Release 9.x.x

Owned by Benjamin von Deschwanden
Last updated: Mar 20, 2023
6 min read

Current release: 9.3.0 / Mar 17, 2023

Initial release date: Feb 23, 2023

This release adds support for Single Sign-On (SSO) using OpenID Connect and improves the handling for long publishing operations.

 

New features

  • Support for Single Sign-On using OpenID Connect

  • Support for configurable data retention strategies (via API)

  • Exporting confidence scores for shared model predictions

  • Retrainable figure step

Usability Improvements

  • Improved support for long-running publish operations

  • Improved user experience for exhaustive dropdowns

  • Status filtering in production listing

  • Improved support for nested listing representations

New Features

Support for OIDC SSO

Release 9 introduces support for Single Sign-On (SSO) using OpenID Connect. To Set it up, enable SSO in the instance settings and enter the appropriate configuration from your SSO provider.

 

Enabling SSO in the Settings

 

  • ClientID, Client Secret, Auth Endpoint and Token Endpoint are mandatory fields

  • Audience and Scope are optional fields

  • The Callback URL is a read-only field

 

Entering the OIDC configuration information

 

 

 

Once you have saved the configuration settings, the Test login button becomes available. Upon clicking this button, you will be redirected to the login screen of the SSO provider you just configured. Complete the login with valid user credentials to obtain an SSO Authentication Success information screen.

 

 

 

 

  • When SSO Integration is enabled, users are provided with an SSO Login button on the start page, redirecting to the login page of the SSO provider.

  • When provided with valid credentials, the SSO provider redirects back to the platform and the user is signed in. If the user is new and there is still capacity, a new account is created automatically on the platform.

  • When the number of user licenses has been exhausted, an error message is shown.

  • The user role and expert mode access can be configured in the Settings for SSO users after they logged in at least once

 

Support for configurable data retention strategies (via API)

Description

In order to have fine-grained control over the transaction-related data retention, the platform allows to define up to 100 different data retention policies per instance via an administration API. Such policies dictate what transactions are being automatically deleted by a periodic cleanup task.

A data retention policy can use the following transaction properties to define eligibility for automatic deletion:

  • An optional workflowId. If the parameter is set, the rule will only apply to the given workflow ID. If none is given (i.e. field is omitted or set to null), the rule applies to all workflows.

  • An optional categoryId. If the parameter is set, the rule will only apply to the given category id. If none is given (i.e. field is omitted or set to null), the rule applies to all categories.
    Note: This makes most sense in combination with a workflowId

  • An optional status set containing any combination of failed , processed , exported . If set, will only apply to transactions of the given states. If it is not set or empty, will apply to all states.

  • A required ageInSeconds value which indicates how old the transaction has to be in seconds (i.e. how many seconds have to be elapsed since the file was uploaded)

 

Example:

{ "ageInSeconds": 2592000, "categoryId": null, "status": [ "exported" ], "workflowId": null }

This rule says that Transactions

  • of any workflow

  • with any category id

  • that were exported

  • are older than a month

will be cleaned up automatically.

 

By default, a single rule is set that indicates that all transactions older than a month are cleaned up (i.e. the single example rule from above).

 

Usage

In order to set specific retention policies, contact our support at support@acodis.io.

Exporting confidence scores for shared model predictions

Description

It is now possible to include confidence scores for shared model predictions in the export. To do this, enable “Export Confidence Scores” in the “Export Step” settings.

 

  • The following steps support the export of confidence scores:

    • Element (only when using a shared model)

    • Figure

    • Structure (shows the confidence score for predicting the header, NOT the header level)

  • Tables and elements/templates that are not using the shared models are currently don’t supported exporting a confidence score

 

The result of enabling confidence scores in the export is a float-valued attribute confidence with values in [0,1] representing the shared model confidence for the respective XML element.

  • For all steps where a tag name can be specified → The confidence attribute is added to the same XML element as the class attribute

  • For steps with an effect type header (structure and header X and element with effect type header) → The confidence attribute is added to the <section> element

 

Retrainable Figure Step

It is now possible to train the figure classification model to recognize specific parts of a page as a figure.

Usage:

When a figure step is included in the analysis:

  • Creating a model also fine-tunes the figure step starting with the pretrained figure recognition model

  • The statistics regarding the figure step also appear in the model screen

  • The user can also validate pages when the figure step is selected

  • The model only considers user-validated pages for the training

Usability Improvements

Improved support for long-running publish operations

When publishing a workflow, the platform now gives you a much better overview of the currently running publishing operation. A progress status indicator in the workflow’s title bar provides you with detailed information about the current state of the ongoing publishing operation. Hovering over the status indicator reveals additional information, such as the currently running publishing step, the user who triggered the publishing as well as the time when the operation was started.

 

 

In case of an error during publishing, this information is also being conveyed by the publishing status indicator, with additional details provided when hovering.

 

Furthermore, to mitigate the risk of interrupting an ongoing publishing operation, the Publishing button at the end of the workflow wizard turns yellow to alert the automation manager to the ongoing publishing operation. When clicking the button in this state, a confirmation dialog is shown that informs and prompts the user to ensure they want to interrupt the ongoing publishing operation.

 

 

 

Improved user experience for exhaustive dropdowns

Selection dropdowns that potentially contain many options now have an explicit filter input field at the top. This field allows you to quickly filter the options of the dropdown using text input. The provided text is being matched against the entries of the dropdown and yields all options that contain the provided text as a substring.

Status filtering in production listing

It is now possible to filter transactions in the production listing by their status. The status filter allows you to select one or multiple specific stati to only display those transactions that match that filter. The four individual stati that can be selected are

  • Interrupted

  • Processed

  • Processing

  • Failed

 

 

Improved support for nested listing representations

Complex nested lists - especially across page boundaries - are being structured more truthfully to their original document representation. The newly supported cases might look similar to the following examples:

 

Usage:

  • At least the following steps must be configured to observe the improved output

    • OCR/NativeText

    • Enumerations

    • Text Aggregation

    • List Formations

    • Export Step

  • The enumerations step needs to be run → “Run Analysis”