V37 Release Notes

Prev Next

37.0.23 (28 Aug 2024)

Updates

This version includes a number of updates that optimize our internal testing and deployment processes.

37.0.22 (16 Aug 2024)

Connections

Fixed

CURL_CA_BUNDLE and ActiveMQ connections – We've fixed an issue that caused the ActiveMQ Message Queue Listener and Notifier Output Blocks to fail when the CURL_CA_BUNDLE ".env" file variable did not have a value.

Informative error messages from UiPath Notifier Output Blocks – We've resolved an exception-handling issue In UiPath authentication that made it more difficult to troubleshoot failures in the UiPath Notifier Output Block.

37.0.21 (17 Jul 2024)

Updates

This version includes a number of updates that optimize our internal testing and deployment processes.

37.0.20 (1 Jul 2024)

Releases 

Fixed

Creating releases with field customizations – We've fixed a query-related issue that caused delays and timeouts when creating releases with a large number of field customizations.

Flexible Extraction

Fixed

Transcribing fields in manually reclassified Structured documents – We've resolved an issue that prevented fields from appearing in Flexible Extraction tasks for Structured documents that had been manually reclassified.

Security

Fixed

Compiling Docker 7.1.0 – We've fixed an issue that prevented Docker 7.1.0, introduced in v37.0.19, from compiling.

37.0.19 (19 Jun 2024)

Models

Fixed

Using v37 Field Identification models in v39 instances – We've fixed an issue that prevented Field Identification models created in v37 from being used in v39 instances in some circumstances.

Security

Fixed

Addressing security vulnerabilities – To increase the functionality and security of your system, we've upgraded:

  • requests to 2.32.2,

  • docker to 7.1.0,

  • types-requests to 2.31.0.6, and

  • idna to 3.7.

37.0.18 (6 Jun 2024)

Training Data Management

Fixed

Showing anomalies in table annotations – We've fixed an issue that prevented detected anomalies in table annotations from being shown in the application in some situations.

37.0.14 (23 Feb 2024)

Layouts

Fixed

"Latest version is not live" message for Semi-structured layouts – We've fixed an issue that caused a "Latest version is not live" warning message to appear on the layout details page for Semi-structured layouts, even though the latest locked version of the layout was live.

Submissions

Updated

Support for EML files and their attachments – You can now extract data from EML files and their attachments. When an EML file is ingested, the system creates a PDF file from the email's body and processes each of the file's attachments as a separate document in the submission.  

Training

Fixed

Performance of Table Identification models and select Field Identification models – We've resolved an issue that caused the performance of Table Identification models and Multiple Occurrence and Generic Freeform Text models for Field Identification to decrease. As part of this update, a feature that stopped training when the system determined that additional training would not improve model performance has been disabled by default. 

​​Reporting

Fixed

Filtering the Transcription Sampled Errors report by user – We’ve fixed an issue in the Transcription Sampled Errors report that caused all errors for a specific field to be displayed, even though the report was filtered to show errors made by a specific user only. 

API

Fixed

Respecting X-Forwarded-Proto HTTP headers – We've resolved an issue that prevented the system from respecting the X-Forwarded-Proto HTTP header when generating absolute URLs (e.g., URLs for previous and next pages of results) in API responses. This issue caused http:// URLs to be returned instead of https:// URLs in instances that used HTTPS and local authentication.

37.0.13 (11 Jan 2024)

Submission Processing

Updated

Windows-1252 encoding and HTML_SUPPORTED_ENCODING_TYPES – You can now specify the HTML encoding types supported in your instance or alter their preferred order by using the HTML_SUPPORTED_ENCODING_TYPES “.env” file variable. This variable lists the order in which encoding types should be used when processing HTML files. Hyperscience supports UTF-8 and Windows-1252 encoding types, with HTML_SUPPORTED_ENCODING_TYPES having a default value of utf-8, windows-1252.

Reporting

Fixed

Calculating the time taken to complete Flexible Extraction and Custom Supervision tasks – We've fixed an issue that caused completion-time calculations for Flexible Extraction and Custom Supervision tasks to depend on the contents of the tasks (e.g., fields, tables, decisions).

