23.07 Release notes

Instabase 23.07 is a major release that introduces new features, enhancements, and bug fixes.

Subsequent patch releases typically contain bug fixes along with testing, optimizations, and other minor internal changes.

Release 23.07.42

This patch contains testing, optimizations, and other minor internal changes. User functionality is unchanged.

Release 23.07.41

This patch contains testing, optimizations, and other minor internal changes. User functionality is unchanged.

Release 23.07.40

This patch contains testing, optimizations, and other minor internal changes. User functionality is unchanged.

Release 23.07.39

• Service account names can now be upper- or lower-case.

Release 23.07.38

• Annotations weren’t carried over when importing an annotation set from the file system into Solution Builder, or when exporting an ML Studio dataset in the same environment.

Release 23.07.37

• In ML Studio, the field list failed to load in certain circumstances.

Release 23.07.36

This patch contains testing, optimizations, and other minor internal changes. User functionality is unchanged.

Release 23.07.35

This patch contains testing, optimizations, and other minor internal changes. User functionality is unchanged.

Release 23.07.34

This patch contains testing, optimizations, and other minor internal changes. User functionality is unchanged.

Release 23.07.33

• All v2 API endpoints now return a 401 status code for authorization errors and a 403 status code for license errors. Previously, a 200 status code was returned, with an error message in the response body, for v2 API errors. See API errors for more information, including v1 API error responses

Release 23.07.32

This patch contains testing, optimizations, and other minor internal changes. User functionality is unchanged.

Release 23.07.31

This patch contains internal changes and testing. User functionality is unchanged.

Release 23.07.30

• If multiple open files have errors in Flow Review, only one appeared in the file list.

Release 23.07.29

• In Flow Review, when you used the pop-out icon to open a document in a separate window, the first file in the file list was displayed, rather than the selected document.

• Table editor cells and validations for the field within the table editor did not update correctly when cells were edited to fix validations.

Release 23.07.28

This patch contains internal changes and testing. Functionality is unchanged.

Release 23.07.27

This patch contains internal changes and testing. Functionality is unchanged.

Release 23.07.26

This patch contains only internal changes and testing and does not change functionality for users.

Release 23.07.25

This patch contains internal changes and testing. Functionality is unchanged.

Release 23.07.24

This patch contains only internal changes and testing and does not change functionality for users.

Release 23.07.23

This patch contains only internal changes and testing and does not change functionality for users.

Release 23.07.22

This patch contains only internal changes and testing and does not change functionality for users.

Release 23.07.21

This patch contains only internal changes and testing and does not change functionality for users.

Release 23.07.20

• If one row remained in a list or table in the Table Editor view, the values could not be deleted.

• The model service did not allow the clean up of model artifacts from failed downloads, resulting in wasted storage space.

• The job service sometimes crashed under large loads.

• Scheduling a job in the Solution Dashboard appeared successfully, but the scheduled job didn’t show up in the dashboard or the Scheduler app.

• When you run a solution from the Solution Dashboard, you can now specify runtime configuration and email notifications for the job. Optional flow run settings are accessible in Advanced options when you run a solution.

• A new run sync endpoint ({URL-BASE}/api/v1/solution/run_sync) in the Marketplace and Solution API lets you run a solution in sync mode.

Release 23.07.19

• Digitization of rich text format (.rtf) files is now supported.

Release 23.07.18

• Checkpoint errors were not always correctly displayed in the Job Status section of the Flow Dashboard.

Release 23.07.17

• During digitization, Reader—and the process files step in Flow—now automatically sizes table columns to prevent truncation when converting CSV to PDF.

• The Ray model is being progressively deployed for our SaaS customers. By the week ending on October 23rd, all SaaS customers will have the Ray model deployed in their Development environment. By the week ending on October 30th, the Ray model will be deployed for all SaaS customers in both the UAT and Production environments.

• After pausing a stage during Deployment Manager upgrades, a new Cancel stage button now displays. Clicking Cancel stage lets you force cancel an in-progress stage, rather than waiting for the stage to complete before the upgrade pauses.

