Release Notes

• OAuth Connections: The system is now more robust in how it retrieves authentication tokens, improving the reliability of OAuth-based connections.

• Google Services Authentication: Resolved an authentication error that prevented users from connecting to Google services, such as Gmail. This was caused by a duplicate parameter being sent during the OAuth process, which resulted in a connection failure.

• Process Execution: Resolved an issue that prevented some processes from starting. The fix corrcts how the system interprets phrases like "the procedure" and "the organization".

• Dashboard Reports: Resolved an issue in dashboard report generation that caused data duplication in exported reports.

• Guidance Center: Resolved an issue where the code editor in the Guidance Center would display empty tabs when switching between different exceptions.

• Learning Suggestions: Resolved an issue where a learned answer from one context (e.g., a specific field in an "Extract Data" step) was incorrectly being suggested in an unrelated context.

• Null Characters in Files: Resolved an issue that caused automations to crash when trying to process files (e.g., CSVs) containing null characters.

• OCR Control Characters: Resolved an issue where special control characters in text from OCR could cause the system to crash. This fix prevents these crashes, ensuring automations no longer get stuck in an endless "processing" state.

• Extracting Tables: The extract a table procedure now supports a file_upload mode for PDF documents, allowing you to send a file directly to an LLM for significantly improved performance.

• Playground Responsiveness: We've optimized the Playground editor by improving data handling and limiting undo/redo history size, resulting in a more responsive experience.

• API Key Deletion: Resolved an issue preventing the deletion of an API key when its associated agent no longer exists.

• File Naming: Resolved an issue where files with a hash symbol (#) in their filename could not be copied, causing "File not found" errors.

• SQL Queries: Resolved an issue where SQL queries using variables would fail due to invisible non-breaking space characters. These characters are now automatically normalized.

• Starred Data: Resolved a pagination issue with searching starred data. Search filters are now correctly applied when scrolling.

• Run Details: Resolved an issue in the run report that caused duplicate rows when procedures had multiple versions.

REST API Details

• You can now toggle between v1 and v2 (default) when clicking Copy API Endpoint for a given process, enabling easy access to both REST API endpoint versions.

Data Processing

• We've resolved an issue that could cause data integrity problems during database operations. The fix ensures accurate and consistent data handling across all transactions.

• File Extensions: Downloaded documents now preserve their original file extension instead of always being converted to .pdf.

• Performance Dashboard: Contract Details now shows a loading indicator while fetching data and is hidden for users without view permissions.

• Dashboard Timeout: Resolved an issue where dashboard details would time out for large organizations. Query performance has been optimized to load in seconds.

• Contract Details Display: Resolved an issue where Contract Details would appear blank on data load failures. The section now displays an error message with a retry option when this happens.

• Organization Ownership Transfer: Resolved an issue preventing ownership transfers to users in other agents. User selection now includes all organization members.

• Table Extraction with Gemini: Resolved failures when extracting tables using Gemini models. Table extraction now works reliably with Gemini.

• Table Row Insertion: Resolved workflow failures when inserting rows containing null or invalid numeric values. Table operations now handle empty or invalid data gracefully.

• File Conversions: Resolved conversion failures caused by a missing dependency.

• Document Processing: Optimized file processing for LLM analysis, reducing token consumption and enabling successful processing of larger documents that previously exceeded token limits.

• Editor Performance: Improved editor performance and stability when working with large processes, including fixed cursor jumping, enhanced responsiveness, and smoother scrolling.

• Error Handling: Resolved an issue where encountering an unknown procedure caused infinite learning loops or duplicate requests.

• Learnings Display: Resolved an error that occurred when viewing learnings associated with a draft process during the publishing workflow.

• PDF Conversion: Resolved a runtime error that prevented successful conversion of documents (e.g., .docx, .xlsx) to PDF format.

• Copy API Endpoint: Resolved an issue where the Copy API Endpoint screen displayed the incorrect agent_id.

• Connections Page: Resolved an issue that caused the Connections page to crash when a connection to an integration was present after the integration itself had been uninstalled.

• Performance Dashboard: Resolved an issue that occurred when saving contract configurations. The character limit for the "Contract details" field has also been increased from 200 to 1,000 characters.

• Publishing Processes: Resolved an issue where new processes and their learnings were not displayed in the publish summary when publishing for the first time.

• System Stability: Resolved a dependency compatibility issue to prevent potential application crashes and ensure system stability.

• Browser Automation: Browser automation now supports file uploads and downloads. We've also added a new procedure to close a browser.

• Table Extraction: Table extraction is now more accurate and reliable with a new strict mode and two-pass approach for improved header and value accuracy.

