0.33.0

What's Changed

• [feature] Enable the Grafana plugin in the new UI
• [feature] Show user privacy in the new UI
• [feature] Support expand/collapse all in the new UI artifact display
• [feature] New landing page in the new UI
• [feature] Add --no-cache option for Docker builds
• [improvement] Display cloned state icon for cloned runs in the new UI
• [improvement] Changes to backend data model in preparation for upcoming features
• [improvement] Improve run cancellations for local runs
• [improvement] Minor visual improvements to new DAG view UI
• [improvement] Add Llama 2 to fine-tuning example
• [bugfix] Make the recent DB upgrade script more robust
• [bugfix] Allow changing back to old UI from unauthenicated local executions
• [bugfix] Resolve an issue with node expand/collapses in new DAG view
• [bugfix] Miscellaneous bug fixes in the new UI
• [bugfix] Fix prompt display in fine-tuning example
• [bugfix] Avoid collisions with public Docker repos for Docker builds

Full Changelog: v0.32.0...v0.33.0

Compatibility

Sematic Servers on this version can support pip packages back to v0.30.0.

Corresponding Helm Chart Release

1.1.14

0.32.0

What's Changed

• [feature] Publish a new version of the Dashboard UI, which is currently in "Beta". You can
  switch to the new version by clicking on the pop-up banner, and between the two versions
  through your profile window pop-up.
• [improvement] Add constraints to the DB schema in order to improve validations. In case you
  get any errors during the upgrade, please contact us on
  Discord so that we can assist you.
• [improvement] Added documentation for custom user metrics¹
• [improvement] Improve error messaging for unschedulable pipeline runs
• [improvement] When running the CLI via Bazel, use the current directory as the working
  directory
• [improvement] Add Hugging Face model types, visualizations, and documentation
• [example] Add an example pipeline which fine tunes LLMs that summarize a text
• [deprecation] Ended backwards-compatibility support for Calculator, which had been renamed
  to Function in v0.30.0
• [deprecation] Ended backwards-compatibility support for pre-v0.27.0 log message sourcing and
  for an API response serialization
• [bugfix] Fix a bug where long running jobs' durations were not correctly rendered
• [bugfix] Fix a regression where the URL generated in Slack updates was incorrect
• [bugfix] Redact the DB URL from the migration logs
• [bugfix] Fix a bug where the Native Docker Build System did not support image URIs in quotes
• [bugfix] Add missing configuration to deployment documentation
• [bugfix] Fix clipping of metrics graphs in the Dashboard when new data points are received

Full Changelog: v0.31.2...v0.32.0

Compatibility

Sematic Servers on this version can support pip packages back to v0.30.0.

Upgrade Instructions

Please see these instructions on how to ensure you can use the new version correctly!

Corresponding Helm Chart Release

1.1.13

New Contributors

• @twitchax made their first contribution in #944

Thank you @twitchax!

---

1. This feature is for Sematic's "Enterprise Edition" only. Please reach out if
   you are interested in using Sematic EE. ↩

0.31.2

What's changed

• [improvement] Add ability to customize images for Ray workers
• [improvement] Add image pull secrets and pull policy to migration pod

Full Changelog v0.31.1...v0.31.2

Compatibility

Sematic Servers on this version can support pip packages back to v0.30.0.

0.31.1

What's changed

• [improvement] Add index on edges for source/dest run ids

Full Changelog v0.31.0...v0.31.1

Compatibility

Sematic Servers on this version can support pip packages back to v0.30.0.

0.31.0

What's changed

• [feature] Enable remote execution using pure-Docker, without bazel
• [feature] Support live-metrics during Sematic Function execution¹
• [feature] Add visualization for Prompt/Response pairs
• [example] Add Hacker News summarization example
• [improvement] Expose all Kubernetes classes in the base sematic module
• [improvement] Switch from a WSGI server & gevent to an ASGI server (uvicorn)
• [bugfix] Ensure UI-reruns don't automatically "rerun from here" for the root run
• [bugfix] Fix an issue that prevented rendering of matplotlib figures
• [bugfix] Remove a password that could be shown in cleaner logs
• [bugfix] Include missing information from a local storage error message

Full Changelog v0.30.0...v0.31.0

Thank you @kaushil24 for your contribution!

Compatibility

Sematic Servers on this version can support pip packages back to v0.30.0.

---

1. This feature is for Sematic's "Enterprise Edition" only. Please reach out if
   you are interested in using Sematic EE. ↩

0.30.0

What's changed

• [feature] Grafana dashboards tailored for Sematic installable via Helm
• [feature] User metrics SDK
• [feature] Add horizontal pod autoscaling and pod disruption budget support
• [feature] Health endpoint to display DB health in dashboard
• [feature] Add support for set types
• [feature] Add GitHub commit check support
• [improvement] Have resolver continue in in a new resolution if resolver restarts
• [improvement] Log request IDs in server
• [improvement] Early resolution failure is more robust
• [improvement] Make API retries more robust
• [improvement] Enable run search deep links
• [deprecation] Calculator renamed to Function
• [bugfix] Fixed a DAG view display issue
• [bugfix] Fixed application logs being duplicated
• [bugfix] Adjust the Name column width distribution
• [bugfix] Add missing call to init to Function

Full Changelog: v0.29.0...v0.30.0

Thank you @kaushil24 for your contribution!

Compatibility

Sematic Servers on this version can support pip packages back to v0.24.1.

Upgrade Instructions

Default Kubernetes deployments of the Sematic server will now run with 2 pods for the API server, in order to enable high availability. As such, the memory and CPU requests and limits for each pod has been halved.