Release 23.07.16

• During Deployment Manager upgrades, a Skip Turn Down toggle lets you skip the turn down step during an upgrade. Skipping the turn down step is not recommended during major version upgrades.

Release 23.07.15

This patch contains only internal changes and testing and does not change functionality for users.

Release 23.07.14

This patch contains only internal changes and testing and does not change functionality for users.

Release 23.07.13

• A new run sync endpoint ({URL-BASE}/api/v1/solution/run_sync) in the Marketplace and Solution API lets you run a solution in sync mode.

• ML Studio models using Gaussian calibration models did not work after upgrading to 23.07.

• Trace was not working correctly in the Flow Dashboard.

Release 23.07.12

• Split classifier steps now function correctly when preceded by more than one process files step.

Release 23.07.11

This patch contains only internal changes and testing and does not change functionality for users.

Release 23.07.10

• When choosing files, the browser might crash if the file was too large to display a preview.

• If you had two Flow Review tabs open with the same output folder of results, you might see irregularities.

Release 23.07.9

• Automation metrics incorrectly assumed that if a validation rule was configured for a document, that the field must exist. This bug fix checks to see that the field also exists in that document.

Release 23.07.8

• Under certain circumstances, when copying or moving files to a destination in an encrypted drive, the file service sometimes crashed.

• Restricted file extensions are now case insensitive.

• Cell-level validations were not highlighted in Flow Review for extracted tables and extracted table lists.

Release 23.07.7

• File retention jobs were not correctly purging files after 60 days.

• Under certain circumstances, the model training and output set creation portions of the Solution Builder onboarding tutorial would break.

Release 23.07.6

This is the first generally available release of Instabase 23.07.

New features

Platform

• The Instabase platform has been upgraded from Python 3.7 to Python 3.9. This upgrade brings performance improvements and access to new libraries and features. Because of this upgrade, Instabase versions 23.07 and later require .ibformers v2.2.0 or later.

• Public Preview | RabbitMQ high availability (HA) improves reliability for Instabase deployments by reducing downtime and disruptions caused by restarts of the broker. HA uses a multi-replica setup with a lease-based active replica election process.

  RabbitMQ brokers are run using a new deployment object in a Kubernetes cluster with one replica, with the existing statefulset object set to 0 replicas. The lease-based active replica election uses the database as a backing store for lease and active replica information. This means that RabbitMQ pods need to connect to the database, and credentials have been added to the workload definition. Similarly, the network policies shipped with Instabase have been updated.

  RabbitMQ HA doesn’t add any infrastructure requirements, as the same Persistent Volume Claim (PVC) is used with the new deployment.

• Instabase now supports OpenSearch v2. OpenSearch v2 has a much smaller vulnerability footprint than OpenSearch v1.x. This change is backwards-compatible across backends. However, data indexed with OpenSearch v2 isn’t compatible with OpenSearch v1 or ElasticSearch. You can upgrade to OpenSearch v2, or you may continue using OpenSearch v1 or ElasticSearch 7.

• Compressed license_metrics data from the database can be archived into the file system. The license metrics data is archived as part of the license_maintainance job, a periodic cron job. To enable this functionality, set the environmental variable EnableLicensing to true.

• Password and email account requirements have been updated. The + (plus) tag in email addresses is no longer supported. Passwords must now contain at least:

  • 1 uppercase letter

  • 1 lowercase letter

  • 1 number

  • 1 special character

  • 8 characters

• Weaviate, a vector database, is added to the platform to support backend work on indexing and querying results when working with large language models (LLMs).

• You can now select service accounts instead of or in addition to users in user selection screens. You can add service accounts to spaces, subspaces, and groups, and assign ACL permissions on the site permission and app permission pages.

• Public Preview | This release introduces workload autoscaling, letting Instabase autoscale data services based on demand. When enabled, autoscaling optimizes service resources to maximize efficiency and performance for any workload at a given time. Workload autoscaling also removes the need to manually size services and presents cost saving opportunities.

  Autoscaling is performed with Kubernetes HorizontalPodAutoscalers (HPAs) based on CPU usage for conversion-service, ocr-msft-lite, ocr-msft-v3, and ocr-service. Benchmarks for a standard cluster with autoscaling enabled show ~1.4x-2x improvements in cluster throughput.