OpenID Connect (OIDC)

New

Redirecting users during ID token renewal – By default, when renewing OIDC ID tokens, the application no longer redirects users to the identity provider’s token endpoint. To allow this step to be bypassed, we have introduced the HS_OIDC_RENEW_ID_TOKEN_WITH_REFRESH_TOKEN “.env” file variable. When this variable is set to true, the renewal transaction occurs without redirecting users out of the application, enhancing the overall user experience. See OpenID Connect (OIDC) for more details.

Fixed

Renewing ID tokens – We’ve fixed an issue with the HS_OIDC_RENEW_ID_TOKEN_EXPIRY_SECONDS “.env” file variable that prevented the OIDC ID token from being renewed. Now, the ID token is renewed as intended.

SaaS

Fixed

Reading large text files – We've resolved an encoding-detection issue that sometimes caused large text files to be read very slowly.

37.0.12 (16 Nov 2023)

Flows

Updated

Default timeout for blocks – We've increased the default timeout for block requests from 60 seconds to 180 seconds.

Training Data Management

Updated

Resource management – We’ve made some optimizations for uploading documents, as well as for saving and editing annotations on existing training documents. Memory usage should be lower during these tasks, and a small increase in speed can be observed, especially when working with larger documents.

Fixed 

Columns not present in current layout – We’ve fixed an issue where Table Identification model training failed due to the presence of annotations for columns that were not in the latest live layout version.

Task Queue 

Fixed

Effects of changing filters – We’ve fixed an issue where changing the filters in the Task Queue would only apply the changes and deselect any filters originally applied. This issue led to incorrect filtering when choosing both a date range and a filter from the Filters list. 

37.0.11 (27 Oct 2023)

Custom Supervision

Fixed

Case-level decisions for unassigned documents and pages – We've fixed an issue that allowed keyers to make case-level decisions for documents and pages that were not assigned to cases.

Reporting

Fixed

Accounting for differences between browser and server times – We've updated our task-completion-time calculations to account for the difference between the server's timestamp and the browser's timestamp.

Input Connections

Updated

Number of subfolders scanned by Email Listener for Microsoft 365 Outlook – We've updated the number of subfolders scanned by the Email Listener in Microsoft 365 Outlook accounts from 10 to 100.

37.0.10 (25 Oct 2023)

Layouts

Fixed

Filtering layouts by tags – We’ve fixed an issue related to a query-filter limitation on the Layouts page that occurred when users filtered the list of layouts by layout tag. Users can now see all matching results when they filter by a tag. 

Submissions

Fixed

Responsiveness when viewing documents – We’ve fixed a query-plan  issue in deployments using MSSQL databases that caused delays when opening the document viewer in some instances.

Training Data Management

Fixed

“Remove Training Documents” button and empty lists – We’ve fixed an issue in Training Data Management where users could click the Remove Training Documents button for an empty selection of items. 

Showing groups with excess documents – We’ve resolved an issue that prevented the Groups sidebar from being opened when users clicked Show more groups

Flows

Fixed

Recycling workflow engine workers – To prevent potential memory leaks, we now recycle each workflow engine worker after it processes 10,000 flows.

Input Connections

New

Kafka Listener – When you add a Kafka Listener connection to a flow's Input Block, you can ingest, send, and schedule submissions from a Kafka queue for processing in Hyperscience. Kafta messages must be formatted in JSON in the same structure as Submission Creation API requests. You can configure the block's output to be sent to a Kafka topic of your choosing via an HTTP connection.

To allow you to configure your Kafka consumer options to meet your needs, we've introduced several ".env" file variables. You can also scale the number of consumers in your instance.

For more information on the Kafka Listener and its configuration options, see Kafka Listener.

SaaS

Fixed

Logging in when self-service user management was enabled – We've fixed an issue that prevented users who weren't System Admins from logging in to Hyperscience when self-service user management was enabled.

Permission group after deactivation and reactivation – We've resolved an issue where, if a user's permission group was changed and then the user was deactivated, the user would be assigned to their original permission group upon reactivation. This issue impacted deployments where self-service user management was enabled. 

Security

Fixed

Addressing security vulnerabilities – To ensure security, we've updated:

  • urllib3 to 1.26.17,

  • mpmath to 1.3.0, and

  • scipy to 1.10.0.