• Asking for User Input: Resolved an issue where automations would incorrectly pause to ask for user input instead of automatically retrying after certain transient errors, such as network failures.

• Providing Values: Resolved an issue where providing a value for one missing field in a document would incorrectly apply the same value to all other missing fields in the same request.

• Table Extraction: Resolved an issue causing table extraction failures for certain documents due to malformed data and complex structures.

• Browser Automation: We've reduced the width of the Execution Steps panel to prevent overlap with the browser automation window. We've also improved the "Connecting to Browser" status messages for clearer loading and reconnection feedback.

• Internal Logging: Our internal logging has been enhanced to better diagnose issues with triggered runs and improve our troubleshooting capabilities.

• Excel Files: Resolved an issue that caused Excel files (.xlsx) to become corrupted after being generated or updated in an automation.

• Markdown Widgets: Resolved an issue with adding multiple markdown widgets to an automation.

• Table Extraction: Resolved an issue that caused table extractions to fail when missing values were provided for a table, particularly when using the Select from Document option.

• Browser Use: Improved instructions for getting started with browser automation.

• Exception Routing: Documented an existing process configuration setting that allows you to set up assignment and notification rules for exceptions.

• Improved Sorting for All Books: All Books now displays installed books first, making it easier for users to locate and manage them.

• Connections: Removed the non-functional "Reconnect" option for OAuth Client Credentials connections, as this action is not currently supported.

• Extract Data: Resolved an issue where extract data would trigger the Guidance Center for fields with default values.

• SharePoint: Resolved an issue that prevented users from creating tables from SharePoint lists using retrieve the slist as a table.

• Table Extraction: Resolved an issue with table extraction where headers specified in a prompt were being incorrectly parsed and merged.

• Tables (v2) – Row Retrieval: You can now use get the table's rows to retrieve all rows (previously limited to the first 1,000).

• Tables (v2) – Conversion Options: When converting a table to Excel, you can now set a custom filename and worksheet name.

• Automation Editor: We've optimized the rendering of several editor components, resulting in smoother scrolling and more responsive interactions.

• Book Logos: Resolved an issue where the logos for the Airtable and Windows Legacy books were incorrectly displayed.

• Parallel Processes: Resolved an issue where some completed runs in a parallel process would incorrectly display as In Progress after the page was refreshed.

• Process Scheduling: Resolved an issue with scheduling a process to run on the "Last day of the month".

• Table Creation from Singular Row: Resolved an error that occurred when creating a table from a single row of data.

• API Key Usage: You can now see your API usage and limits directly in the API Keys page (found under Settings → Admin & Preferences).

* **Enhanced Excel Table Extraction**: [**Table extraction**](https://docs.kognitos.com/excel/tables/extract-a-table-from-a-worksheet) from large Excel worksheets is now significantly faster with support for specifying a table's exact location. * **Custom Blank Values in Table Operations**: You can now specify **custom blank values** when [**casting**](https://docs.kognitos.com/tables/columns/cast-a-table-column) or [**rounding**](https://docs.kognitos.com/tables/columns/round-a-columns-values) table columns.

• Empty Table Cells: Resolved an issue where numeric columns with empty cells were incorrectly converted to a decimal format.

• Submitting Guidance: Resolved an issue in the Guidance Center regarding the Submit button being hidden.

• Table Extraction: Resolved an issue where table extraction would fail if the headers specified in a prompt did not match the headers found in the document.

• Agent Exports: Agent export file names now include a timestamp to improve file organization.

• Numeric Conversions: We've enhanced formula calculations to automatically convert numbers formatted as text into numeric types.

• User Preferences: The Enable Document Plan toggle has been removed from Admin Preferences to simplify the user interface.

• Loop Iterations: Resolved a bug where loops and conditional steps in a process run were incorrectly showing a View Runs link.

• Process Renaming: Resolved an issue where the Run button was disabled after a process was renamed and saved with leading or trailing spaces.

• Table Data: Resolved an issue where formulas and conditions in tables failed to correctly handle empty or null date and time values.

• Improved Table Extraction: Table extraction is now more reliable thanks to stricter data validation and smarter internal prompts. When extraction fails, users will see clearer, more descriptive error messages.

• Numeric Operations: Resolved an issue when performing numeric operations, such as casting or summing, on numbers formatted with commas.

• Table Extraction: Resolved an issue where table extraction would fail intermittently or result in misaligned table headers and rows.

• Expanded Color Palette for Charts and Plots: You can create charts and plots with an expanded color palette. All HTML named colors are now supported.

• Document Processing: Resolved an issue where processing documents with unsupported file types resulted in a "job failed" error. The system now validates the file type and returns a clearer error message.