  Workload autoscaling is in public preview and is disabled by default. To learn more about workload autoscaling, see the feature documentation. For information on how to enable workload autoscaling, see the deployment guide section.

Administration

• Administrators with file storage or site admin permissions can now configure or disable global drives mounted across all subspaces on the site. You can also disable all Instabase Drives across the site mounted by default. You can find these options in the Admin app’s File Storage tab. (The File Storage tab is available only in SaaS deployments.)

• The Service Accounts tab in the Admin > Site Permissions page has been removed. To update service accounts permissions, use the Users tab instead.

Observability

• You can now more easily configure observability alerting and federation settings in Deployment Manager, following the introduction of ConfigMap templates. ConfigMap templates introduce a simpler, UI-based way to manage select ConfigMaps that were previously managed using complex patching. See the observability documentation for a list of editable parameters and the configuration management documentation for guidance on editing ConfigMap templates.

Solution Builder

• You can now export Solution Builder projects for use in another deployment, so you can continue developing a solution in a different environment, or demonstrate a solution in an environment where it wasn’t originally built. When you export a solution, you can specify which models to include, and you can exclude all data, including documents, annotation sets, and ground truth sets.

Deployed solutions

• A new Solution Dashboard provides management and metrics for deployed solutions. From the app, you can add solutions, run and schedule solutions, and view automation metrics related to validation and correction rates, processing volume, and overall automation value. The app is now generally available for all users. In addition to the GUI-based app, the Deployed Solutions and Automation Metrics APIs are available for managing and monitoring deployed solutions.

• The Deployed Solutions API now enables running solutions by name and version, rather than UUID only.

ML Studio

• Model training in ML Studio now runs on Ray, a distributed computing framework that offers scalability, advanced observability, and performance enhancements. For details, see Ray model training.

  Enhancements introduced with the switch to Ray include:

  • GPU allocation that lets you specify partial, single, or multiple GPUs for training jobs, optimizing resources based on the specific requirements of your training tasks. For details, see Distributed Multi-GPU training.

  • In ML Studio, more detailed error messages related to model training task failures, so you can quickly diagnose issues without the need for manual investigation. Additionally, model training tasks are now integrated into audit logs.

  • In ML Studio, queue size is reported when you start a new training job, so you know your position in the queue and can estimate training duration.

  • In Grafana, a new Ray Cluster dashboard that offers insights into model training tasks, so you can monitor and optimize resource allocation.

  • In the Training Job Count API, new details about the number of Ray model training jobs by status, and in the Model Training API, updates that facilitate running custom model training jobs.

  • Optional binary autoscaling automatically scales GPU nodes as needed based on model training requests, saving costs particularly in SaaS environments.

• You can now download aggregated metrics in JSON format for a given ML Studio job, including training, evaluation, incremental training, and pruning jobs. Aggregated metrics include dataset stats, f1 scores, confusion matrices, and more. To get aggregated metrics, on the Metrics tab for any ML Studio job, click Download, or use the Models API.

Flow

• You can now specify a minimum and maximum percentage of workers used for flow processing based on groups you specify. Allocating resources based on groups allows multiple business units, solutions, or use cases to operate with minimal interference or interruption. For details, see Resource management.

Flow Review

• Updates to the Flow Review user interface give reviewers more control over their display. Reviewers can now:

  • Pop out the single record view to a separate window and optionally drag it to a second monitor. Both windows are synced so edits in the single record view appear in the main window.

  • Resize the file list and field list to maximize screen space and avoid truncating data.

  • Filter, sort, and search fields in the field list to quickly identify or locate specific fields.

  • Use icons to identify at a glance a field’s type and whether it has provenance or edit history.

• You can now export Flow Review results as an annotation set in order to create a new project or iteratively train a previously trained model. For details, see Creating annotation sets based on human review.

Enhancements

Deployment Manager

• You can now configure SAML connections without needing to apply patches yourself. Using new Deployment Manager APIs, you can configure a SAML connection during or after installation. During installation, you can also instead use the new SAML configuration UI in Instabase Installer. For more information, see the authentication documentation or the new SAML-related Installer APIs.