37.0.9 (10 Oct 2023)

Languages

Updated

Transcriptions of Korean multiline fields – We’ve made optimizations to transcriptions of multilined Korean text. Now, the model achieves higher accuracy when reading Semi-structured text fields with multiple lines.  

Layouts

Fixed

Sorting layouts by "Last Updated" – We've fixed an issue that caused the table on the Layouts page (Library > Layouts) to become empty when it was sorted by the contents of the Last Updated column. 

Flows

Updated

Default timeout for blocks – We've increased the default timeout for block requests from 20 seconds to 60 seconds. 

Transcription Supervision

Fixed

ResizeObserver loop exceeded error in Chrome on Macs – We've fixed an issue that caused ResizeObserver loop exceeded errors to occur during Transcription Supervision in Mac Chrome browsers in some instances. 

Table Identification

Fixed

Completing submissions containing nested tables with orphan rows – We’ve fixed an issue that allowed submissions with a “Confirm that all pages are reviewed” warning message to be completed with orphan rows, leading to those submissions halting.  

SaaS

Fixed

Saving changes to user information – We've fixed an issue that prevented changes made through self-service user management from being saved in Okta. 

Security

Fixed 

Addressing security vulnerabilities – To ensure security, we've updated pygments to 2.15.0 and sentry-sdk to 1.31.0.

37.0.8 (13 Sept 2023)

Data Types

New

Capitalized Names – We've added a Capitalized Names data type that expects names that have the first letter of each name (e.g., first name and last name) capitalized.

For more details on Capitalized Names and other default data types, see Supported Characters and Default Data Types.

Input Connections

Fixed

Generating images from Email Listener – We've fixed an image-generation issue that caused data to be missing or incorrectly formatted in submissions created by the Email Listener.

Flows

Fixed

Timeouts for flow-block polling requests – We've added timeouts for flow blocks' polling requests, preventing failed requests from continuing indefinitely.

Security

Fixed

Updating cryptography – To address security vulnerabilities, we've updated cryptography to 41.0.3.

37.0.7 (30 Aug 2023)

Tasks

Fixed

Tasks after "Mark Layout Variation Incorrect" is clicked – We've fixed a race condition that sometimes prevented the Task Queue from loading after a keyer clicked a document's "Mark Layout Variation Incorrect" button during Field Identification.

Releases

Fixed

Loading of Releases page – We've optimized the loading of the Releases page (Library > Releases), resolving an issue that prevented the page from being displayed in some instances.

Reporting

Updated

Definition of dt_started – We've changed dt_started from the time the task was first assigned to the time when the task was opened. This update creates a more accurate measurement of the time taken to complete tasks.

Output Connections

Fixed

"transformed" API version in HTTP Notifier settings – We've fixed an issue that caused the following warning message to appear when the API version was set to transformed in the HTTP Notifier Output Block settings: Update deprecated output payload API version: transformed.

Default authentication method for UiPath Notifier – We've fixed an issue that caused OAuth to be the default authentication method for UiPath Notifier connections. The issue caused flows that used Basic Authentication for these connections to fail.

37.0.6 (17 Aug 2023)

Data Eligibility Filtering 

Fixed

Checking for consecutive page breaks – We’ve fixed an issue with training data eligibility where the "Consecutive Page Breaks" check was conducted for all model types and not only the applicable ones.

Flows

Fixed

Logs for blocks – We've resolved an issue that could potentially cause PII data from block inputs to be stored in task logs.

Deleting or canceling submissions with tasks in progress – The system now waits for all tasks related to submission to be completed before allowing that submission to be deleted or canceled. Because PII data could be stored in the database temporarily during task execution, this update prevents a submission's PII data from being stored in the database after the submission is deleted or canceled.

Subprocesses from pagination – Previously, pagination tasks sometimes created subprocesses that wouldn't time out if they couldn't be completed. To resolve this issue, we've added timeouts to these subprocesses.

UUIDs of block definitions – We've fixed an issue that caused block definitions to have duplicated UUIDs in some circumstances. This issue resulted in errors when initiating the application.

Connections

Fixed

Logs for HTTP REST Blocks – We've resolved an issue that caused authentication information to be included in logs for HTTP REST Blocks in some circumstances.