• Tables with Special Characters: Resolved an issue that caused automation runs to crash when processing tables with special characters (such as newlines) in the column headers.

• Ask Koncierge: Resolved an issue where ask koncierge would fail if the LLM returned an unexpected response format.

• Multiple Choice Questions: Resolved an issue where multiple-choice question dropdowns remained active after submitting an answer.

• Run Parallelization: Resolved an issue where parent processes failed to resume when child runs were archived.

• Table Filtering: Resolved an issue with filtering multi-word column headers and complex Excel file structures.

• Updating Table Rows: Resolved an issue that caused processes to fail when updating table rows with specific numeric data.

• Guidance Center: The Guidance Center now displays the number of pending questions for each group. A new Stop Generation control is also now available.

• Table Extraction: Extracting tables from documents is now more reliable. When Kognitos detects uncertainty in extracted table data, it will prompt you for guidance.

• Page Extraction: Resolved an issue where a request to extract a single page from a document would incorrectly process the entire document.

• Playground Results: Resolved an issue where results for a completed process run were sometimes missing in the Playground.

• Table Extraction: Resolved an issue where extract a table caused an error when a table could not be found in a document.

• Our documentation has been updated with all current BDK Books — offering complete coverage for 40+ integrations.

• Run Parallelization Improvements: We've made updates to run parallelization. After waiting for child processes, the parent process is automatically notified of which child runs have completed and continues without requiring manual action.

• Enhanced Table Querying: Users can now perform SQL queries that join two tables together using the query a table procedure.

• Adding Collaborators: Resolved an issue where adding a collaborator would fail if their email address was entered with different capitalization.

• Document Previews: Resolved an issue where the document preview for a starred data element would fail to load if it was closed and then reopened.

• Extract Data: Resolved a caching issue in Playgrounds where modifying and re-running an extract data step could incorrectly use results from a previous extraction.

• Process Creation: Resolved an issue with saving empty processes. The Save button is now disabled when a process is empty.

• Process Visualization: Resolved an issue with process visualizations within demo agents.

• Abort Long-Running Guidance: Users can now abort long-running guidance computations that are taking long to process.

• Extended Character Limits: The character limit for a single learning has increased from 10,000 to 15,000.

• Process Widget Search: The process widget dropdown now includes a search bar.

• Agent Imports: Resolved an issue where large agent imports caused a timeout error. We've also improved the import process to be significantly faster.

• API Key Creation: Resolved an issue where creating multiple API keys simultaneously could fail.

• Department Box Search: Resolved an issue where the department box search failed to find items.

• Editor Performance: Resolved performance issues in the editor that caused lagging, unresponsiveness, and unexpected cursor jumps.

• Process Creation: Resolved an issue where the Create Process button in the Playground was disabled if the first line in the automation was empty.

• REST API (v1): Resolved an issue where invalid requests to the legacy REST API caused internal server errors.

• Updated Terminology: The Email Trigger Configuration dialog now uses clearer, more consistent terms.

• Improved Inline Results: Long text results are now truncated with an ellipsis instead of being hidden.

• Connection Timestamps: The system now tracks when a connection was last updated.

• Table Access: Resolved an issue that prevented some tables from being viewed or downloaded.

• Excel Sheets: Resolved an issue where processing an Excel file with only one worksheet would cause an error.

• Column Name Handling: Resolved an issue where AI-generated queries would fail when interacting with table columns containing spaces in their names.

• Bulk Database Inserts: Resolved an issue that caused bulk database inserts to fail.

• Timestamp Accuracy: Resolved an issue where the "Last Updated" timestamp for draft processes on the process list page was incorrect.

• Faster Process Search: Improved speed and responsiveness for searching and filtering on the Processes page.

• Subprocess Navigation: When viewing a process in the editor, users can now hover over a subprocess widget to see a View Process option.

• Outlook Email Support: We've added support for uploading email files with the .msg format.

• Process Visualization: Fixed a bug that prevented process visuals from loading correctly.

• Activation Flow: Resolved an issue where the confirmation message wouldn't close after activating or deactivating a process.

• Email Trigger Permissions: New processes will default to stricter email trigger settings, allowing only agent members to trigger processes by email.

• Fixed Run Output Display: Resolved an issue where run output wasn't displayed for certain processes.

• Improved Schedule Reliability: Resolved an issue where creating or editing a process schedule would cause the selected time of day to be ignored upon saving.

• Enhanced Department Cleaning: Improved handling of rapid department box cleaning operations to prevent version conflict errors.

• Preserved OAuth Connections: Fixed an issue where OAuth application connections would break when publishing a process.

• Maintained Credential Formatting: Resolved automatic capitalization of credential names.