• If Kubernetes doesn’t allow a configuration change to a running Kubernetes service, Deployment Manager can now detect this failure, delete the running service, and then automatically recreate the service with the updated configuration.

• The infra dashboard has several enhancements:

  • Two new tabs, Daemon Sets and HPAs, let you monitor usage and view information about your Kubernetes DaemonSets and HorizontalPodAutoscalers.

  • When viewing a list of pods in the infra dashboard, you can more easily identify which pods are in the process of terminating. Pods with a status of Terminating have a color background.

  • The Deployments and StatefulSets tabs now include graphs displaying total CPU usage and total memory usage across the namespace.

  • You can restart StatefulSets from the infra dashboard. From the StatefulSets tab, select a StatefulSet then click Restart on its details page to perform a zero-downtime restart.

  • When viewing the list of containers in a deployment’s pod, you can now also see any init containers.

• On the Configs tab, the new search tool lets you search by keyword through all configs or patches.

• You can now right-click Deployment Manager tabs to open the tab in a new browser tab or window.

• An update to the behavior of post-install actions improves the reliability of post-install action reporting during upgrades. If any post-install actions triggered by the upgrade fail, the upgrade itself also fails. Previously, post-install action errors could be ignored even if an underlying issue existed. <!–CINFRA-1987>

Licensing

• License usage details are now broken down by deployed solution in both License Manager (Admin > Licenses) and in the License Usage API. This breakdown gives you increased visibility into how licenses are apportioned across your deployment.

• Licenses now update automatically for multi-year contracts in which a new license is required each year.

Solution Builder

• You can now version refiner, script, validation, and flow modules in Solution Builder rather than creating a copy. Versioned modules are grouped in the user interface, better indicating relationships between module versions. New Solution Builder projects fully support module versioning. Due to a one-time change in directory structure, versioning modules in an existing Solution Builder project might trigger errors in downstream modules. To fix these errors, in Flow, re-link the versioned module, which updates module path to the new directory structure.

• Accuracy reports in Solution Builder provide more granular insight due to changes to how accuracy is calculated.

  • Classification accuracy is now reported by page, rather than by record, providing more useful insight into classification of multi-page files.

  • Extraction accuracy is now reported for records, in addition to fields. Previously, straight-through processing rates were measured at the field level only, which skewed higher.

ML Studio

• Pending and in-progress training jobs now appear in a separate Added to queue table, visually differentiating queued jobs from completed or cancelled jobs.

• If you have a mix of automatic and manual test / train split annotation sets, a warning is displayed indicating that the automatic test / train split doesn’t apply to the manual annotation sets.

• When you expand a document in ML Studio, a spinner icon is displayed while the page is still loading.

• In annotation sets, you can now filter for documents with no class assigned.

Flow Dashboard

• The summary page for each flow now display settings used to launch the flow.

• A new restart option in Flow Dashboard lets you start a new flow with the same settings as an existing flow, so you can quickly re-run flows that encounter errors.

Flow Review

• When correcting extracted data in Flow Review, you can now select the appropriate value directly in the record view for nested field types, including tables, lists, extracted tables, and extracted table lists. Previously, this method of correcting data was supported only for text fields. Using this method specifies provenance for the field, which can improve accuracy; however, you can’t clear provenance after it’s applied.

• The interface for editing table data is now more streamlined and usable. A new header offers buttons for common actions such as adding and deleting rows and columns, and undo and redo. The header also includes a list of cell validations which allow you to jump quickly to a specific issue. Additionally, cell validation messages are now displayed on hover, and table lists are now paginated.

• The Automation Metrics API now reports time-based review metrics, including how long a job spends in the review queue and review time per user on specific fields or classes.

Reader

• Reader output now indicates handwritten words within the style for each word along with confidence. This feature is only available for MSFT OCR engine.

Text Editor

• Debugging capability was removed to prevent running UDFs in Text Editor, which can pose security risks.

Metrics

• With this release, the Metrics app displays comprehensive statistics for model training tasks. You can view metrics such as the number of jobs executed within a specific time range, the specific user that triggered each model training job run, and advanced statistics such as grouping by username or function name.

