Resolwe

===================

Added
-----
- Add ``descriptor`` and ``descriptor_schema`` fields to the ``Relation`` model
- Setting ``KUBERNETES_DISPATCHER_CONFIG_LOCATION`` specifying the location of
the kubernetes config to load in the ``Kubernetes`` workload connector

Changed
-------
- Retry loading ``Kubernetes`` configuration couple of times before giving up
in the kubernetes workload connector
- Add model observers that notify clients about model changes via a websocket
connection


===================

===================

Fixed
-----
- Clear ``custom_messages`` array on auditlog reset call


===================

===================

Changed
-------
- **BACKWARD INCOMPATIBLE:** Always try to load kubernetes configuration via ``load_kube_config`` befofe
falling back to ``load_incluster_config``
- Rename ``docker-compose.yml`` to ``compose.yaml``
- Listener can always modify the data object that it is processing
- Add original objects to ``post_duplicate`` call
- Install ``asgiref`` version based on the version of the installed ``Python``
interpreter

Fixed
-----
- Fix a typo in ``listener`` permission handling when creating data model: a
check was performed on the wrong object type

Added
------
- Make requests and limits for the ``communication`` container configurable


===================

===================

Added
-----
- Support custom user model in Python processes


===================

===================

Added
-----
- Send custom signal ``post_duplicate`` when objects are duplicated since
regular Django signals are not sent
- Add ``auditlog`` application to log user actions


===================

===================

Added
-----
- Custom signal ``resolwe.flow.signals.before_processing`` is sent when data
object is ready for processing
- Support setting ``descriptor`` and ``DescriptorSchema`` on Data during
process runtime
- Support filtering Data, Entity and Collections based on permissions (view,
edit, owner, group, shared_with_me)
- Support filtering Data and Entity objects by relation id
- Create ``upload_config`` API endpoint that specifies upload connector type
and credentials client can use for optimized upload


Fixed
-----
- Do not return multiple version of the same process while checking for
permissions in Python processes
- Change misleading error message when importing file if the response with
status code indicating error was received from the upstream server


Changed
-------
- Use ``data_id`` instead of ``data.id`` when notifying dispatcher to avoid
potential database query inside async context


===================