User Experience
Filtering by date in instances whose UI is not configured in UTC timezone – If your instance’s UI is not configured in UTC, you might not see all results when you filter by date in the application. The database records are in UTC, and the filtering occurs on the database side. For example, if your UI’s timezone is set to EDT (UTC−04:00), and you upload a submission at 10 PM on October 22, filtering your Submissions table by that date will not retrieve this submission. The reason for this issue is that the given submission’s creation date will be October 23 in the database.
Languages
Low confidence for fields with Portuguese "ã" character – The system transcribes Portuguese fields containing "ã" with low confidence, and sometimes it marks these fields as illegible.
Transcription of Korean characters, numbers, and punctuation – The system may translate certain Korean characters incorrectly. It may also
Specifically:
"것" may be transcribed as "거 K" in a Freeform Characters field or as "거" in a Generic Text field.
"많은" may be transcribed as "–" in a Freeform Characters field or as "ᆭ" in an Address field.
"제" may be transcribed as "ᅰ".
"[" may be transcribed as "(" in fields with Korean text.
Numbers in the format XXX.XXX,XXX are transcribed with low confidence.
Layout Editor
Expanding a table column with the same name as a field – Currently, if you expand a table column in the Layout Editor that has the same name as a field in that layout, that field is also expanded.
"Not in [layout language]" option for checkboxes and signatures in Semi-structured layouts – We're fixing an issue that causes the Not in [layout language] option to be shown for checkboxes and signatures in Semi-structured layouts. This option does not apply checkboxes or signatures.
Confirmation of deleted fields in Semi-structured layouts – When a user deletes a field in the Layout Editor for a Semi-structured layout, they do not receive explicit confirmation that the field has been deleted.
Deactivated fields in Structured layouts – Deactivated fields may not always appear in the Layout Editor's field list, depending on whether changes to the layout have already been saved or are in the process of being saved.
Selecting fields in a Semi-structured layout – If a user selects fields in a Semi-structured layout, and they scroll down the page and select additional fields, the fields they initially selected become inaccessible.
Copying fields in different languages in Structured and Semi-structured layouts – If a user copies a field from a layout in one language to a layout in a different language, the result depends on the type of layout the user is copying from and pasting to:
In Semi-structured layouts, the field's language changes to the language of the layout it is pasted in.
In Structured layouts, the field retains its original language.
Layouts
Renaming Field Customization Group in instances with MSSQL databases – We're working to resolve an issue that prevents users from renaming Field Customizations that have at least one field selected. This issue occurs in instances with MSSQL databases.
Sorting the Layouts table by "Last Updated" – Currently, if a user sorts the Layouts table by the Last Updated column, no layouts appear in the table.
Potential Layout Variations
Sorting the Potential Layouts table – Clicking the Job ID, Ran On, or Pages column headings does not sort the table. However, users should be able to sort the table by these columns.
Pagination after clicking “Find Layouts” – If there are more than 25 pages in the jobs listed in the Potential Layouts table, and a user clicks Find Layouts, only 25 pages are listed on the Run Potential Layout Variation Job page, and there is no way for the user to view the rest of the pages.
Model Details Page
Redirection after downloading Classification models – When a user clicks the Download model button for a Classification model, they are redirected to a page that shows the download_uuid of the download request. Instead, the user should receive a notification when the download is ready.
Training document count on the Model Details page – The training document count in the Identification Model card on the Model Details page does not automatically update after you’ve annotated or deleted training documents. To update the training document count, refresh the Model Details page.
Deleting a training document during training data analysis – If you delete a training document during training data analysis, the training data analysis will fail.
Releases
Empty shared_field_id values in releases imported from older versions – When importing releases from older versions of Hyperscience, fields' shared_field_id values are empty, which causes all the fields on a page to have the same bounding box.
Workaround: Copy the release in v35 and deploy the copy.
Identical uuid and origin_version_id values for layouts in downloaded releases – The layouts contained in the templates folder in a downloaded release have identical uuid and origin_version_id values in the template_versions element.