Scheduler

• You can now schedule flows by specifying a corresponding solution name and version number, or a solution directory. Previously, Scheduler supported running flows only by specifying a flow binary.

Test Runner

• Public preview | Test Runner API V2 provides more robust and reliable execution of tests. Before you migrate to Test Runner API V2, test your current Test Runner API usage. See the Test Runner API V2 documentation for more information.

• To streamline comparing test results to ground truth sets, Test Runner now supports comparisons between file_comparator_easy and ibocr_comparator_easy. Specify the folder containing the files, and Test Runner selects the files from the specified folder at runtime.

• You can now run and test solutions using the Test Runner app by selecting the Run Solution option as the test method.

Bug fixes

Platform

• Granting a group pipeline permissions could fail but still return a successful response.

• After mounting a database, the database tables weren’t visible in the interface.

Solution Builder

• After copying a validations rule and performing a field swap, the validations card in Solution Builder displayed the same affected fields for both the original rule and the copied rule.

• Ground truth sets were limited to 20 sets. Now, ground truth sets are paginated to display all available sets.

• Uploading files with certain attachments produced errors and halted the upload process. Now, errors are logged the upload process continues with the remaining documents.

ML Studio

• A warning about an annotation set being modified in another tab appeared unexpectedly under certain circumstances.

• Class schemas could be imported only from extraction models, not classification models.

• Keyboard shortcuts for marking and unmarking a record as annotated weren’t usable in the classification annotation view.

• The expanded page view displayed a loading spinner instead of the thumbnail image.

• In incremental training, no warning was displayed when a required hyperparameter was removed from the hyperparameter form.

• Lines in model training logs displayed out of order under certain conditions.

Flow

• Models imported from Marketplace that included split classification failed to produce output for the apply classifier step.

• Nested field configurations weren’t allowed for extracted table lists.

Flow Review

• Direct selection of values in the document viewer didn’t work for fields with the type ANY.

• Invalid fields didn’t include an error message in certain circumstances.

• List field types weren’t showing in edit mode.

Deployment guide

• When upgrading from release 22.10 to 23.01 or later, if you have flows that have extraction and refiner steps, you might see that they start failing during refiner execution. To fix this:

  • If your flow is using a published model, you can use the ML Studio Utilities app to fix your published models. Go to the Migrate Published Models tab and click Migrate.

    • You need ML Studio Utilities version 2.0.4 to do this. If you don’t have this version, reach out to the Instabase team.

  • If your flow is using model projects (that is, unpublished models), you must fix them manually in the file system.

    1. In the file system, open the extraction module folder in your flow modules. This folder contains a JSON file that contains information about the model project. Go to the folder specified in the "model_fs_path" field.

    2. In the folder, open the package.json file, and check to see if the "result_type" field has the value "ner_result". If not, edit the file so that the "result_type" field has the value "ner_result" (that is, "result_type": "ner_result",). Your flow now works correctly.

• With the introduction of ConfigMap templates, during upgrades and installations you now must upload the default_patches.zip file included in the release’s installation.zip file. For previous releases, uploading default_patches.zip was optional.

  If you have previously made changes to a non-template version of a ConfigMap (such as config-prometheus-server or config-alertmanager) that now exists as a ConfigMap template, those changes are not automatically migrated upon your first installation or upgrade to an Instabase version that supports ConfigMap templates. After installing or upgrading to an Instabase version that supports ConfigMap templates, you must manually update the corresponding ConfigMap template with any previously defined custom values. The non-template version of a ConfigMap that now has a template version is no longer used or supported.

  See the observability documentation for a list of affected ConfigMaps and ConfigMap templates and the configuration management documentation for guidance on editing ConfigMap templates.

• Release 23.07 introduces workload autoscaling as a public preview feature that’s disabled by default. Workload autoscaling has several infrastructure requirements. Read the workload autoscaling feature documentation for more information.

