cloudflare/workers-sdk miniflare@4.20260415.0

on GitHub

Patch Changes

• #13354 854d66c Thanks @courtney-sims! - Prepares asset-worker for a more gradual rollout by refactoring and separating out the invocation from the business logic. In the future, this will provide space for us to route requests to new versions of asset-worker based on their plan, but should make no functional difference today.

• #13450 6f63eaa Thanks @petebacondarwin! - Fix POST/PUT requests with non-2xx responses throwing "fetch failed"

  Previously, sending a POST or PUT request that received a non-2xx response (e.g. 401, 400, 403) would throw a TypeError: fetch failed error. This was caused by an undici bug where isTraversableNavigable() incorrectly returned true, causing the 401 credential-retry block to execute in Node.js and fail on stream-backed request bodies. This has been fixed upstream in undici v7.24.8, so we've bumped our dependency and removed the previous pnpm patch workaround.

• #13447 aef9825 Thanks @dependabot! - Update dependencies of "miniflare", "wrangler"

  The following dependency versions have been updated:

  Dependency	From	To
  workerd	1.20260410.1	1.20260413.1

• #13475 eaaa728 Thanks @dependabot! - Update dependencies of "miniflare", "wrangler"

  The following dependency versions have been updated:

  Dependency	From	To
  workerd	1.20260413.1	1.20260415.1

• #13472 58292f6 Thanks @roerohan! - Fix wrangler dev crash when using a Flagship binding with remote: true

  In remote mode, the flagship binding is backed by a generic proxy worker that only has a default export. The plugin was requesting a named entrypoint "FlagshipBinding" which doesn't exist on it, causing workerd to reject the binding at startup. The named entrypoint is now omitted in remote mode so workerd routes to the default export, which correctly proxies all RPC calls to the remote Flagship service.

• #13386 5e5bbc1 Thanks @mksglu! - Make startup network requests non-blocking on slow connections

  Wrangler makes network requests during startup (npm update check, request.cf data fetch) that previously blocked the CLI indefinitely on slow or degraded connections (airplane wifi, trains), causing 10+ second delays.

  • Update check: The banner now races the update check against a 100ms grace period. On a cache hit (most runs) the result resolves in <1ms via the I/O poll phase; on a cache miss the banner prints immediately without blocking. A 3s safety-net timeout caps the update-check library's auth-retry path.
  • request.cf fetch: The fetch to workers.cloudflare.com/cf.json now uses AbortSignal.timeout(3000), falling back to cached/default data on timeout.

• #13476 d5ff5a4 Thanks @roerohan! - Fix wrangler dev crash when using a Stream binding with remote: true

  In remote mode, the Stream binding is backed by a generic proxy worker that only has a default export. The plugin was requesting a named entrypoint "StreamBinding" which doesn't exist on it, causing workerd to reject the binding at startup. The named entrypoint is now omitted in remote mode so workerd routes to the default export, which correctly proxies all RPC calls to the remote Stream service.

• #13426 89c7829 Thanks @edmundhung! - Reject non-local /cdn-cgi/* requests in Miniflare

  Miniflare now validates Host and Origin on /cdn-cgi/* requests before request rewriting. Requests are still allowed for localhost, configured route hostnames, and the configured upstream hostname, but non-local hostnames can no longer reach internal development endpoints such as platform-proxy, handler routes, live reload, and the local explorer.