Vantage 3.1.0 release notes

Documentation Index

Fetch the complete documentation index at: https://docs.abbyy.com/llms.txt

Use this file to discover all available pages before exploring further.

​

New Features

​

PDF Processing Mode

​

Improvements

​

Security

• Critical security updates have been applied to address identified vulnerabilities.

​

Fixes

​

Advanced Designer

• Fixed an issue in Advanced Designer where the visual layout of pipeline arrows for IF activities was cluttered. Branches from an IF block now render in a clearer, more organized structure consistent with the 2.7 experience. (656385)
• Fixed an issue in Advanced Designer where “Export Set with Labeling” failed for some documents with multiple pages. Multi-page documents now export successfully. (659342, 659546)
• Fixed an issue in Advanced Designer where importing large skill files failed with 500 Internal Server Error and caused a CPU-consumption spike from repeated retries. The per-request timeout has been increased from 90 seconds to 10 minutes, and HTTP error retries are now capped at 3 attempts (previously unlimited). Large skills now import successfully without retry loops.
• Fixed an issue in Advanced Designer where fields that had been disabled in a derived skill’s Activity Properties were still appearing as enabled in the Extraction Rule Activity Editor. Only fields explicitly selected for the activity are now shown. (598060)
• Fixed an issue in Advanced Designer where ICR recognition results were poor on the initial processing of a form, but improved after manually adjusting the extraction region. Fields with correctly defined regions now produce accurate recognition results without requiring manual adjustment. (589813)
• Fixed an issue in Advanced Designer where adding an Extraction Rules activity to a derived skill would re-enable fields that had been disabled in Field Management, causing them to appear in the activity editor and require mapping. (604372)
• Fixed an issue in Advanced Designer where the FlexiLayout import documentation lacked guidance on differences between ABBYY FlexiLayout Studio and Vantage matching results. The documentation now advises users to review matching results on several documents and adjust search elements as needed before publishing the skill. (569285)

​

API

• Fixed an issue where Quality Analytics transactions were silently dropped from the Reporting API (/api/reporting/v2/qa/process-skills/documents) when a dimension field value was 257 characters or longer. Dimension field values are now truncated to 100 characters before being written to the reporting database, ensuring all qualifying transactions appear in the API response. (652805, 652966)

• Fixed an issue where the GET /api/reporting/v2/transaction-steps endpoint returned an empty response regardless of the parameters specified. Transaction step data is now returned correctly when valid query parameters are provided. (605421)

​

Built-in Skills

• Fixed an issue in derived Document skills based on the Invoice JP built-in skill where field values were incorrectly merged with adjacent field values on unseen documents, despite correct extraction training. The issue was caused by a limitation in Technology Core 2.0 and is resolved with the updated Technology Core in Vantage 3.0. (596472)
• Fixed an issue where the Vendor Name field in the Invoice EU skill could extract extraneous label text (such as “Seller:” or “Sprzedawca:”) along with the vendor name, even after training. (566406)
• Fixed an issue in the Form 1099 Composite skill where values from the “Wash Sale Loss Disallowed” column in the “1099-B Summary of Sale Proceeds” table were incorrectly extracted into the “Net Gain or Loss” column. (587114)
• Fixed an issue where a quantity value of “1” in invoice line items was not recognized or extracted by the invoice skill, while all other quantity values were extracted correctly. (605975)

​

Classification Skill

• Fixed an issue where retraining a classification skill after upgrading the Technology Core version from 2.0 to 2.4 could fail with a training error, preventing the skill from being updated successfully. (593819)

​

Data Catalogs

• Fixed an issue where a No records found in Data Catalog error was incorrectly displayed in Document skills configured with a Data Catalog Lookup rule, even when the associated field was empty or a valid record had been manually selected from the catalog. (601613)
• Fixed an issue where Data Catalog lookup rules produced errors or returned incorrect results when the lookup field’s value was populated by a script rule or other business rule. The catalog lookup was executed before the rules that set the field value had run, causing the lookup to operate on an empty or stale value. Data Catalog lookup rules are now re-evaluated after field values are updated by scripts, script activities, or other business rules. (530164)
• Fixed an issue where Data Catalog lookup rules auto-populated dependent fields even when the lookup key field had no recognized value on the document. Fields associated with a catalog lookup rule are no longer populated when the source field value is absent from the document. (570252)
• Fixed an issue where creating a new Data Catalog or uploading and updating an existing Data Catalog could fail with an error, preventing users from managing catalog data.