Security

Fixed

Addressing security vulnerabilities – To ensure security, we've updated:

  • Django to 3.2.20,

  • certifi to 2023.7.22, and

  • pyJWT to 2.7.0.

37.0.5 (28 Jul 2023)

Training Data Management

Updated

Appearance of “Analyze data” button – We’ve changed the color of the Analyze data button to be more visible when the requirements for data analysis are not met.  

AnalyzeDataButtonV37.0.5.png

Design improvements for the Table ID Anomaly Detection – We’ve changed the icon of the “missing column” label to a checkmark for consistency.

MissingColumnsLabel.png

We’ve also added a dotted line around labels that indicate column-level anomalies. 

ColumnLevelAnomalies.png

Fixed

Deleting training documents that are being processed – We've fixed an issue that caused errors to occur when training documents that were being processed were deleted.

Transcription Supervision

Fixed

ResizeObserver loop exceeded error in Chrome on Macs – We've fixed an issue that caused ResizeObserver loop exceeded errors to occur during Transcription Supervision in Mac Chrome browsers in some instances.

37.0.4 (18 Jul 2023)

Document Classification

Updated

Displaying Submission ID – We’ve added the Submission ID to the top of the page for easier traceability of your uploads (“Document Classification: Submission ”).

File Storage

Updated

Custom domain for Azure Blob Storage – If you've set up a custom, private domain for your Azure file store, you can add it to your Hyperscience configuration by including the IMAGE_STORE_AZURE_CUSTOM_DOMAIN variable to your ".env" file.

To learn more, see Azure Blob Storage.

Output Connections

Updated

"Additional notification request headers" setting for HTTP Notifier Output Blocks – We've added an Additional notification request headers setting. This setting allows you to enter HTTP headers for notification requests as key-value pairs in JSON format.

Fixed

HTTP Notifier Output Block requests and responses – We've fixed an issue that prevented HTTP Notification connections from sending POST requests. The issue also caused connections to return 401 errors in some circumstances.

OpenTelemetry

Updated

Default data-emission interval – We've updated the default data-emission for OpenTelemetry from 60 seconds to 30 seconds.

37.0.3 (14 Jul 2023)

RHEL

Fixed

lspci: command not found – We've fixed an issue that caused lspci: command not found errors to occur when ./run.sh was run on RHEL 7.9.

unable to load image when deploying trainer – We've resolved an issue that caused an unable to load image error to occur when deploying a trainer on instances running RHEL 8.4.

37.0.2 (12 Jul 2023)

User Experience 

Updated

Maximum number of files per upload – We've increased the default maximum number of files per upload from 100 to 1000. This value can be customized with the DATA_UPLOAD_MAX_NUMBER_FILES ".env" file variable. The maximum applies to both training-data pages for models and submission pages.

Fixed

Tooltips overlapping with header bar and sidebars – We’ve fixed the tooltips in the interface to be displayed over sidebars and header bars rather than under them when they overlapped. 

TooltipsOverlapping37.0.2.png

Languages

Fixed 

Text segmentation for non-Latin languages – We’ve fixed the language parameters for segmentation for languages outside of the Latin language family. We’ll use the language of the submissions during segmentation rather than linking them to the Latin language family by default. For example, we won’t use Latin segmentation if you upload Korean or Korean-English documents.

Layout Editor

Fixed

Keyboard shortcuts – We’ve fixed several minor issues with the shortcuts in the Structured Layout Editor that prevented them from working as expected:

  • Pressing CTRL + Y redoes the last change.

  • Double-clicking on a bounding box and pressing ALT or Option + Backspace or Delete causes a deletion-verification message to appear. 

Layouts

Fixed

Changing names of Semi-structured layout variations – We’ve fixed an issue where users had to refresh the page to see changes they made to names of Semi-structured layout variations. 

Creating variations with the same names as deleted variations – We’ve resolved an issue that caused a warning message to appear when creating a new layout variation with the same name as a deleted one. 

Find Potential Layout Variations

Fixed

Submissions with at least five empty or non-matching pages – We’ve fixed an issue where an error message appeared when trying to use Find Potential Layout Variations on submissions containing at least five empty or non-matching pages. 