• You can now filter an agent's procedures to show only those that have been run.

• Missing credentials for a learned Book can now be entered directly in the Questions panel.

• CSV previews now support smoother scrolling and faster performance on large files.

• Flowcharts have been enhanced with larger fonts and improved line spacing for better readability.

• Resolved an issue on the Runs page where the procedure name was displayed incorrectly.

• Resolved an issue where the Playground editor and visualization panels did not resize correctly.

• Resolved an issue where inserting a new, empty column into an Excel worksheet would fail to apply the specified column name.

• Faster Load Times in the Guidance Center: We've optimized how data is fetched and rendered in the Guidance Center, making it 4x faster.

• Improved Department Dropdown Responsiveness: The department dropdown is now much more responsive.

• Broken Links: We've corrected a system URL issue that was creating broken internal links.

• Editor Stability: We've resolved an issue that caused an error in the Playground editor header.

• Added support for the process run widget when starting parallel runs.

• Enabled bulk management of department members.

• Implemented automatic email notifications for new department users.

• Improved handling of non-existent table columns.

• Enhanced OAuth authentication for BDK Books.

• Tables (v2): Resolved an issue where empty column names caused an infinite loop.

• Tables (v2): Resolved an issue where automations got stuck when accessing a column that didn't exist in a row.

• Processes: Resolved an issue that caused processes to get stuck in a "preparing worker" state.

• Processes: Resolved an internal issue related to the run archiving process.

• Files & Documents: Resolved an issue with filling in .docx templates that included tables.

• Files & Documents: Resolved an issue with extracting pages from encrypted documents.

• Files & Documents: Resolved an issue where non-English characters were not displaying correctly in JSON attachments.

• Tables (v2): Resolved an issue that resulted in a "Could not get a table's column's values" error.

• Processes: Resolved an issue affecting scrolling behavior on the Processes page.

• Processes: Resolved an issue where processes did not load correctly after searching.

• Automation Stability: Resolved an issue that caused automations to stop unexpectedly, particularly when handling large files.

• First Edition Books: The First Edition Outlook, SharePoint, GitHub, and ServiceNow Books will no longer be available for learning. Instead, use the newer BDK version of these Books.

• Run Parallelization: We've updated the syntax for retrieving information from child runs.

• Visual Enhancements to Questions: We've enhanced the Questions tab with improved visuals for displaying questions, hints, and errors.

• Tables (v2): Resolved an issue where get a table's row resulted in an index error.

• Tables (v2): Resolved an issue where the First Edition Database Book could not read data created or formatted with Table 2.

• Mini-Playground: Resolved an issue where mini-playgrounds entered an infinite creation loop.

• Learnings: Resolved an issue where some learnings weren't loading due to a pagination error.

• Tables & Excel (v2): Resolved an issue where retrieving an Excel worksheet's cell resulted in an out-of-bounds error.

• Tables & Excel (v2): Resolved an issue where empty values were not preserved when converting a CSV to a table.

• Data Extraction: Resolved several issues with data extraction when using Gemini models.

• Exception Handling: Resolved an issue where the system threw an unknown exception on a random loop iteration.

• Exception Handling: Resolved an issue where a run crashed and users were repeatedly prompted to answer the same exception.

• Platform Stability: Fixed an issue where not all processes were loading on the Processes page when scrolling.

• Platform Stability: Resolved an issue where files passed from a parent process to parallel runs were not uploading correctly to SharePoint.

• First Edition Books: First Edition HTTP, Slack, and S3 Books will no longer be available for learning. Instead, use the newer BDK version of these Books.

• Playgrounds: The Restart button is now hidden for new Playgrounds until they've been run at least once.

• Table & Excel Enhancements: Table and Excel procedures (v2) have been optimized for table extraction and overall performance.

• Document Viewer: We've improved the experience for viewing and navigating large documents in the Document Viewer.

• Markdown Support: We have added support for handling Markdown content in the Microsoft Outlook Book (First Edition).

• Data & Document Processing: Resolved a bug where Select from Document was not extracting text properly.

• Data & Document Processing: Resolved a "No space left on device" error when document processing with LLMs.

• Data & Document Processing: Resolved an issue resulting in an UnidentifiedImageError when working with image files.

• Playgrounds & Processes: Resolved an issue where the Runs page was not loading.

• Playgrounds & Processes: Fixed a bug where a process incorrectly appeared as "Paused" despite completing successfully.

• Playgrounds & Processes: Fixed a bug where subprocess results were not visible.

• Playgrounds & Processes: Fixed a bug with HAL where the "Apply" button was not functioning as expected.

• Excel: Addressed an issue with the First Edition Excel Book where a cell's value could not be retrieved.