Resolved issues
This topic lists previously reported issues that are resolved in Tungsten TotalAgility 2026.3.
ID | Description | Solution |
224820 | After setting a process in a job query and selecting save, then editing it again, the process was not saved. | The process is now saved. |
2245244 | In Transformation Designer, when a user configured a Table Locator and selected an input locator for the "Find the start of the table" and "Find the end of the table" options, the value displayed in the drop-down at the time of selection was not the value saved to the project. On reopening the Table Locator properties, the value differed from the previously selected one. | The value of the input locator that was selected during configuration is saved to the project. |
2244384 | If a query was saved using the "contains" operator with criteria provided, after saving and reopening the configuration, the Data Access configuration search criteria did not work:
| The Data access search criteria work as expected. |
2244006 | In TotalAgility 2026.2, the Generative AI activity failed to make a Generative AI call due to the installer pointing to the incorrect AI Proxy endpoint URL. | The installer now points to the correct AI Proxy endpoint URL. |
2243457 | When converting PDF to TIFF using the Document Converter, if the resulting TIFF exceeded 2 GB, the job was suspended. | A different overload is used to split pages. |
2242726 | When TotalAgility on-premise multi-tenant 2026.1 was used with Windows authentication for a tenant, the "Update Tenant Database Server" function did not update the connection string inside the Server_Data table > CEBPM settings, resulting in an Error deleting chunks error in Message Connector. | The connection string in the Server_Data table > CEBPM settings is now updated when Windows authentication for a tenant is used. |
2242496 | After the upgrade, the Classification or Extraction training automatically saved the project to the TotalAgility database without the user explicitly selecting to save it. | Unintended database version increments are prevented during auto-saves triggered by training commands. |
2242290 | When using DoHoleFill in eVRS strings, bars were removed from the barcodes. | The EvrsCheck.exe utility is updated to use eVRS version 3.7.0.0.0.523. |
2242025 | The Data access activity that worked correctly at design time failed at runtime with the Azure.Identity assembly binding is missing in Agility.Server.Azure.WorkerRole error. | The Azure identity binding was added to the Worker role configuration. |
2239871 | Recursive onTextChanged actions were executed when loading Validation. When loading a document validation form, changing a field value triggered the onTextChanged actions. Under certain conditions, this caused the action to be executed recursively, and the form became unresponsive. | The onTextChanged actions are not fired when the field is changed programmatically. |
2239202 | When Display rules were configured to control table column visibility (such as table1.Column1, table1.Column2), using Display rule actions to hide/unhide table columns returned the following error during page load or debug: Cannot read properties of undefined (reading 'getWidth') | The column element’s existence is checked before attempting to retrieve its width, preventing an undefined property error. |
2239071 | In a form designed to retrieve and update resources, GetWorkerResource2 and DataObject failed with umlauts, converting them to "?" (question marks). | UTF-8 encoding is used to support Unicode characters (including umlauts and other special characters). |
2238690 | Out-of-the-box document validation rules caused the Classification node to suspend with a Document with ID '' can’t be found error. The error disappeared if the separation profile or the document validation rule was removed from all document types in the extraction group. | No error occurs when the document validation rule or separation profile is used. |
2238220 | When performing the "Restart Next" action on an activity in the Job Properties form, the resource name was always incorrectly recorded as the job creator's name in the audit log, whereas the job history correctly displayed the resource that performed the action. | The logged-on user is recorded in the audit log when the “Restart Next” action is performed. |
2238046 | Following an upgrade to 2026.1.2.24177.204, merge fields in generated documents were not being resolved correctly.