Training Data Management

Updated

Training status of submissions that have completed QA tasks – We’ve implemented changes to ensure that the submissions processed through Quality Assurance have a training status of “Never” in Training Data Management and in the Training Data card on the model details page.

Titles of Field Identification documents – For consistency and clarity, we’ve updated the title in the document viewer for Field ID Training Data Management from “Document [ID]” to “Training Document [ID].” 

TrainingDocumentTitle37.0.2.png

Releases

Fixed

Exporting locked releases We've resolved an issue that prevented users from exporting locked releases. Doing so resulted in a Could not export archived layout release with UUID: error message.

Layout variations contained in exported releases We’ve fixed an issue where ZIP files for exported releases contained layout variations that were not part of the releases. Now these files contain only the layout variations associated with a particular release. 

Machine Classification

Fixed

Storing pre-calculations for classifying Structured documents – We've resolved an issue that caused invalid memory alloc request errors when the system attempted to store pre-calculated values for the release's Structured layout variations in the database. The issue affected instances with PostgreSQL databases. 

Classification Supervision

Fixed

User interface for Classification Supervision tasks We've made the following fixes to the Classification Supervision user interface:

  • We've widened the right-hand panel, enlarging the image of the page being categorized.

  • We've fixed an issue that caused the screen to flicker each time a keyer clicked on a thumbnail in the left-hand panel.

  • We've resolved an issue that caused the right-hand panel to be hidden when a keyer clicked on a page group in the middle panel.

Perform Tasks links for documents in submissions undergoing Manual Classification We've fixed an issue that caused Perform Tasks links to appear for documents whose submissions were in Manual Classification. Because Classification occurs at the submission level, Classification tasks for single documents don't exist. When a user clicked on a Perform Tasks link for a document whose submission was in Manual Classification, they would be taken to a "Loading" page that would load indefinitely.

​​Table Identification

Fixed

Keyboard shortcut for deselecting all columns We’ve fixed an issue that prevented all columns from being deselected when keyers pressed the ESC button.

Transcription

Fixed

Submissions with empty transcriptions We’ve resolved an issue that caused submissions containing fields with empty, or null, transcriptions to halt.

Manual Transcription

Fixed

Normalization of Date table columns with column-specific languages – We've resolved an issue related to normalization after Manual Transcription. Date table columns with a different language from the one assigned to the layout are now normalized correctly. For example:

  • Before: MM/DD/YYYY was normalized as YYYY/MM/DD.

  • After: MM/DD/YYYY is normalized as MM/DD/YYYY.

Reporting

Fixed

Reporting time spent on Classification Supervision tasks We've fixed an issue that prevented time spent on Classification Supervision tasks from being reported in Document Classification Supervision Time Spent (Seconds). This metric appears in the KeyerPerformance.csv file in the Keyer Projection Report and previously had a value of 0.

Databases

Fixed

Notifications and deadlocks – We've resolved an issue that caused database deadlocks to occur if the user and system made changes to notifications at the same time.

API

Updated

Increasing maximum submission payload size – We've increased the maximum payload for submissions from 10MB to 50MB.

37.0.1 (22 Jun 2023)

Training Data Management 

Fixed

Showing loading documents in “Training Data Health” card – We’ve fixed an issue where uploaded documents that were still loading weren’t accounted for in the numbers shown in the Training Data Health card. To clarify the information, we’ve updated the messaging to “0 documents added (documents may still be loading), 0 documents removed since the last analysis on [date]. Reanalyze data to update results.”

Manual Transcription

Fixed

Normalization of Date fields with field-specific languages – We've resolved an issue related to normalization after Manual Transcription. Date fields with a different language from the one assigned to the layout are now normalized correctly. For example:

  • Before: MM/DD/YYYY was normalized as YYYY/MM/DD.

  • After: MM/DD/YYYY is normalized as MM/DD/YYYY.

Upgrades

Fixed 

Defined batch sizes for database migrations – We've resolved a memory-allocation issue that occurred during database migrations when upgrading to v37. The issue caused out-of-memory or Invalid memory alloc request size errors in some instances.

37.0.0 (20 Jun 2023)

Languages

Updated 

Improvements to processing of Korean documents – We've made the following enhancements to increase automation for Korean documents:

  • Dynamic optimization of text-segment size for Korean documents – The system now optimizes text-segment length in Korean documents based on the aspect ratio of the documents' pages. This update ensures that text segments are predicted and transcribed by the machine in their entirety.

