Skip to content
  • There are no suggestions because the search field is empty.

November '25 Release Notes

Product release notes for November 2025

New Features

  • Standard Video Intake Workflow: A new system workflow and rule, "Standard Video Intake", is available to process video files. This workflow extracts metadata, generates web-compatible video and audio renditions, creates poster and thumbnail images, transcribes audio content, and summarizes videos.
  • Hierarchical Collections: You can now create nested collections by adding one or more collections as children to a parent collection. This enables more complex and organized content structures.
  • Conversation Analysis Tools: Now AI agents can search past agent conversations and perform AI-powered analysis on them using the new built-in tools: "search_conversations" and "analyze_conversation".

Improvements

  • Agent Runner Redesign: The Agent Runner has been redesigned with a new landing page that lists all active agent workflows, offering a more consistent experience. You can now initiate new agents from a modal and navigate to a dedicated detail page for each agent.
  • Agent Runner Topic: A new "Topic" field has been added to conversation workflows to provide a brief summary of the conversation displayed in Agent Runner.
  • Run Detail Page: We've introduced a dedicated page for viewing detailed information about a specific run. This provides a larger view of run data, in addition to the side panel view.
  • Standardized Table Navigation: We've unified the user experience across all data tables. Clicking on a row now consistently navigates to the detail page, while a new eye icon allows you to preview items like Runs and Content Objects in a side panel. Checkboxes are now only visible on hover or selection to reduce visual clutter. 
  • Dynamic Breadcrumb Navigation: The UI navigation has been enhanced with a dynamic breadcrumb system that works across the side menu navigation.
  • File Replacement Workflow: When replacing a file in a content object, the dialog now defaults to "Create new version".
  • Workflow Configuration Templates: We've added configuration templates to quickly configure common workflow rules like Semantic DocPrep on document intake and the model for video analysis on video intake.
  • Search Documents Tool: The tool is now more robust with improved error handling and clearer documentation. 
  • Fetch Document Tool: Agents can now focus on relevant sections of large documents without processing the entire file by using a page_range parameter to retrieve content from specific pages or a range of pages within a document. 
  • PDF Table of Contents Generation: The workflow for generating a table of contents from PDFs has been significantly overhauled, significantly improving performance and reliability processing large documents. 
  • Optimized Embeddings Status Endpoint: The performance of the embeddings status endpoint has been significantly improved. 
  • Error Fallback Page: A robust error fallback page has been implemented for handling critical loading errors, such as network issues or CDN failures.

Bug Fixes

  • Type Facet Counts in Object Search: Fixed a bug where facet counts for content object types were limited to 10,000. The counts now accurately reflect the total number of objects for each type. 
  • Infinite Scroll for Content Objects: Resolved an issue where the "load more" functionality would sometimes fail to trigger when scrolling through the content objects list. 
  • Text Display in Content Overview: To prevent browser freezing with very large text files, the content overview panel now displays a preview of the first 128KB of text and shows a warning when the content is truncated. 
  • Multi-file Upload and Progress Display: Fixed a bug that prevented uploading multiple files at once via the "Upload" button and displaying the progress.
  • Collection Metadata: Resolved an issue where the Collection Metadata properties editor was not correctly saving changes for array-based properties.
  • Model Loading in Playground: Fixed a bug where the Playground would not correctly load the default model from an Interaction's environment.
  • Blank Responses from Gemini in PDF Transcription: Addressed an issue where Gemini models could return empty responses during PDF transcription. 
  • Private Conversation Visibility: Private agent conversations are now clearly marked with a lock icon in the Workflow Executions list for users who do not have access. 
  • Graceful Shutdown for Max Iteration Errors: Corrected a failure condition in conversation workflows where reaching the maximum iteration limit with pending tool calls would cause an API error. 

Deprecated

  • Output Modality Configuration: We have removed the need to manually specify the "Output Modality" (e.g., text, image) in Interaction configurations. The platform now automatically detects the output type based on the model's response, making it easier to work with multimodal models.