## Release notes

### New features

- Standard explainers: for supported libraries (XGBoost, Scikit learn, LightGBM) it is now possible to deploy an explainer without any additional configuration. Currently comes with support for TreeSHAP.
- Deployment tab navigation now has an indicator to display the number of deployments.

### Bug fixes

- Fixed a bug on Events pagination.
- Fixed a bug which could lead to existing compliance templates not showing properly in the UI.

✨ Major new features and improvements

Additional performance monitoring metrics

We improved our monitoring with additional monitoring metrics! On top of that, we created new ways to monitor your Deployment using our monitoring cards and bins to improve your experience with visualizing your Deployments' data.

MLFlow
Deploy your MLFlow models and explainers on Deeploy using our new MLFlow integration! You can make use of MLFlow's model stages or deploy specific model versions.

Deployment overview
The new Deployment overview offers a better insight in your Deployments. Check out the alerts triggered in the last 24 hours, or get an overview of your Responsible Deployment progress for all Deployments in your Workspace.

UI changes
The response time for requests is presented in seconds instead of milliseconds Removed the possibility to archive Repositories Added the possibility to redeploy Deployments without any changes

🐛 Bug fixes
- Fixed issues during installations on Google (GCP, GCS, GKE)
- The password requirements in the sign-up flow are always visible during input
- Fixed an issue where the wrong event name was visible when viewing logs from an event
Inference errors are again saved on your prediction log
- You can update your AWS credentials for SageMaker
- Fixed a bug where Deployments that used a deleted branch could not be updated

v1.31.1

Release notes

🐛 Bug fixes

Use correct annotations for blob storage
Use correct namespace for crds

v1.31.0

Release notes

✨ Major new features and improvements

Compliance insights
Track and document the Deployment's compliance with responsible AI standards using the compliance insights. Find the compliance insights in the create Deployment flow, or in the Compliance tab of your Deployment. Completing the compliance insights is fully optional.

Transformers
We added support to deploy tranformers! Select 'Custom Docker' as your tranformer type in the Deployment create flow, or update an existing Deployment with your transformer.

Added evaluatedOnly to the prediction log endpoint
Filter retrieved prediction logs on whether they are evaluated or not using the evaluatedOnly query parameter.


🐛 Bug fixes

Save a prediction log for all the given inputs when inferencing the model fails
Throw a 400 (bad request) exception when request bodies do not satisfy our data payload while interacting with the Deployment

# Release notes v1.30.0 20230809

**Breaking changes**
In the helm values the tag has to be set separately per image. e.g. `images.tag` is now `images.frontend.tag`, `images.backend.tag` and `images.kserveclient.tag`. 

## ✨ Major new features and improvements

### Alignment SaaS and Enterprise
We now have a shared codebase for SaaS and enterprise.

### Blob and Docker credentials can now be selected as part of the deployment step
Previously, a created credential ID had to be added as part of the reference.json
in the git repository. This limited using the same repository branch for multiple deployments.
Blob credentials can be selected with all model frameworks and Docker credentials with
the custom docker option. Existing deployments can be updated in the model details page.

### Improved metadata.json experience
The metadata.json in your repository is now the single source of truth for your Deployment's example in- and output, tensor shapes, and custom ID. Your existing Deployments that defined these values in their Deployment metadata through the UI remain untouched, but when updating your Deployment, the metadata.json is now used instead of these values.

### Improved model and explainer error handling
We added more descriptive error messages when interacting with deployed models and explainers,
which are now directly propagated from their respective containers.
In the interact tab the error message is shown in the response field and via the API 
it is part of the response body. They can also be viewed in the prediction logs.

### Deployment tokens and personal key pairs performance and usability
We drastically improved the performance of our Deployment tokens and personal key pairs, allowing for less overhead on interacting with our API.

### Redeployed failed Deployments
Our UI now allows for easily redeploying a Deployment that failed after creating it. Simply select the changes you want to make through the Deployment details and update the Deployment in order to retry it.

### Optional uri in the Docker reference.json
The uri in the reference.json of a custom Docker image is now optional. If no uri is provided, it defaults to /v1/models/model-name:predict, where the model-name is a variable based on your the name you give to your model.

### UI changes
Change the Repository status directly after adding the SSH to your repository Updated the links to our documentation Specific error when there are no members to select on the Workspace members page


## 🐛 Bug fixes

Fixed the default custom resources for an Explainer
Delete all the Docker and blob credentials on deleting a Workspace

Release 1.29.1

Breaking changes
Requires Kubernetes 1.23+ due to a dependency on the HorizontalAutoScaler Api V2.

Release notes - Deeploy - v.1.29.1

New Features

- We have changed the way our KServe Client works. The KServe Client used to loop, retrieving Deployments from the database. In the new implementation, the KServe Client is a FastAPI, meaning it's event-driven. This not only improves deployment time but also reduces the possibility of errors
- Our deployment flow now supports PDP and MACE explainers
- A new option has been added to our Helm chart that allows enabling or disabling both KServe and SageMaker as deployment backends. Disabling a deployment backend stops the specific resources from being deployed on your cluster
- We have reworked the Workspace selector in the navigation to accommodate more Workspaces
- A new pagination component has been implemented

Bug Fixes

- The Swagger documentation for the Usage microservice is now up and running
- The time it takes to load your repository in the create and update Deployment flow has been improved
- Using MinIO as an object storage now works as expected
- We have fixed an error message that was thrown while retrieving branches and commits for initially failed Deployments
- We have prevented duplicate Introspectors for the same Deployment, which could cause issues while inferring your Deployment
- Code snippets in every language for prediction evaluations have been fixed
- An issue with the explain endpoint for PyTorch models has been resolved
- We've improved the pod start-up order for license checks, which removes errors like 'Request entity too large' from occurring when a valid license is present
- The count in the pagination for prediction- and request logs has been removed to improve performance on loading the logs

Improvements

- We now throw a specific error when the model file name doesn't adhere to our contract
- It's now possible to specify the 'from' email address from which emails are sent in the Helm chart
- The styling of the toggle button in our UI has been improved

Fixed a bug in querying the Prediction Logs
Fixed retrieving the secret in the Usage microservice
Fixed a false optional CRD for KServe

- Improved Prediction Log retrieval time
- Fix re-ordering of Deployment events
- Animated in progress events icon
- Highlight failed log in Deployment events
- Improved pop up placement
- Improved Kubernetes restrictions
- Change the Workspace owner as admin
- MyDeeploy access via toolbar