Patch Changes
-
#11753
b8f3309Thanks @ruifigueira! - Added the following improvements to local Browser Rendering binding in Miniflare:- Local Chrome version upgraded to 126.0.6478.182
- Reciprocate browser websocket close events
-
#12986
5aaaab2Thanks @petebacondarwin! - fix: allow mixedd1Databasesrecords containing both string and object entriesPreviously, passing a
d1Databasesconfig that mixed plain string values and object entries (e.g.{ MY_DB: "db-name", OTHER_DB: { id: "...", remoteProxyConnectionString: ... } }) would cause Miniflare to throw an error. Both forms are now accepted and normalised correctly. -
#12986
5aaaab2Thanks @petebacondarwin! - fix: allow mixedkvNamespacesrecords containing both string and object entriesPreviously, passing a
kvNamespacesconfig that mixed plain string values and object entries (e.g.{ MY_NS: "ns-name", OTHER_NS: { id: "...", remoteProxyConnectionString: ... } }) would cause Miniflare to throw an error. Both forms are now accepted and normalised correctly. -
#12987
f8516ddThanks @petebacondarwin! - fix: allow mixedpipelinesrecords containing both string and object entriesPreviously, passing a
pipelinesconfig that mixed plain string values and object entries (e.g.{ MY_PIPELINE: "pipeline-name", OTHER_PIPELINE: { pipeline: "...", remoteProxyConnectionString: ... } }) would cause Miniflare to throw an error. Both forms are now accepted and normalised correctly. -
#12952
9c9fe30Thanks @petebacondarwin! - fix: allow mixedr2Bucketsrecords containing both string and object entriesPreviously, passing an
r2Bucketsconfig that mixed plain string values and object entries (e.g.{ MY_BUCKET: "bucket-name", OTHER_BUCKET: { ... } }) would cause Miniflare to throw an error. Both forms are now accepted and normalised correctly. -
#13015
6a6449eThanks @petebacondarwin! - fix: disable undici Pool request timeouts for local devMiniflare's undici
Poolinstances were using the defaultheadersTimeoutandbodyTimeoutof 300 seconds (5 minutes). Any request taking longer than that — streaming responses, large uploads, long-polling, or compute-heavy Workers — would be silently killed with a "request failed" error.Setting both timeouts to
0disables them entirely, which is the correct behaviour for a local development tool where there is no reason to enforce request timeouts.