Unnecessary template_versions folder in downloaded releases – Each downloaded release contains a template_versions folder, which is not needed, as version information is contained in the templates folder's JSON files.
Flows
Complete Blocks in custom flows – If you have a Complete Block in a custom flow that does not have all of the blocks that are included in pre-built flows (e.g., Machine Classification), submission processing may fail in that flow.
Workaround: Include all of the blocks that are included in pre-built flows, even if they do nothing in the custom flow.
Error codes for missing flows – When a user attempts to retrieve a flow that does not exist, the system returns a 500 status code rather than a 404 status code.
Machine-related warnings for flows – In some circumstances, the system may show a warning message that says "Your workflow is now live, but needs machines to power it," even though there is a machine assigned to the flow's submissions.
Notifications sent from disabled notification flows – If you add an output connection to a notification flow (e.g., Submission State Notifications for Document Processing V35) and enter its secrets in the connection's settings, notifications may be sent from that connection even when the flow is disabled. These notifications are sent when submissions are processed through the pre-built Document Processing flow that corresponds to the notification flow (e.g., Document Processing V35).
Responsiveness of Flows page in instances with many flows – The Flows page can become unresponsive if there are 80 or more flows in the instance.
Flow Studio
UI for blocks with long descriptions – If a block's description contains 126,000 or more characters, some text may appear on the canvas rather than within the block's box.
UI for custom blocks with multiple Flow Identifier inputs – When multiple flows are linked to a single custom block, some of them may not appear to be directly connected to that block in Flow Studio.
Showing the same validation messages multiple times – The same validation error and warning messages are shown to the user each time they save or view a flow in Flow Studio, even if those messages have already been shown to the user in the same session.
Discrepancies between warning messages for live and disabled flows – Warning messages for untrained Field Locator models do not appear in the Flow Studio for disabled flows. These messages only appear for live flows.
Flow Execution
Unexpected errors when viewing the execution of blockless flows – If a user attempts to view the execution of a flow with no blocks (e.g., a flow that provides some custom logic), an "unexpected error" occurs. Instead, an empty Flow Execution page should be shown.
Duplicate exception messages when custom logging is enabled – When logging for Custom Code or Python Blocks is enabled, each exception that occurs in those blocks may be logged twice.
Halted submissions with all blocks "Completed" – We're fixing an issue that causes all blocks in a halted submission's Flow Execution diagram to be shown as "Completed" in some circumstances.
Submission Processing
Documents sent to Flexible Extraction after Document Classification – After all of a document's pages are matched to the document's layout variation in Document Classification, the document may be sent to Flexible Extraction rather than Transcription.
Indefinite “loading task” screen for halted submissions – If a submission halts after Document Classification, an indefinite “loading task” screen appears.
Submissions Table
Displaying names of disabled flows – Disabled flows appear with incorrect names in the Submissions table.
Submission Output Page
Layout variation names for partially machine-classified documents – If a submission contains a document with both machine-classified and manually classified pages, its layout variation name may not be shown in the Layout Variation column of the Documents table. Instead, that column may contain the layout variation's version name.
Cases
Clearing the filters on the Cases page – On the Cases page (Submissions > Cases), if you manually remove filters, the number of applied filters shown in the Filters button will not decrease.
Tasks
Task type names in the Supervision Tasks card – The task type names in the Supervision Tasks card (Tasks > Overview) overlap on screens with small resolutions.
Completing Supervision tasks from the Perform Tasks tab – Completing Supervision tasks very quickly from the Perform Tasks tab (e.g., pressing the Enter or Return key immediately after you’ve entered a Transcription task with a single field that you need to transcribe) can sometimes force you out of the Supervision queue even if you have remaining tasks to complete.
Document Classification
Structured Layout Match Threshold not respected during Manual Classification – We're fixing an issue that prevents the system from respecting a flow's Structured Layout Match Threshold value during Document Classification tasks. For example, users may not be able to restore pages that are deleted from a machine-classified Structured document to their original positions in the submission; instead, they may be appended to the submission's end.
Creating a layout variation with the same name as a deleted variation – We're resolving an issue that prevents users from creating layout variations with the same names as deleted layout variations.