​

Document Skill

• Fixed an issue where field property changes in a Document Skill — including toggling Required field — would not save or cause publish to fail. (660923, 662700)
• Fixed an issue in Document skills where line-item field values were only partially extracted even when the full text was contained within a single markup bounding box. (586848)
• Fixed an issue where using the Fast Learning activity in a derived Document skill could degrade extraction accuracy for out-of-box fields inherited from the base skill, such as Ship To Address, Bill To Address, and Line Items. Extraction quality for these fields is significantly improved with the updated Technology Core in Vantage 3.0. (598328)
• Fixed an issue where applying a Check Sum Rule to a table column with a large number of line items (approximately 120 or more) would fail with a constraints error, causing the document to be sent to Manual Review unexpectedly. (637858)
• Fixed an issue where skill extraction accuracy decreased after upgrading to technology core 2.4 due to table column values being incorrectly merged, causing overall recognition results to drop significantly compared to previous technology core versions. (583312)
• Fixed an issue in derived skills where field extraction regions could capture unintended extra characters, causing extracted values to include text beyond the intended field boundary (for example, a product code extracting as “FSE32407Z Ge” instead of “FSE32407Z”). (592190)

​

Manual Review

• Fixed an issue where pressing Enter on a field containing validation errors (red text) during Manual Review caused the field status to spin on Updating, and then fail and revert the value. Now, the field edits during review save correctly. (660270)
• Fixed an issue in Manual Review where a document would disappear after being dragged and dropped onto its own document section or back to its original position. Rejecting and reopening the task would restore the document. (612839)
• Fixed an issue in Manual Review where clicking on a field in the extraction form did not reliably navigate to the corresponding extraction zone in the document. This was particularly problematic for multi-page documents, where it was unclear which page a field had been extracted from. (613065)
• Fixed an issue where opening a multi-page document in Manual Review or the skill editor could cause the page to go blank when scrolling to the second page, making the document unviewable. (590977)

​

OCR

• Fixed an issue where OCR recognition accuracy was significantly reduced for documents containing large monospaced fonts, such as airway bills. Characters were frequently misread or not recognized at all. (494673)
• Fixed an issue where Chinese characters were missing from OCR output when processing documents that contained mixed Chinese and other language content. Affected characters were absent from both the extracted text and the searchable PDF output. (501997)
• Fixed issues with OCR recognition accuracy on Thai documents where recognized text contained incorrect characters and extraneous spaces inserted between syllables and words. (540457)
• Fixed an issue where table cell values were incorrectly merged with adjacent column values in OCR output. This caused multi-word text spanning table separator rows to be combined into a single cell rather than recognized as separate values. (582286)
• Fixed an issue where processing a document with an OCR skill and exporting to Word (editable) could result in tables being collapsed or rendered incorrectly in the output file. (583580)
• Fixed an issue where rotated or side-oriented text on a page was not detected by the OCR skill when the Handwritten text option was disabled. (583808)
• Fixed an issue where specific characters and numbers were intermittently missing from OCR output when processing PDF documents created with macOS applications. (584604)
• Fixed an issue where OCR failed to consistently recognize table column headers and dash characters (”-”) used as cell values in certain document layouts. Affected columns were systematically missing from the extracted output. (584729)
• Fixed an issue where text in adjacent table columns was incorrectly detected as a single text object when column values were printed in close proximity, even when a vertical separator was present between the columns. This caused incorrect field extraction results and affected training accuracy. (585161)
• Fixed an issue where empty rows at the end of a table were omitted from OCR output, causing the extracted table structure to differ from the source document. (589173)
• Fixed an issue where enabling both English and Arabic languages in a document skill’s settings caused digits to be recognized incorrectly, returning garbled characters instead of the expected numeric values. (603706)
• Fixed an issue where processing a document with an OCR skill configured to use French with handwritten text recognition enabled could fail with an internal processing error, causing the document to be rejected instead of processed successfully. (604916)

​

Process Skill

• Transaction updates with very large document counts are now more reliable under heavy processing loads. (638555, 658752)
• Fixed an issue where a Document skill remained visible in Process skill settings after being removed from all activities. This occurred when the associated class was deleted from the Classification skill before the Document skill was removed from the Assembly activity, which caused the Assembly activity cache to retain a stale reference. The Document skill is now correctly removed from Process skill settings once all references have been deleted. (609737)
• Fixed an issue where document images were incorrectly cropped when processed through a Process skill, causing fields to be partially or fully outside the visible area in Manual Review. Images processed directly through a Document skill were not affected. (549978)
• Fixed an issue where processing a document containing a long table could cause the report worker to run out of memory and hang, resulting in the document failing to process. (579890)

​

User Management

• Corrected the role description shown when assigning the Skill User role in the Users tab. It now reads “Can process documents using existing skills and view the skills in the Skill Catalog” (previously referenced “Skill Monitor”, which Skill Users cannot access). This change applies to Vantage 3.0 only. (661123)

​

Resolved Known Issue in 3.0.4

• Removed Azure OpenAI option from the Configuration > Connections page.