Skip to content

What's New

Platform 2026.11.12

Platform · 2026.11.12 · 12 November 2026

New FeatureImprovementBug FixSecurityOn staging sites

On staging sites

A preview of what is coming to production sites within a few weeks. Learn more about the Insites release process

This release lets you stop a running background job, speeds up saving large pages, and makes Liquid filters safer and more consistent.

Stop a running background job

The new admin_background_job_stop GraphQL mutation stops a job created with the background tag within a few seconds, wherever it is running. Changes inside an unfinished transaction are rolled back, the job is not retried, and the stop is logged. You cannot stop deploys, data exports, data imports or other jobs the platform runs for you.

A background job running, then stopped part way through with the new mutation

First, find the running job. A job that has just started can take about ten seconds to appear in the list.

query {
  admin_background_jobs(per_page: 20, filter: { type: RUNNING }) {
    results { id source_name started_at }
  }
}

Then stop it with its id:

mutation {
  admin_background_job_stop(id: "YOUR_JOB_ID") {
    id
  }
}

Saving large pages is much faster

Creating or updating a page with several megabytes of content no longer takes seconds.

Database queries stop when a request times out

If a page or GraphQL request runs past the request timeout, the database query it started is now canceled. Before, it kept running in the background and slowed down later requests.

Liquid filters handle empty input the same way

extract_url_params, expand_url_template, humanize, translate, translate_escape, video_params, querify, useragent and url_to_qrcode_svg now return an empty value for blank input, instead of reporting an argument error. Many other filters now accept nil without failing.

Also in this release

  • The parse_json tag now treats an empty body as an empty hash, including a body made only of whitespace HTML entities. This matches the parse_json filter
  • Liquid filters that generate data, such as random_string, pad_left and format_number, now have upper limits on size, and gzip_decompress output is capped at 50 MB
  • array_add, array_prepend and hash_add_key now also catch a value that would end up containing itself, for example adding a hash that already contains the target. Before, rendering failed
  • is_gpg_valid and the GraphQL GpgKey validation give the same results as before, and validating many keys no longer slows them down
  • force_encoding, encode, advanced_format, gzip_decompress, pad_left and digest now give clearer error messages, and hash_dig returns nil when called with invalid arguments
  • A deploy now reports an error when processing its assets fails. Before, the CLI was not told and the failure went unreported
  • When an app page and a module page share the same route, removing the app page no longer removes the module page. Before, redeploying did not bring it back
  • After a data clean that includes the schema, asset URLs now change. Before, redeploying the same asset files could keep serving old cached content
  • After a data clean that includes the schema, the constants cache is now cleared. Before, context.constants could return old values until a new constant was added
  • Sync now returns an error when a file cannot be deleted because records still depend on it. Before, it reported success
  • Security improvements