New features, improvements, and fixes from each release of e! by Lexemo.
These updates have completed testing and will be deployed to production in the next release.
These updates have completed testing and will be deployed to production in the next release.
Fixed a critical issue where workflow steps that call an external API with an optional file attachment would fail the entire run if no file was provided. These steps now complete normally whether or not a file is included.
Fixed an issue where account administrators could not see bots that other administrators had added to a shared user group. Bot access for user groups is now visible and manageable by any administrator in the company.
Fixed a display issue in multi-section forms where navigating past certain sections could briefly clear the progress indicator and section highlights. Your position and progress now stay accurate as you move through the form.
Fixed an issue in the variable mapper where declining the confirmation for “Remove unused” would still delete the unused variables. Cancelling now correctly leaves all variables in place.
These updates have completed testing and will be deployed to production in the next release.
API Connector now supports multipart/form-data requests, including file fields. You can connect to APIs that require file uploads or a mix of structured form fields and files.
Runs now save automatically. Your progress is preserved about once a minute, so you no longer lose work if your browser closes or a session ends unexpectedly. Resuming an auto-saved run works exactly like resuming one you saved yourself.
AutoMate file uploads are more reliable. Supported file types (txt, PDF, JPG, JPEG, PNG) are accepted, unsupported files now show a clear error instead of being silently dropped, and you can attach up to 20 files in one go. New per-user and per-conversation upload limits keep usage in check, with clear feedback when a limit is reached.
Clearer wording when editing in AutoMate. Editing prompts now consistently use the term “modify” throughout for a more predictable experience.
The Run Debugger window is easier to reposition. You can now drag it from anywhere along its title bar, and it stays visible after you move it.
Strengthened security around uploaded image files.
Fixed an issue where the SharePoint connection would stop working after a period of time. It now refreshes in the background and stays connected without needing to be set up again.
Fixed an issue in the Report view where reloading a previous run and resending it without changes could send incorrect values (such as 0) for number fields. Reloaded values are now preserved on resend.
Fixed an issue where the CSV multi-process preview only showed the first 10 rows. You can now see all runs both before and after starting.
Fixed an issue where statistics and data reports could fail to load or time out for bots with very large run histories. These reports now load reliably.
Fixed an issue in AutoMate where the full bot was not automatically included as context after new nodes were added. Follow-up prompts now correctly apply to the current bot.
These updates have completed testing and will be deployed to production in the next release.
These updates have completed testing and will be deployed to production in the next release.
You can now trigger multiple runs at once by uploading a CSV file. Each row in the file starts a separate run, with each column mapped to an input field, so you no longer need individual files to process work in bulk.
Working with large workflows is now much easier. A new multi-node mode lets you keep several nodes open at the same time, toggled via the Open Nodes panel or the Ctrl/Cmd + M shortcut. Holding Ctrl/Cmd while clicking a node still opens it alongside the others as a quick shortcut.
Logic Condition nodes now have an expand/collapse toggle that opens or closes all of their nested steps in one click, instead of opening each child node individually.
Agents can now use your configured API connectors as tools, allowing them to fetch and act on data from connected systems during a conversation. Note: this is the first stage of the capability, so expect it to expand in upcoming releases.
Added authentication support for the advoware integration, enabling secure connections to the advoware practice management system.
Fixed an issue where the source document viewer failed to load PDFs, and where clicking a source passage did not highlight the matching text in the PDF, particularly in documents containing umlauts or special characters.
Fixed an issue where agent execution logs were unavailable when a knowledge base search failed during the run. Logs now remain accessible so you can see what happened.
Fixed an issue where the agent tool selection listed tools from other organizations. You now only see your own company’s tools.
Fixed an error that prevented the User Details window from opening in Company Members settings.
Fixed an issue where the search bar on the See Data page loaded endlessly and never became usable.
Fixed an issue where file and image uploads failed when processed with certain AI models.
Fixed an issue where background colors of table cells were removed from rich text content.
Fixed an issue where document generation runs could fail due to font styling in rich text content. Note: custom font-family styling is no longer carried over into generated documents; standard formatting is unaffected.
Fixed an issue where logic conditions could keep an invalid order until they were opened, and where conditions opened at the same time received the same position.
The clause preview now opens in a centered, read-only window instead of taking over the full screen, and no longer shows an unnecessary confirmation prompt.
Fixed a layout issue in the Branding tab where the allowed file types text could overlap the upload label on narrow screens.
This update has completed testing and will be deployed to production in tonight’s release.
These updates have completed testing and will be deployed to production in the next release.
Fixed an issue where custom Azure OpenAI portals were using the default Azure endpoint instead of the endpoint configured in the backend, which caused AI output nodes to fail with a permission error.
Fixed an issue where adding or removing tags on a clause did not update the clause card until the page was refreshed. Tag changes now appear immediately after saving.
Fixed an issue where unpublishing a bot did not update its status in the bot overview. The published indicator now reflects the change right away.
Fixed an issue that prevented the login screen from rendering correctly in some configurations.
Fixed an issue in the text editor where, when two editors were placed in the same section, the fullscreen action always opened the second one regardless of which was clicked.
Fixed an issue where saving a standard agent required two clicks before changes were properly persisted and the unsaved changes prompt would no longer appear.
Improved how pasted content from Word is preserved in the consumer frontend, so tables and formatted text retain more of their original styling.
These updates have completed testing and will be deployed to production in the next release.
Bot templates have a refreshed design that makes browsing and selecting templates easier and more visually appealing.