  • Transcriptions of multi-line fields – We've improved machine transcriptions of multi-line fields that contain Korean characters. Previously, the machine would sometimes transcribe characters as having an extra horizontal line, causing incorrect characters to be included in transcriptions.

  • Detection of 1s without surrounding text – We’ve made text-segmentation improvements that enable the system to better detect the number 1 when it does not have surrounding text.

Layouts

Updated

"Not in " option for table columns – The Not in option that has been available for fields can now also be applied to table columns in Semi-structured layouts. This option allows you to assign languages on a per-column basis, giving keyers the ability to enter transcriptions that are not in the language assigned to the document's layout. It also tells the system to transcribe that column’s text without applying the rules of the layout’s language.

To learn how to use the Not in option, see Creating Semi-structured Layouts.

Training

Updated

Training performance enhancements – We've optimized the training of Field Identification, Table Identification, and Unstructured models, reducing training time by 20-30%. The trainer ends the training process as soon as it determines that additional training would not improve model performance.

Training Data Management

New 

Training Data Curator – The Training Data Curator for Field Identification and Table Identification models suggests training documents that you should annotate first in order to reach your desired target accuracy much faster than before. When you run data analysis for a model, the Training Data Curator marks documents as having either High or Low Importance. You should give higher priority to the High Importance documents in your annotation efforts.

By identifying high-impact documents, this feature can potentially reduce the number of documents required to train a model from 400 documents to 100-200 documents, depending on the complexity of the task and the variation within the training data. In this way, the Training Data Curator allows you to implement new use cases—and maximize the benefits of Hyperscience—more quickly.

For more information, see Training Data Curator.

Updated

Training Data Management (formerly Keyer Data Management) improvements – We've made the following enhancements to Training Data Management in v37:

  • Anomaly Detection for Table Identification models – Running data analysis for a Table Identification model now identifies potential anomalies in the annotations for the model's training data. Reviewing these anomalies and making any necessary fixes increases model performance. More information about Anomaly Detection can be found in Detecting and Correcting Anomalies in Table Annotations.

  • Model Validation Tasks (MVTs) in Training Data Management – Because MVTs and Anomaly Detection both flag potential errors in training data, we've removed MVTs from the application in v37. Any anomalies found in training data are shown in Anomaly Detection results on the model details page.

  • Document Eligibility Filtering – Not all documents can be used as training documents. In previous versions of Hyperscience, there was no explanation in the application for why certain documents could not be used to train models. In v37, clicking See details >> next to the count of documents that are eligible for training reveals a list of reasons why certain documents are ineligible for training. For each reason, a count of affected documents or pages appears, along with a link to a list of those pages or documents where possible. For more information on Document Eligibility Filtering, see Document Eligibility Filtering.

Flows

Updated

Flow Groups – To make the connections between flows and subflows more visible in the application, we've introduced the concept of flow groups in v37. A flow group consists of a flow and its subflows. Each flow group in your instance appears as a card above the Flows table on the Flows page.

With the introduction of flow groups, we've made the following updates to the management of subflows:

  • If a flow is a subflow of other flows, its deployment status—whether it's live or disabled—is automatically managed based on the status of flows that depend upon it. That is, you cannot make a subflow live or disabled directly; but, a subflow becomes live when you make at least one flow that depends on it live.

  • Flows that are subflows of other flows cannot be archived. To archive a subflow, first remove it from the flows that use it as a subflow.

To learn more about managing grouped flows, see Managing Flows.

Flow Run (formerly Flow Execution) improvements – We've made the following enhancements to the Flow Run page:

  • Auto-refresh option for Processing submissions – If you are viewing a flow run for a submission with a Processing status, the Flow Run page automatically refreshes every second to show the submission's progress across the flow. You can toggle the auto-refresh option on or off by clicking Auto refresh is on or Auto refresh is off at the top of the page and toggling the switch that appears.

  • “View List” link for Halted submissions with subflows – When a submission has a Halted status, at least one of the flows initiated for that submission has failed, timed out, or been terminated. To view a list of the failed flows related to the submission, click the View List link below the submission's status in the left-hand sidebar. Doing so redirects you to the Flow Run list page, which is automatically filtered to show the submission's flow runs that have caused the submission to halt. You can then click on a flow's run ID to go to the Flow Run page for that flow.

