[2.10.0] - 2026-08-28

This release is the first to support Django 6.1, and no longer supports Django
6.0.

Remember to migrate, and this time there is a manual step to clean up some
migration cruft as well, see NOTES.

Big new features, in brief

• Heartbeat functionality
• Kiosk mode, see the new link in the filter box
• About-page, with a button that can check for a for new release

See the copious NOTES.

Removed

• Removed outdated migrations files. See NOTES!
  (#2022)
• Removed support for testing on Django 6.0, as that version is EOL-ed.
  (#2025)

Added

• Added warnings to be shown for destinations and notification profiles where
  the medium is not installed
  (#1453)

• Added INCIDENT_RESTART event type to allow source systems to reopen incidents
  (#1502)

• Add maintenance API endpoints
  (#1774)

• Implemented detailed Slack URL validator
  (#1919)

• Add docs for using the base apprise plugin directly
  (#1921)

• Added a kiosk mode to the main view, providing a minimal version of the
  incident list suitable for passive Argus usage.
  (#1938)

• Added new API endpoint to explicitly update the last_seen field for
  a source system. (#1939)

• Added a way to to store expected heartbeat frequency (as a Postgres interval,
  minimum 60 seconds) to a source system, and made the new field visible in the
  admin. (#1941)

• Added management command suitable to run as a cronjob that looks for dead
  incidents (as per heartbeat frequency and last seen heartbeat). Dead sources
  are reported as incidents, formerly dead sources get their respective
  incidents auto-closed.
  (#1954)

• Show last_seen and heartbeat_frequency in the source list and allow
  altering heartbeat_frequency.
  (#1962)

• The time zone is now configurable via the TIME_ZONE environment variable
  (default Europe/Oslo, unchanged). The TIME_ZONE variable already shipped
  in docker-compose.yml previously had no effect.
  (#1963)

• Added reference documentation for Argus's own incidents.
  (#1964)

• Added the boilerplate for API v3 for incidents, with an altered SourceSystem
  serializer showing the last_seen field as read only to serve as an example.
  (#2010)

• Added support for testing on Django 6.1.
  (#2025)

• Add an "about"-page (#2042)

• Added function to return a ticket identifier for an incident to the base
  ticket class (#2047)

• Added button to about page to trigger a check for new versions. Staff only!
  (#2052)

• A new endpoint dedicated to health checks was added, on "/.still-alive/".
  This is to make it easier to filter out in access logs.

• Added zabbix-argus-glue to documented list of known glue services

• Added a new log level: TRACE, for stuff even less important than DEBUG. In
  the process, moved logging helpers to a dedicated home in anticipation of
  supporting better logging in production.

  This new level is really not a good match for use in production, expect a lot
  of noise. See NOTES.

• The Docker image publishing workflow can now be triggered manually to
  (re)build any past release, and tags the newest release as latest.

Changed

• Changed API to allow for an incident to have multiple INCIDENT_END events
  (#1502)

• Represent tickets using a per-plugin identifier.
  (#1710)

• Made Apprise an optional dependency
  (#1920)

• Collected the validation functionality of the media plugins in the base
  class, removed the ability to version a media plugin, and changed the plugins
  that are shipped with Argus to use new validation functionality.

  This will break existing 3rd party media plugins. See NOTES.
  (#2023)

• Updated CONTRIBUTING file to be more naggy, we really want everyone to
  use the PR template and to write proper commit messages!
  (#2026)

• Changed versioncheck to use the PyPI upload time instead of Argus' visit time
  to register new versions.
  (#2051)

• Whether a destination medium is marked as installed or not is now visible and
  filterable in the admin.

Fixed

• Simplified the tags API (in version 3) so that the internal structure of tags
  is properly hidden. This will make altering the actual structure a smaller
  task. (#136)
• Slimmed down the production Docker image with a multi-stage build, so the
  build toolchain and the full source tree (including its git history) are no
  longer baked into the published image.
  (#923)
• Added check for whether notifications are globally enabled before sending
  email and SMS notifications
  (#1918)
• Fixed incident list not reacting to filtering changes when a saved filter is
  selected (#1940)
• Fixed planned maintenance filter selection freezing the drop-down
  (#1958)
• Added check on startup for if an Apprise-derived medium is activated without
  the Apprise package itself being installed.
  (#1965)
• Fixed searching by description being case sensitive
  (#1978)
• Fixed state- and ack-field being empty after selecting and deselecting
  filters (#2000)
• Fixed filter tag dropdown closing when unselecting last item
  (#2005)
• Sorted the destination dropdown by media when adding/updating a notification
  profile (#2017)
• Show the destination label set by the user in the dropdown when
  adding/updating a notification profile
  (#2018)
• Correct the input format of v3 of the destination API
  (#2029)
• Fixed incident table becoming unresponsive when a bulk action dialog box was
  open during a table refresh.
  (#2030)
• Fixed the incident list sorting configuration not being preserved after a
  refresh and after changing the page
  (#2036)
• Ensured that kiosk mode always resets to page 1 to avoid strange behaviour
  (#2044)
• Fixed connecting ongoing planned maintenance tasks with incidents that are
  being covered by them (#2049)
• Pinned django-psycopg-infinity to 0.1.0. Version 0.1.1 changed infinity
  timestamps loaded from timestamp with time zone columns from local time to
  UTC. Since infinity is represented as a wall-clock sentinel rather than
  a true instant, the two are not equal, so an incident's end_time no longer
  compared equal to the infinity constant after being reloaded from the
  database. Among other things this made notifications for open incidents
  render an explicit far-future timestamp instead of "Still open".