 d36b39d Thanks @jamesopstad! - Write Build Output to .cloudflare/output/v1/ and record the producer

    vite build and the experimental wrangler build --experimental-cf-build-output now write Build Output v1, with account settings nested under settings in the root config.json. They also record themselves as the producer, so readers can name the tool that wrote the output in error messages. Previews from Build Output read the account settings from their new location.

  • #16246 349fc4e Thanks @leostera! - Start local Workflows from Artifacts events

    Declare Artifacts events on exports.workflow({ name, events }) in cloudflare.config.ts to start the owning Workflow locally in cf, Vite, and Workers Vitest. Wrangler configurations continue to use triggers.events. Native config conversion derives those targets from the Workflow export, and cf migrate attaches existing event subscriptions to their matching Workflow exports.

    Local Artifacts namespaces publish the documented cf.artifacts.* messages for repository create, import, fork, and delete operations and token creation and revocation. Successful create, import, and fork operations also publish the initial token-created event. Delivery follows the committed mutation, is unordered, and does not fail the repository operation if a destination rejects it. Push, clone, and fetch events are not yet published locally. This feature does not provide local Artifacts-to-queue subscriptions.

Patch Changes

  • #16803 b219078 Thanks @petebacondarwin! - Run experimental cf projects with their installed Vite or Vite+ commands

    cf dev, cf build, and deployments that build your project can use Vite or Vite+ even when no framework command is detected. This removes the need for the internal cf-vite delegate and supports dependencies declared and hoisted at the root of a configured workspace. A workspace application using a custom-named Vite config should declare @cloudflare/vite-plugin in its own manifest, or add vite or vite-plus as a dependency or dev dependency to identify its tooling; this keeps shared root tools from selecting Vite for non-Vite siblings. Configs passed later via --config are evaluated by Vite when the command runs. Existing framework commands keep their priority. Consumers resolve bare executables using the detected nodeModulesPaths in order, adding .bin to each directory. Lookup stops at the validated workspace root; standalone applications use only their own node_modules/.bin.

    Project inspection can recognize an existing Cloudflare configuration before dependencies are installed. Running a command provides guidance when the required tools are missing or the plugin version does not support cf profiles (>=2.0.0-beta.v.13 <3). Vite evaluates the project configuration when the command runs, including dynamic configs and custom config files. Repair guidance uses the project’s detected package manager. Automatic setup asks users to convert CommonJS .cjs and .cts Vite configs to ESM imports and exports in .mjs or .mts before adding the ESM-only Cloudflare plugin. Native builds also work with the Windows shell. Pages migration safeguards, unsupported-framework setup checks, and Wrangler command selection retain their existing behavior.

  • #16944 db3c071 Thanks @danlapid! - Stop wrangler dev from missing config file changes made soon after a reload

    wrangler dev replaced its config file watcher on every reload without waiting for the new one to start, so a change saved in that gap was never applied. It now keeps one watcher while the set of watched files is unchanged, and reports a config as applied only once its files are being watched. A second change saved within 50 ms of the previous one is no longer dropped either.

  • #16920 81e56dd Thanks @benbuzbee! - Use the effective max_instances value when choosing the default container rollout steps

    When max_instances was omitted, Wrangler deployed 20 instances but chose rollout steps as if there were 0 instances, so the rollout used a single 100% step. Wrangler now uses [10, 100] for omitted max_instances, while still using a single 100% step when max_instances is 0 or 1.

  • #16946 e97c9f8 Thanks @dependabot! - Update DEFAULT_COMPAT_DATE to 2026-11-16

    Use the bundled workerd release's compatibility date by default when scaffolding new projects or when no compatibility date is configured.

  • #16946 e97c9f8 Thanks @dependabot! - Update dependencies of "miniflare", "wrangler"

    The following dependency versions have been updated:

    Dependency From To
    @cloudflare/workers-types ^5.20261113.1 ^5.20261116.1
    workerd 1.20261113.1 1.20261116.1
  • #16874 4335eac Thanks @petebacondarwin! - Avoid redundant OAuth refreshes after a sibling process updates credentials

    Check access-token expiry and select the refresh token from the same fresh credential snapshot. When a sibling has already saved a valid access token, reuse it without another exchange or write. This prevents an unnecessary rotation that can let an older response overwrite a newer refresh token.

  • #16917 d36b39d Thanks @jamesopstad! - Detect Cloudflare Build Output of any version before deploying

    wrangler deploy, wrangler versions upload and wrangler preview refuse to run in a project that has Cloudflare Build Output, because that output belongs to cf. Detection now looks for any .cloudflare/output/v<N>/ directory rather than only the root config of the current version, so output left by an older or newer cf is still caught.

  • Updated dependencies [e97c9f8, e97c9f8, 349fc4e, 0bf81ac]:

Don't miss a new workers-sdk release

NewReleases is sending notifications on new releases.