Tags

v1.53.0

v1.52.1

v1.52.0

v1.50.0

v1.49.0

v1.48.0

v1.47.1

v1.47.0

Release notes

✨ New features

- Custom monitoring metrics
Visualize your own metrics in the new Custom tab in Deployment monitoring. Add up to four custom metrics; calculate the data points you want to visualize and upload them using the Deeploy python client or API.

- Flexible input and output schema mapping  
Leverage Deeploy's full monitoring capabilities for external and custom Docker models with request and response mapping. Define a JSON mapping in your metadata for models that do not adhere to Deeploy's standard.

🛠️ Improvements

- Created a new navbar and Workspace selector to easily navigate between Workspaces
- Improved error handling for connecting to an external object storage
- Detached the selection of nodes from selecting custom resources (Private cloud only)
- Improved validation for the `validUntil` property of tokens and personal key pairs
- Added state for missing metadata in the Deployment details
- Stricter validation on environment variable keys to prevent confusing Deployment errors
- Improved error handling for failed Deployment updates

🐛 Bug fixes

- Fixed an issue with upgrading Deployments to external or managed
- Fixed an issue where Deployments with a large amount of prediction logs couldn't be deleted

v1.46.0

v1.45.1

Release notes

🛠️ Improvements

- Updated the SHAP explainer to 0.46. Older SHAP kernel explainer objects are not supported by version 0.46. Use the new explainer framework selector to deploy a different version of kernel SHAP.
- Workspace owners and operators can now change the status of alerts
- Inverted the trend line color for alerts and errors on the team overview page
- The unit of measurement for alerts is now saved and displayed in the triggered alerts overview

🐛 Bug fixes

- Fixed an issue with upgrading a Deployment to an authenticated external Deployment
- Automatically close a popover if an option from the list is selected
- Fixed an issue where alerts were triggered when `skipLog` was used to inference
- Disabled the ability to upgrade archived Deployments
- Fixed an issue with upgrading a Deployment to an authenticated external Deployment

v1.45.0

v1.44.0

# 1.44.0

## Release notes

### ✨ New features

- **Deployment approval**
Ensure your Deployment quality by requesting approval from your Workspace members before deploying it. The Deployment version will be pending until it has been approved, after which you can deploy it.

- **Select compliance templates per Deployment**
You can now add and remove specific compliance documentation and checklist templates on a Deployment level.
Tailor your compliance towards what is important for a specific Deployment.

- **Upgrade Deployments**
Upgrade your Deployment to unlock more of Deeploy's functionalities! Go to your registration or external Deployment details, click upgrade, and follow the upgrade flow to upgrade your Deployment use case.

- **Filter monitoring graphs**
Filter your monitoring graphs to refine your Deployment monitoring graphs even better. Go to your Deployment's monitoring page and click **Filters** to add the desired filters to your graphs.

### 🛠️ Improvements

- Okta has been added as an option for OIDC Authentication (Private Cloud only)
- Introduced a status and status code filter for the request logs
- Added an endpoint type filter to the prediction logs
- Improved handling of a failed Deployment in the UI
- Added a loader to the drift monitoring graphs

### 🐛 Bug fixes

- Fixed incorrect "custom id" name in a filter of predictions.
- Fixed an issue when switching between Deployments sometimes showed the old Deployment details
- Fixed the notification pagination
- Fixed slow loading of external Deployments in the Deployment overview

v1.43.0

v1.42.0

v1.41.1

v1.41.0

Release notes

✨ Major new features

- Support for onboarding external Deployments
Connect your externally hosted model to Deeploy using an API endpoint. Experience the benefits of monitoring and alert services by using the Deeploy API to inference your model.

- Support for onboarding registration Deployments
Register any model. You can also create compliance documentation for your model, and optionally link a repository to use model cards, data cards, and metadata.

🛠️ Other changes

- Reworked the create and update Deployment endpoints. Sagemaker and Azure Machine Learning Deployments can now be created and updated using the same endpoint, meaning their old endpoints are deprecated. Check out the swagger documentation for the new API spec.