0.29.0

What's changed

• [feature] Garbage collection CRON job
• [feature] Enable customization of local storage path
• [feature] Add Sematic Grafana dashboard as helm package
• [feature] Add support for function timeout
• [improvement] Enable backward logs scrolling
• [improvement] Rename inline=False to standalone=True
• [deprecation] Remove direct support for matplotlib figures, use Image instead
• [bugfix] Constraint plotly version for MNIST example
• [bugfix] Fix Kuberay autoscale

Full Changelog v0.28.1...v0.29.0

Compatibility

Sematic Servers on this version can support pip packages back to v0.24.1.

New contributors

Many thanks to @kaushil24 and @v-pwais.

0.28.1

What's Changed

• [improvement] Allow selecting S3 paths in UI
• [improvement] Backend logging improvements
• [bugfix] Ensure gevent import doesn't monkeypatch standard lib late when importing Sematic
• [bugfix] Resolve issue with S3 links in the UI for S3 "directories"
• [bugfix] Eliminate one situation that could lead to duplicated logs in the UI
• [bugfix] Fix a casting issue with floats

Full Changelog: v0.28.0...v0.28.1

Compatibility

Sematic Servers on this version can support pip packages back to v0.24.1.

0.28.0

What's Changed

• [feature] Display metrics for pipelines (success rate, run count, runtime)
• [feature] Added support for setting and memorizing a dev debug flag for the Dashboard
• [deprecation] Deprecate Kubernetes 1.22 support
• [improvement] Backend improvements to syncing with Kubernetes job states
• [improvement] Several minor logging improvements and fixes
• [bugfix] Enable local server to run with python 3.10
• [bugfix] Ensured different users can rerun a pipeline
• [bugfix] Ensured pipeline reruns use the submitting user's credentials
• [bugfix] Ensure canceled/terminated runs have proper runtime display
• [bugfix] Properly display duration for cloned runs
• [bugfix] Fix matplotlib figure serialization, use Sematic Image type for support
• [bugfix] Make storage object URL redirects consistent
• [bugfix] Remove possible infinite reconnect loop when canceling local runs
• [bugfix] Fixed a bug where the Resolver Socket.io client would not be cleanly closed
• [bugfix] Fix corner case in the comparison of sqlite versions
• [bugfix] Wrap long pipeline import paths in pipeline/run display

Full Changelog: v0.27.0...v0.28.0

Compatibility

Sematic Servers on this version can support pip packages back to v0.24.1.

0.27.0

What's Changed

• [feature] Added new S3Location and S3Bucket types that render S3 links in the Dashboard, and
  documented them
• [feature] Added a new Image type that render images in the Dashboard, and documented it
• [example] Added TorchLightning + Resnet + Ray example
• [example] Added Ray AIR example
• [example] Added various enhancements to the example Testing Pipeline
• [improvement] BREAKING CHANGE: image_layers field in sematic_pipeline bazel macro
  now ONLY gets passed to image layering, and not also to the Sematic binary target.
  if you are using image_layers to express dependencies of a pipeline, you will now
  need to duplicate them in the deps field.
• [improvement] Added support for Python 3.10
• [improvement] Added support for Python 3.10-style type hints
• [improvement] The Dashboard login flow will now redirect to the requested page instead of to
  the homepage
• [improvement] Displaying the user who started a Run in the Dashboard Run list page and in the
  Run history drop-down
• [improvement] Various documentation updates and fixes
• [improvement] Various log message improvements and enhancements
• [improvement] Sped up Docker image generation through various improvements
• [improvement] Automatically update npm dependencies when building the wheel
• [improvement] Added deep-linking selected panel stickiness in the Dashboard
• [improvement] Filtered only pipelines launched by the current user in the Dashboard homepage
  last run widget
• [improvement] Documented Ray integration architecture
• [improvement] Improved wheel building portability by defaulting to bash
• [improvement] Improved reaction time of the CloudResolver after a very long-running cloud Run
  has ended
• [improvement] Improved handling of incorrect cleanup of or errors in subprocess spawned by
  user code
• [improvement] Various internal refactorings, cleanups, and build improvements
• [improvement] Switched from MB and GB to GiB and MiB in the Ray configurations, for
  consistency and standardization reasons
• [improvement] Added checks and documentation for the new libmagic dependency
• [improvement] Resolution failures caused by Resolver errors now get marked as failed instead
  of canceled
• [bugfix] Pinned MNIST example pipeline dependencies after a new dependency release broke the
  execution
• [bugfix] Fixed a bug that prevented the browser back/forward buttons from working correctly
  in the Dashboard
• [bugfix] Fixed a bug that sometimes prevented the Dashboard Run tree panel from updating
• [bugfix] Fixed sorting of the Runs in the Dashboard Run tree panel by creation time
• [bugfix] Fixed a race condition where Kubernetes job updates would be incorrectly handled,
  and added safeguards for incorrect Run state transitions
• [bugfix] Fixed a bug where killing a subprocess spawned by user code would prematurely
  terminate the Run, or leave the Resolver in an inconsistent state
• [bugfix] Fixed a race condition that would prevent a fresh deployment on Kubernetes due to
  missing resources
• [bugfix] Fixed a bug in the Dashboard that prevented the log panel error messages from
  refreshing when seeking to a different Run
• [bugfix] Fixed a bug where failure to notify the Dashboard or Resolver with Run updates via
  socketio messages would cause the Resolution to fail

Full Changelog: v0.26.0...v0.27.0

Compatibility

Sematic Servers on this version can support pip packages back to v0.24.1.