|
|
|
|
|
|---|---|---|---|
| 2319 | Train custom AI models with new Enterprise Extended license |
Agiloft's new Extended Enterprise license allows you to create and train your own machine learning models, giving you full access to the capabilities of the Agiloft AI Core. Users with an Enterprise Extended license also gain the added benefit of unlimited custom modules and around-the-clock named support.
To discuss how your company can acquire an Enterprise Extended license, contact your sales representative or email sales@agiloft.com. |
204080 |
| 2320 | Directly map Machine Learning model output data to records in Agiloft | Agiloft AI Core Machine Learning actions can now be used to directly map outputs from clause and metadata extraction models into one or more records due to new sections that have been added to the Task tab of the Machine Learning action wizard, and the addition of a new Import Action feature.
If "Update a single record" is selected, an option to "Choose the Record Based on ID" appears which always you to choose the record that should be updated.
If "Create/update multiple records" is selected, the output is converted into a JSON field. A new feature of the Import Action allows you to map the JSON output into the table and fields of your choice. This can create multiple new records or be used to update specific fields in existing records.
For information on how to configure a Machine Learning action using the new features, see the Extracting Content section of Extracting Data with AI. |
203466 |
| 2309 | Retrieve artifacts after model training |
In the Training tab of an AI Model record, data scientists can now download the full set of training
artifacts, or outputs, for more detail on the model training metrics. You can now download the full set of
training artifacts to a zip file.
There are separate zip files for each instance of training that the model has undergone. These zip files
contain text documents with detailed information about the model's training data, such as the model's
confusion matrix, confidence scores per label, total scores, and validation results.
To download these zip files, simply navigate to the Training tab of the relevant AI Model record and
click the artifact file you would like to download. Once you have the file downloaded, unzip it to access
the individual documents.
|
209169, 209165 |
| 2328 | Import JSON text to other Agiloft fields |
Pull data directly from a JSON source field into other fields in Agiloft with a new Import action function.
JSON source fields are text fields that are designed to contain JSON objects. This enhancement is
especially useful for processing Machine Learning models where the output is JSON-formatted text into
multiple records, although JSON text can be useful in scripting and other contexts as well. After the JSON
text is added to the JSON source field, you can map relevant values to other fields.
When an Import action runs on a text field that contains data, the system parses the content and
outputs the selected data as text to a new record. This data, seen under the Sample Data column in the
image below, can be mapped to relevant text fields in the new record by configuring the Fields tab of
the Import Action. For example, in the second image below the JSON attributes for the job title of the
contract signee and the name of the contract signee is the Sample Data, and can be mapped to fill fields
such as Job Title and Employee, respectively.
To set up an Import action to parse a JSON Source field:
 |
205166, 204831, 204781, 204700, 204127 |
| 2317 | Access Amazon SageMaker with an Agiloft-managed AWS account |
If you're interested in using Amazon SageMaker with the Agiloft AI Core, but are unable to provide your
own AWS account, Agiloft can configure and manage a dedicated AWS account to host your machine
learning models. Agiloft AI Core uses Amazon SageMaker to build, train, and deploy machine learning
models. To deploy the AI Credential for Amazon SageMaker in Agiloft, you need to connect to an active
AWS account during setup.
It is recommended to use your own AWS accounts when possible. If you choose to have Agiloft create
and maintain an AWS account for you, the cost of maintaining this account is passed on as part of your
regular invoice. An Enterprise Extended license is not required for integration with Amazon SageMaker.
To discuss how your company can request an AWS account, contact your sales representative or email
sales@agiloft.com.
|
|
| 2306 | Office 365 Retiring Basic Authentication |
If you use a Microsoft Exchange or Office 365 email account as an inbound email address with Agiloft, you must update your configuration by October 2020 to Microsoft's new authentication requirements. Otherwise, when the new requirements are put in effect, these email accounts will stop working.
To update your configuration, you need to configure Agiloft to connect with Azure Active Directory, and then update the affected email accounts with your Azure credentials to perform authentication. Refer to Microsoft Exchange and Office 365 Email Accounts for detailed steps. |
208895, 208889, 207972, 207971 |
| 2327 | Easily export HTML and XML content as plain text |
HTML and XML formatting can be difficult to work with in exported data, especially if that data is used for reporting or passed on to another system or process. Now, KB content that is represented in HTML or XML can automatically be exported as plain text instead, stripping out unnecessary code and tags. This option is available in several export workflows:
This option is turned off by default.
|
|
| 2366 | Use OCR actions to generate plain text output |
Optical Character Recognition (OCR) actions in Agiloft have become more flexible thanks to a new
option that allows users to store the output as a plain text file. Plain text files are much easier to work
with if you need to do additional processing on the OCR output, such as parsing the text or passing it to
an integrated system. Previously, all OCR output was in PDF format.