  To enable workload autoscaling during your upgrade to 23.07:

  1. Before upgrading, edit the release’s control-plane.yml file to enable the ENABLE_AUTOSCALING and IS_CUSTOMER_HOSTED_AUTOSCALING environment variables. The required steps are:

     1. Unzip the installation.zip file for the new release.

     2. On the command line, navigate to and open the control-plane.yml file (installation/control-plane/control-plane.yml).

     3. Change the value of the ENABLE_AUTOSCALING environment variable to True.

     4. Change the value of the IS_CUSTOMER_HOSTED_AUTOSCALING environment variable to True.

     5. Save your changes. This updated control-plane.yml file is what you apply when updating Deployment Manager at the start of the upgrade or installation process.

  2. Add the base configurations that enable autoscaling to the release’s base_configs.zip file:

     1. Unzip the base_configs.zip file contained within the release’s installation folder.

     2. Locate the autoscaling folder in the installation folder (installation > additional_configs > autoscaling).

     3. Move the config files in the autoscaling folder to the now unzipped base_configs folder.

     4. Select all files in the base_configs folder and compress them, creating a new .zip file of base configs.

     5. Rename the file base_configs.zip. This updated .zip file is what you upload during the upgrade.

  3. (Optional) If your deployment uses custom resourcing sizing, create patches that define the minReplicas and maxReplicas values for each autoscaled service’s corresponding HPA service. For example, a patch targeting autoscaler-conversion-service sets the autoscaling range for conversion-service. The required steps are:

     1. Calculate the minReplicas values for all autoscaled services’ corresponding HPA services. See the workload autoscaling documentation for instructions.

     2. In the release’s installation file, locate the custom-hpa-patches folder (installation > optional_patches > custom-hpa-patches). This folder contains patches to configure minReplicas and maxReplicas values for each HPA service.

     3. Edit each HPA service patch to define the minReplicas and maxReplicas values. Use your calculated minReplicas values and define the maxReplicas values based on your preferred resourcing sizing.

     4. Add the edited HPA patches to the default_patches.zip file:

        1. Unzip the default_patches.zip file contained within the release’s installation folder.

        2. Add your edited HPA patches to the now unzipped default_patches folder.

        3. Select all files in the default_patches folder and compress them, creating a new .zip file of patches.

        4. Rename the file default_patches.zip. This updated .zip file is what you upload during the upgrade.

  4. During the upgrade, upload the updated base_configs.zip file and the (optionally updated) default_patches.zip file. The default_patches.zip and base_configs.zip files contain all patches and configurations required to configure autoscaling in your deployment.

  Info

  You must upload the default_patches.zip file during the upgrade even if you didn’t add custom patches to it.

  Note

  If you have previously decommissioned (set replicas to 0) conversion-service, ocr-msft-lite, ocr-msft-v3, or ocr-service, to maintain this decommissioned state you must do the following when enabling workload autoscaling:

  • When updating the base_configs.zip file to include the HPA config files found in the autoscaling folder, don’t include the decommissioned service’s corresponding HPA config file. For example, if you have decommissioned conversion-service, don’t include the autoscaler-conversion-service.yml file in the base_configs folder.
  • After the upgrade or installation is complete, you must set the decommissioned service’s replicas count back to 0. (When autoscaling is enabled the value is automatically set to 1.) Run the following kubectl command: kubectl scale --replicas=0 <name of decommissioned service> -n $IB_NS, where $IB_NS is your Instabase namespace.

Deprecations and removals

For a complete list of deprecations and removals within the past year, see the deprecations and removals page.

Classifier

Training in the Classifier app is deprecated and scheduled for removal in 24.01 or later. Use ML Studio to train classifiers instead. You can still use the Classifier app to write custom code classifiers that don’t require training, such as heuristic-based classifiers.

Additionally, a planned Python upgrade in 23.07 might result in breaking errors in classifiers trained with the Classifier app.

Deployed documentation

Instabase documentation for versions 23.04 and later is now public and available at instabase.com/docs. The static documentation shipped in Instabase deployment images is deprecated and planned for removal in version 23.10 or later. If you are running an earlier supported version of Instabase, the documentation in your deployment will continue to be updated when you patch your deployment. Availability of Instabase platform documentation is as follows:

• 23.10 and later: online only

• 23.04 and 23.07: online and in-product

• 23.01 and earlier: in-product only