🐛 Bug fixes

- Fixed issue where non-admin users could incorrectly manage alerts and added tests to ensure correct authorization behavior.

v1.40.1

v1.40.0

Release notes

✨ Major new features

- Job schedules
Create automated request flows for your Deployments using the new job schedules! Job schedules create cron jobs on the cluster that request instances from your transformer (or model) to prediction or explain. To create a job schedule, go to the job schedules page and click on **create schedule**

- Automated drift detection
Continuously monitor your production input data distribution and compare to a baseline distribution. View and analyze trends on the **Monitoring** page of your Deployment and distributions through interactive graphs and set alerts to get notified when significant drift is detected. The monitoring functionality has been expanded to support Jensen-Shannon divergence as a metric.

- Deployment dashboard
Navigate to a Deployment to find the new Deployment dashboard. The Deployment dashboard shows the most important Deployment information in a single overview.


🛠️ Other changes
- Workspaces can now have multiple owners. Team admins are automatically owners of every Workspace, meaning they have the same rights in every Workspace. Regular users can still be made Workspace owner via the Workspace members page.
- Select which autoscaling type you want to apply to your Deployment to finetune your model reachability. Enterprise and SaaS Scale users can choose between CPU and Concurrency based scaling.


🐛 Bug fixes

- Fixed an issue where restoring SageMaker Deployments sometimes failed to update the proper Deployment status
- Fixed an issue where the container logs only displayed a part of all the logs
- Fixed an issue where the Deployments order would change after a while
- Fixed an issue where the unsaved changes weren't properly saved when going back from the Deployment summary
- Fixed an issue where the password visibility toggle in the webhook dialog didn't work
- Fixed an issue where updating a Deployment didn't reuse the environment variables from the previous version of the Deployment

v1.39.0

✨ Major new features

- Team overview landing page
Admin users of a team can access the team overview page to view statistics, alerts, events, and compliance for all active Deployments across all Workspaces in the team. After logging in, you'll land on the team overview page. You can also navigate to it by clicking your **Avatar** in the top right and then selecting Team overview.

- Webhooks
Automate your workflow by adding a webhook that gets called when an alert triggers. Go to **Integrations**, and underneath **Webhooks** click **Configure**. Add a webhook and connect it to an alert rule by selecting it on your **Deployment alerts** page.

- Data card
Enhance your deployment with a data card, providing specifics on the data used to train the model and explainer. Specify a `data-card.md` within your Git repository so that it will be viewable in the details page of your deployment. 

- Environment variable UI integration and improvements
Environment variables can now be managed in the UI and linked to your Deployment by selecting it in the create Deployment flow or on the **Deployment details** page.

- Test page rework
Test prediction and explain endpoints, or view the explanation visualization on the new **Deployment test** page.


🛠️ Other changes
- Add a new deployment status that shows if deployments remain healthy after a successful initial deployment. In the deployment overview detailed error messages are provided for the model, explainer and transformer
- Improve deployment events error logging, specifically for startup and memory issues
- Improved empty states for tables and pages
- Added BLEU text metric in prediction logs of Deployments with `problemType` "textGeneration" in the `metadata.json`
- Improved the prediction log layout, including the option to directly show the explanation visualization
- Prediction logs table update, removed explanation visualization from prediction logs overview and added explanation, actual and evaluation columns
- Updated repository details page
- Improved the usability of the Workspace integrations
- Increased the expiry time for user invitation links to 7 days
- Added new filters for the Deployment events
- Improved onboarding experience with an example Deployment and realistic example data
- Added support for transformer with standard explainers and integrated explainers
- Added the feature name to input validation alert messages


🐛 Bug fixes

- Fixed an issue with pagination for prediction and request logs
- Fixed an issue with Azure installations that use different app registrations for KMS and Blob
- Workspace operators can delete credentials again
- Team admin can no longer access the deployment create flow in workspaces that the admin is not part
- Fixed an issue where the email field would show an error after inviting a user on the team page
- Fixed an issue where an alert notification would display an incompatible bin state
- Fixed a bug that could result in using default versions instead of custom framework versions
- Fixes captum image visualization

v1.38.3

## Release notes
Fixed an issue with deployments and compliance templates for Azure installations