Changed

• Uploads go to local storage until S3 is configured. Production and staging named :amazon whatever the settings said, so a site with no bucket could not serve an upload at all. Name AWS_BUCKET_NAME and it is S3 as before; leave it unset and files are stored on the server, under the volume a Kamal deploy already keeps. A site can now run before it has object storage.

What’s fixed

• A folder of Markdown with a grid field no longer rewrites itself on every sync. A grid row carries an id the control panel follows it by, a file carries none, and one was minted each time — so the page never looked unchanged, and every deploy wrote a revision, purged its cache and raised events for a page nobody had touched. The row ids already stored are now kept. The same applies to a replicator.

Upgrade

• If you run on S3 without setting AWS_BUCKET_NAME, uploads were going to a bucket named nibble-<environment> by default. That default no longer selects S3 at all: without the variable this release stores files on the server instead, and anything already in that bucket stops being found. Set AWS_BUCKET_NAME before upgrading and nothing changes. A site that has never used S3 needs to do nothing.