  • “View Flow Run” link for nested flows and blocks that contain flows – Some blocks in a flow-run diagram represent smaller flows that contain the tasks performed by that block. To view the flow run for one of these blocks, click its View Flow Run link. This feature can help you debug blocks that have caused a submission to halt.

Submission SLAs

Updated

Defining SLAs for submissions received on business days or non-business days – With the updates in v37, you can define a submission's SLA based on whether it was received on a business day or a non-business day. These tailored SLAs help to set more accurate expectations for keyers processing submissions and stakeholders awaiting completed submissions.

For example, you can define the following set of SLAs:

  • A submission received on a business day after 9:00AM but before 5:00PM is due within 3 hours of when it is received.

  • A submission received on a business day after 5:00PM is due by 12:00PM on the next business day.

  • A submission received on a non-business day is due by 12:00PM on the next business day.

The new Business Calendar settings on the System Settings page (Administration > System Settings) allow you to designate certain days of the week as business days or non-business days and particular days of the year as non-business days. You can export these settings along with the rest of the System Settings and import them into other environments.

More information about the Business Calendar settings can be found in Application Settings Overview. To learn how to define SLAs for submissions, see Prioritizing Submissions.

Machine Classification

Fixed

Classifying Structured documents with extreme aspect ratios – We've resolved an issue that caused out-of-memory errors to occur when the machine attempted to classify structured documents with extreme aspect ratios (e.g., 600 x 2 pixels). As part of this update, the system pre-calculates classification data for releases containing Structured layouts. These calculations may increase the time required to process the release the first time it is used.

Text Classification

Updated

Adding multiple labels from a dataset – With the introduction of multi-label Text Classification models, keyers can apply multiple labels from a dataset to a document during Text Classification Supervision. Multi-label models allow multiple labels to be applied through a single Text Classification Block, reducing implementation and testing time for multi-label use cases.

You can create a multi-label model by providing training documents that have multiple classes assigned to them. The Classes column in the Training Data Management table on the model details page shows the classes that have been assigned to each document.

For more information on multi-label models, see Text Classification.  

Unstructured Extraction

New

Unstructured extraction for on-premise deployments running on Docker or Podman – You can now extract data points from unstructured documents in on-premise deployments of Hyperscience running on Docker or Podman. To do so, your trainer machine needs to have both a GPU (graphics processing unit) and a CPU (central processing unit), as training Unstructured models requires additional processing resources.

For more information about the technical requirements for Unstructured extraction and setting up a trainer with a GPU, see Enabling Trainers with GPUs in On-Premise Deployments.

Custom Supervision

New

Formatting description text in text_block components with Markdown – You can now use Markdown to format description text in text_block components. With Markdown, you can add headings, links, lists, bold text, italicized text, and other elements to your Custom Supervision interface. For example, you can format error messages and output from Custom Code Blocks and Python Code Blocks to make them more noticeable to keyers.

For more information on the Markdown features the Custom Supervision interface supports, see our Flows SDK documentation.

Output Blocks

Updated

OAuth support for UiPath Notifier Output Blocks – We've added support for OAuth connections in UiPath Notifier Output Blocks. To configure OAuth as an authentication method, you need to create an application grant in UiPath and enter the application's ID and secret in the block's settings.

For more information, see UiPath Notifier.

Audit Log

New

Accessing the Audit Log in the application – On the Audit Log page (Administration > Audit Log), users with the View System Health and Settings permission (given to the default  System Admin permission group) can view details about actions completed by the system or its users. This information was previously accessible only in /admin or via the Audit Log API endpoints.

Examples of logged actions include:

  • user logins;

  • model-management activities such as uploading, training, and downloading models;

  • the creation, editing, and deletion of authentication groups and permission groups;

  • the completion of Supervision tasks;

  • …and more.

For each activity, the Audit Log shows:

  • the activity ID,

  • the date and time of the activity,

  • the name of the activity, and

  • the user who completed the activity.