| The barcode merge field replacement logic is modified to apply only to barcode merge fields and not to normal merge fields. This fix ensures that barcode fields are processed correctly while preserving the existing behavior of standard merge fields. |
2238003 | In TotalAgility Azure, the GetDocumentAsFile SDK method returned the full file path rather than the file name only. | The source file path definition is updated to return the file name if the environment is multi-tenant. |
2237926 | When a Kodak i4250 scanner with an A3 flatbed accessory was used with TotalAgility in Automatic mode, the flatbed accessory was unresponsive. This was due to the Web Capture Service not working correctly with flatbed accessories for Kodak scanners. | Flatbed can be used in Automatic mode. |
2236950 | When the Transformation Server executed an AEL using a "Bring Your Own License" Generative AI provider with settings: Provider = Azure OpenAI, Model = gpt-5.2-chat, and Temperature = 1, the following error occurred: Unsupported value: 'temperature' does not support 0.0 with this model. Only the default (1) value is supported. | Previously, Copilot for Extraction did not use the temperature configured in the AI provider. This issue is resolved, and the configured temperature is now properly applied during processing. |
2236891 | When the fields panel was minimized in the scan form (specifically the Document View area), creating a job worked the first time, but it failed on subsequent scans. Selecting the "Create Job" button had no effect; however, maximizing the minimized panel restored the button's functionality. | The issue with job creation after minimizing the fields panel scan form is fixed for subsequent scans. |
2236757 | When creating a folder in the TotalAgility Advanced Studio, field names on a Child field were not inherited from the Parent field. At runtime, the inherited folder fields had default values rather than the expected inherited values. | When the field name is changed in a folder, the same is updated in the child folder type field names. |
2236164 | Web Capture Service returned a corrupt PDF upon import when Capture Client forms tried to upload the entire file or large (such as 40 MB) chunks. | The logic is corrected so that the Web Capture Service works correctly with large files. |
2236155 | Capture Client forms ignored chunk settings from Advanced Studio (System > System settings > Scan forms > File upload > Maximum binary chunk size). | The issue is fixed. |
2235517 | During table extraction, the ATL (Advanced Table Locator) failed on the “Classify” call with the following error: An error occurred while processing the 'Classify' call of the table extraction. Error: "Named pipe operation ReadFile failed. The occurrence of the issue depended heavily on the OCR engine used — for example, it did not occur when Azure AI DI was used to OCR the same document. | The result can be extracted using ATL at design time and runtime. |
2235208 | A Generative AI activity was suspended with the following generic message: Error occurred while processing the document. Either the document failed to be retrieved from the repository, or it is not supported. Several hidden exceptions were recorded in the Transformation Server logs. For example, the Kofax.CEBPM.CPUServer.Common.Exceptions.CpuServerException: Activity timed out error indicates that the MaxSyncProcessingDurationInSeconds timeout was breached before the narrative text could be generated. | The following two changes are applied:
|
2235044 | When displaying the upload percentage, the "Uploading..." progress bar repeatedly showed 100% progress during a single file upload. | The progress of uploading both single and multiple files is based on file size. |
2235016 | When key-value prediction was performed (such as triggered by AEL), the page extension stored references to word objects. If a page was subsequently rotated, OCR was deleted for that page, leaving KVP references pointing to non-existent words. When the document was then processed through Classification, Extraction, Generative AI, or PDF Generation, the activity was suspended with the FoX exception number 8 error. | The key-value pairs of a document page are removed if no text representation exists for that page. This handles scenarios such as a user rotating a page in Document review after the key-value pairs are already been determined in Classification. |
2235013 | PDF text extraction produced inconsistent results between the Transformation Designer and the Transformation Server. For example, the Format Locator returned 14.154.00 for Transformation Designer and 15400 for Transformation Server. This discrepancy sometimes caused locators that worked correctly during design and testing to fail or return unexpected values when documents were processed at runtime on the Transformation Server. | The PDF text extraction in the Transformation Server has been aligned with the Transformation Designer. PDF text layer extraction now produces identical results in both Transformation Designer and Transformation Server. |
2233795 | Two related issues were identified with Job queries in TotalAgility Workspace:
| Functional access is checked instead of Maintenance access during query categorization, ensuring that queries appear correctly under Personal or Shared based on the assigned resources. |
2233393 | Duplicated records appeared in Reporting databases, which produced the Exception: Violation of UNIQUE KEY constraint errors in Reporting logs. | The reporting processing logic is changed to automatically prevent duplicate records. |
Tungsten TotalAgility 2026.3
© 2026 Tromba Technologies, Inc.