Machine-classified pages from v33 – Currently, the indexes of pages that were machine-classified in v33 are missing from the v35 application. This issue may prevent users from recategorizing documents containing these pages.
Manual Classification of pages with dark backgrounds – We're fixing an issue that prevents pages with dark backgrounds from being included in Manual Classification tasks in some circumstances. The system may classify these pages as empty pages.
Drag-and-drop in Document Classification – Valid locations to drop pages may be easy to miss while dragging the pages to their proper locations. The cursor needs to be positioned between two pages, and the space between those pages needs to be colored green before dropping a page in that location.
Table Identification
One-click bounding boxes for the first row’s cell in a column – If a Table ID model predicts a few cells in a column but does not predict the first row’s cell in this column, you will not be able to properly select and create this cell’s bounding box. Hovering over this cell with your mouse does not allow creating one-click bounding boxes from all angles.
Child table cells in Table ID QA – Child table cells that are identified during Table ID Supervision do not appear during Table ID QA.
Two rounds of Table ID QA to reach consensus – If you:
add an extra row to your table during Table Identification, and
submit the document, and
complete Table ID QA without making any changes,
then consensus is not reached. To reach consensus, you need to complete a second round of Table ID QA.
Overlapping bounding boxes of table cells – If bounding boxes of multiple cells overlap, you cannot resize the overlapping parts.
Workaround: To avoid resizing, delete all overlapping bounding boxes and redraw them.
Tooltips in Table ID tasks – Tooltips in Table ID tasks interact with mouse cursors. For example, tooltips appear when you hover over a column button. To continue interacting with the Table ID task, you need to move your cursor out of the tooltip.
Flexible Extraction
Manual transcription of partially machine-classified documents – If a document has both machine-classified and manually classified pages, and it is sent to Flexible Extraction, none of the fields are transcribed. Instead, all of the fields must be transcribed manually.
Connections
Saving connection settings with missing required fields – We're fixing an issue that allows the user to save a connection's settings with missing required fields. In these situations, the system should show an error message indicating that required information is missing. This issue affects the following blocks:
Message Queue Listener
Box Folder Listener
HTTP Notifier Output Block
UiPath Notifier Output Block
Salesforce Notifier Output Block
Message Queue Notifier Output Block
Box Notifier Output Block
Using only “Endpoint URL” in Amazon S3 test connections – We're working to add an option that, when enabled, allows the user to indicate whether both the URL and Endpoint URL should be used in a test Amazon S3 connection or only the Endpoint URL should be used.
Successful test connections for HTTP Notifier Output Blocks with invalid header values – We're working to resolve an issue that allows test connections for HTTP Notifier Output Blocks to be marked as successful when Authorization Type is set to HTTP Header and the Authorization Header Value is invalid.
Training
Checking trainer availability in SaaS deployments – In SaaS deployments, the trainer automatically turns off when not in use. However, the pages for Classification models and finding potential layouts may indicate that there is no trainer connected to the application when the trainer is actually present but turned off.
Reporting
Number of submissions in the System Throughput report – The number of completed submissions in the System Throughput report (Reporting > Overview) is lower than the actual number of completed submissions.
Filtering by deleted flows in the Automation Training Result report – Filtering by deleted flows in the Automation Training Result report (Reporting > Overview) results in an error message.
Administration Settings
Clicking the Reset to Default Settings button – Clicking the Reset to Default Settings button in Administration > Settings sets a few Data Deletion settings incorrectly compared to their actual default values.
Upgrades
Upgrading an instance with Podman containers – Upgrading an instance with Podman containers results in an error.
Workaround: Before upgrading, manually remove prior application containers and images.
API
Creating submissions with a null value for the cases parameter – Creating submissions with a null value for the cases parameter results in a 400 error message.
Flow-restricted information in the Submissions CSV report – Adding a Workflow UUID column to the Submissions CSV report created via our API (e.g., /api/v5/submissions/csv?include_extra_fields=flow_uuid) allows you to view information for all submissions, including the ones whose flows you don’t have access to.