To use this new option, create or edit an OCR action, navigate to the Options tab, and select Text in the
new Output File Format option. For more information, see OCR Action.
|
206339 |
| 2333 | New option to ignore errors in OCR actions |
OCR actions now have improved flexibility for handling errors when the action is run by a rule. The
OCR action wizard now offers the option to ignore errors from corrupted PDFs, which allows the rule to
continue running even if a corrupted document causes a problem with the OCR action. For example, if
you use a rule to run OCR on batches of files, ignoring errors allows the rest of the batch to be processed
after encountering one corrupted file. Previously, these errors caused the rule to stop running entirely.
This option has no effect on actions run by action buttons. If a corrupted document causes an error in
this case, all actions from the action button stop running.
For more information about OCR actions, see OCR Action.
 |
198459 |
| 2310 | New sync utility to identify and repair linked field issues |
If you need to run an entity set sync with linked field auto-propagation disabled, you can use a new utility to analyze and repair linked field issues that arose from the sync. The new LF Propagation Check utility checks the KB for records where the value of a linked field is misleading or missing. These records are listed in a report and broken down by table, making it easy to check individual records. You can use the Repair LF or Repair All LFs button to automatically attempt a repair for the linked field value. The repair options generate their own report on what was changed to repair the record.
To run the utility:
For more information about sync, see Pushing Changes to Production.
|
208031, 208026, 208015, 208014, 208010 |
| 2314 | New warning when running sync without linked field auto-propagation |
Running a sync without linked field auto-propagation now triggers a warning. This warning advises the user that potential side effects may occur, and to make sure this option is disabled on purpose. Linked field auto-propagation is important for sync because it makes sure linked fields are populated correctly when they are transferred or changed. Sometimes this feature is disabled on purpose, such as if the KBs involve a large number of records or a lot of linked field columns. This warning helps prevent users from accidentally leaving the option disabled.
Linked field auto-propagation is enabled and disabled using a Setup Assistant command. For more information, see Pushing Changes to Production. |
|
| 2313 | Entity set dependencies report now includes records |
Syncing an entity set sometimes requires syncing records as well as entities, usually when a record is
used as a default value or search condition. Previously, these records were not visible anywhere before
or during sync, which made it difficult to prepare for or troubleshoot them. Now, when you select an
entity set and click Report Entity Set Dependencies, the report includes a Records tab that lists any
records being synced as dependencies.
 In general, sync is not intended to transfer records. This feature was added for visibility and to make it
easier to troubleshoot sync issues and adequately prepare for any necessary cleanup in the target KB
after sync.
For more information about sync, see Pushing Changes to Production.
|
207638, 207624 |
| 2311 | Final sync reports now include report on transferred records |
After completing an entity set sync, the final sync report now includes a linked report on any records
that may have been transferred. While sync is not generally used to transfer record data, records are
sometimes included as dependencies, usually when a record is used as a default value or search
condition. This report makes it easier to review all the transferred records so they can be checked in the
target KB after sync.
 The record includes the type of record, record ID, and summary value in both the source and target
KBs, the sync operation, and the sync progress.
 |
207626 |
| 2296 | Improved sync performance | Running sync from an entity set which includes one or more tables that are new to the target KB have
been optimized to be quicker and more efficient. This improvement is especially noticeable during an
indirect sync when a carrier KB file is created, as carrier KB files are created from scratch and do not
contain any tables by default. |
202976, 202975 |
| 2299 | Control KB export permission for admins |
Exporting a KB is resource-intensive and can involve a lot of sensitive data. You can now use a global
variable in the Admin Console to control whether administrators have permission to export KBs in Agiloft
format in the Power User Interface. Removing this permission can avoid performance and disk space
problems by preventing multiple or very large exports on servers where several KBs are hosted
together.
By default, this variable is set to Yes, so admins do not lose access unless you change the variable
setting. If you change the variable to No, admins can no longer select the Agiloft option on the Data
Format tab, which prevents exporting complete KBs. Admins are still able to export the data in the KB
using a different data format.
To view or change this global variable, log in to the Admin Console and go to General > Variables
and edit the "Allow KB admins to export their KB" global variable. Set the value to No to restrict access;
set the value to Yes to provide access.
 |
|
| 2318 | Python upgraded to version 3.7.4 | The default version of Python has been upgraded to 3.7.4. By default, previous versions of Python are
not supported. If you need to use a different version of Python, go to Setup > System > Manage Global Variables, locate the "Location of external Python directory" global variable, and enter the location of the appropriate Python version in the format "python.dir=/filepath". For example: externalpythonlocation=/user/bin/ |
|
| 2297 | File size limits now checked before upload |
When a File with Versioning field has a maximum file size defined, the system now checks the size of
the selected files before processing the actual upload. This saves time for the user, who no longer has to
wait for the whole file to upload before learning the file is too large.
 |
201382 |