Newly created bots now start in pipeline design mode by default, giving you the modern layout right from the beginning. Existing bots are not affected.
Fixed an issue where the variable mapper would fail when an external service returned a time value that included seconds. Times are now handled cleanly without needing manual workarounds.
Fixed an issue where logic conditions using clause input could lead a run into an unexpected path or appear to do nothing. Clause input is now validated correctly and the chosen path is reached as configured.
Resolved an error that could cause the AI engine to fail while processing tool results in certain edge cases, leading to a failed run in the bot.
These updates have completed testing and will be deployed to production in the next release.
The AI configuration area has been renamed from “AI Settings” to “AI Agent” throughout the interface for clearer and more consistent labeling.
The AI models available on Microsoft Azure have been updated to GPT-5.4 and GPT-5.4-mini, replacing the previous versions.
Images generated by AutoMate are now displayed properly as rendered images in the chat, rather than as raw inline references.
Text formatting in AutoMate responses now uses consistent, correctly sized fonts for headings, paragraphs, and lists.
Fixed a security issue where logging out did not fully delete the session. Logout now properly ends the session, and a fresh session is created on next login.
Fixed an issue where text field nodes placed after a Send & Result node were not correctly displaying the data retrieved from API calls — the fields now update as expected.
Fixed an issue where the chat window inside bot nodes could grow indefinitely. The window now has a fixed size and is scrollable.
Fixed an issue where selecting the “dashboard” display option in a bot card could not be saved. This option is now fully supported and persists correctly.
Fixed a false error in the AI Agent debugger log that incorrectly flagged successful tool executions (such as sending emails) as failed. The log now correctly reflects the actual outcome.
Fixed a critical issue where copying workflow nodes caused the copied nodes to share internal reference IDs with the originals. This was leading to data mixups at runtime where copied steps were reading output from the original step instead of their own. Duplicated nodes now always get fresh, unique references.
Fixed two AutoMate issues: prompts containing quotation marks no longer cause a failure, and edited plans are now correctly used when generating the workflow instead of falling back to the original version. Note: document support in AutoMate currently works reliably with PDF files.
Fixed an issue where tables rendered by AutoMate would overflow outside the chat window with no way to scroll. Tables are now correctly contained and scrollable.
These updates have completed testing and will be deployed to production in the next release.
Pro Agent Tools: AI Agents can now call external APIs as tools during a workflow step. You define the API connection and describe the tool in natural language in the dashboard. The agent then decides when to use it based on the workflow context. Tools are scoped per agent and per workflow step. This allows agents to write data into external systems, trigger actions, or retrieve information without manual intervention between steps.
New frontend designs: 6 new layout designs are available for the consumer facing bot frontend. You can preview and switch between them in the bot settings. The selected design applies immediately to your live bot.
You can now click on AI generated images in the chat to open them in a full screen view. From there you can download the image directly, making it much easier to save and reuse generated visuals.
Images generated in AutoMate chat now display as proper standalone visuals instead of being rendered as inline markdown. This makes image responses cleaner and easier to view.
Text sizes in chat responses are now consistent across paragraphs, lists, and headings. Previously, some elements appeared larger or smaller than intended, which could make AI responses harder to read.
These updates have completed testing and will be deployed to production in the next release.
You can now download PDF and SVG documents directly from the Transparenzregister integration, making it easier to save and share registry records.
The Transparenzregister integration now connects to the official production environment, replacing the previous test endpoint.
Fixed an issue where unpublishing a clause would fail with a validation error, even when the clause was not used anywhere.
Subscript and superscript formatting now renders correctly in generated Word documents and PDFs. Previously, these were stripped or displayed as normal text in exported files.
These updates have completed testing and will be deployed to production in the next release tonight.
The Transparenzregister integration is now fully operational. You can authenticate and submit requests to the Transparenzregister directly from the application, with automatic token management running in the background.
When you use AutoMate to add or edit nodes in your workflow, the canvas now automatically scrolls to the first node that was changed or added, so you always land in the right place after an AI edit.
AutoMate now automatically saves your bot whenever it creates or edits nodes. You no longer need to manually save after making AI-assisted changes to your workflow.
The action menus on workflow nodes are now easier to read. The background behind those controls was too transparent and could make the icons hard to distinguish — this has been corrected with improved opacity.
Passwords must now be at least 10 characters long when signing up or changing your password, in line with updated security policy. If you try to set a shorter password you will see a clear validation message. Existing passwords do not need to be changed.
Fixed a reliability issue with AutoMate’s node editing where nodes were sometimes incorrectly deleted, replaced instead of added, or duplicated due to ID conflicts. AutoMate should now apply changes accurately and predictably.
Fixed a bug where asking AutoMate to delete a logic condition left the Bot in an invalid state. Only the child step was being removed, leaving the parent disconnected from the rest of the flow.
We updated the underlying text editor to the latest version, which may resolve an issue where applying formatting to a large block of content could cause an error. If you have been experiencing this, please let us know whether it still occurs.
Fixed two issues with annotation widgets: they now render correctly when previewing an annotation node, and the variable picker no longer opens behind the annotation editor when you try to reference another node.
Fixed an issue where the PDF viewer inside the source node modal was not rendering the document. The modal was showing only raw extracted text chunks instead of the actual PDF.
Fixed an issue where the open nodes counter in the floating toolbar did not update in real time after deleting an open node from the canvas. The count now reflects the correct number immediately.
No release notes for this month
Pick another month or view all updates.