System Admins can also filter the Audit Log's contents based on timestamps, activity names, the users who completed the activities, and whether the activities were completed by a human or the machine. They can then download information about the activities that match the filter's criteria.

To learn more about the Audit Log, see Audit Log.

Databases

Notice

Support for PostgreSQL 10.x in Hyperscience v38 – Beginning in v38, the Hyperscience application will no longer support PostgreSQL 10.x. PostgreSQL ended support for 10.x on November 10, 2022.

The following databases will be supported in v38:

  • PostgreSQL 12.x, 13.x, and 14.x

  • Amazon RDS for PostgreSQL

  • Oracle 19c with DBMS_ALERT privileges

  • Amazon RDS for Oracle

  • Microsoft SQL Server (MSSQL) 2016, 2017, and 2019 with Service Broker enabled

  • Amazon RDS for SQL Server

  • Azure SQL Managed Instance

For more information on database requirements, see Infrastructure Requirements (Production).

File Storage

Fixed

Running run.sh init as root – We've fixed an issue that caused HS_PATH/media to be created as root when run.sh init was run as root and the HS_PATH/media directory was missing. This issue prevented the application from initializing the file store. 

As part of this update, the system verifies the presence and ownership of HS_PATH/media when run.sh init is executed. To learn more, see File Storage Overview.

Updated

Enhancements to directory structure – We've updated the directory structure in file stores for faster data retrieval. Files are now stored in a directory with six levels to minimize the number of files stored in any single directory, preventing performance issues that may occur in high-volume instances.

As part of this update, we've added the following as valid values of the FORMS_STORAGE_MODE “.env” file variable:

  • FILE_EX

  • S3_EX

  • AZURE_BB_EX

These values replace FILE, S3, and AZURE_BB, respectively, as valid values of FORMS_STORAGE_MODE. Instances with these values will have their file stores migrated to the new structure upon upgrading to v37. If you do not want your file store to use the new structure, set FORMS_STORAGE_MODE to FILE_LEGACY, S3_LEGACY, or AZURE_BB_LEGACY.

To learn more about configuring file storage, see Network-attached Storage or Local Storage, S3 Cloud Storage, and Azure Blob Storage.

Monitoring

New

OpenTelemetry – You can now expose an OpenTelemetry data stream for consumption by application-performance monitoring tools. This update allows you to monitor the usage and health of your Hyperscience instance alongside that of other applications in your IT infrastructure.

The stream includes metrics for submission volume and throughput, time to completion for blocks and submissions, response times, error rates, and connectivity issues, among others.

To learn more about streaming OpenTelemetry data, see OpenTelemetry.

SaaS 

New

Self-service user management for deployments with built-in user management – Rather than relying on Hyperscience to add, edit, and disable users, System Admins can perform user-management tasks in the application without assistance from Hyperscience. The self-service user-management updates in v37 provide an interface for maintaining existing user data. 

When this feature is enabled, System Admins can complete the following tasks on the Users page (Administration > Users):

  • Add users

  • Change users' permission groups

  • Revoke a user's access to the system

  • Reset a user's password

  • Reset the additional authentication factor for a user

Individual users can also reset their own passwords and additional authentication factors, regardless of whether they are in the System Admin permission group.

Note that Hyperscience can still perform user-management tasks on your behalf, even when self-service user management is enabled.

To enable self-service user management, contact your Hyperscience representative.

For more information about this feature, see Adding, Editing, and Deactivating Users.

Autoscaling of computing resources – With the autoscaling capabilities in v37, the system maintains the same latency during spikes in submission volume as it does when submissions arrive at the typical rate. 

API

New

Initiating and monitoring flow runs – With the new /api/v5/flow_run endpoints, you can create scripts to automate frequently performed tasks, such as deploying and testing flows.

You can complete the following tasks with these endpoints:

  • Run a flow

  • Monitor a flow run

  • List flow runs

More information about these endpoints can be found in our API documentation.

Updated

Preparing for the sunsetting of API v4 – We will sunset v4 of the API in Hyperscience v38. In v37, API v4 can only be enabled by going to /admin/common/systemconfig/1/change and selecting Activate Deprecated API v4, which is disabled by default.

While you can use API v4 with Hyperscience v37, we strongly recommend migrating to API v5, which includes all of the latest features and will continue to be supported after the release of v38.