September '25 Release Notes

New Features

• Plugin System for Apps: Implemented a plugin system and published a library of UI components, allowing developers to extend the platform's functionality with custom AI apps. 
• Apps Portal: Created a new portal and landing page for authenticated users to access any number of Vertesia, third-party, or custom apps.
• AI Assistant App: Introduced the AI Assistant app, a solution for individuals, teams, or the entire organization to securely interact with AI agents that have knowledge and context to autonomously complete your business-specific tasks.
• Review Center App (beta): Introduced the Review Center app, a comprehensive solution fit for a broad range of use cases focused on document review. This app provides a unified workspace for organizing documents, creating and managing checklists, and interacting with AI-powered agents. 
• Group Management: Added support for managing users as groups, simplifying permission management and user access to Vertesia projects and applications.
• SCIM 2.0 Support: Users and groups can now be externally provisioned through identity and access management services like Microsoft Entra and Okta, centralizing and simplifying user and group management for large organizations.
• Plugin System for Custom Agent Tools: Introduced support for tool plugins, allowing developers to create and expose custom tools for agents. This expands the capabilities of the agent workflow system and enables integration with external services and resources.
• Remote MCP Tools: Added an MCP client to access tools hosted on remote MCP tool servers. This enables agent access to a virtually unlimited number of MCP tools offered by third-party software vendors.
• Object and Collection Security: Granular permissions can now be assigned to users and groups on collections and their contents, providing greater flexibility and control over managing user access in Vertesia projects.
• Azure AI Foundry Support: Added Azure AI Foundry as an inference provider, including support for model families from OpenAI, Grok, Deepseek, Mistral, and Microsoft. This replaces the now deprecated support for the older Azure OpenAI.
• Dynamic Rate Limiting: Introduced a new rate-limiting mechanism that considers both the environment and the specific model being used. This enhancement optimizes execution of AI workloads based on the available capacity of models. 

Improvements

• Hybrid Search: Combined similarity (vector) and full-text searches into a single interface, adding flexibility to the search and improving search results for users and AI agents.
• Object Page Enhancements: The content object detail page has been improved with resizable panels and a panel switcher for left and right sections. This allows users to customize their view of properties, JSON, text, and images, enhancing usability.
• Studio Settings Management: Organization and project settings are grouped and displayed based on the user's respective organization-level and project-level admin access, making administration more intuitive. 
• Role and Permission Updates: Introduced the new `consumer` role designed for application end users, an `app_member` role for user-app management, and a `content_superadmin` role for managing collections and objects. Refined permission names for consistency, and added new permissions and permission checks across the API. 
• Bulk Add Users to Multiple Projects: A new bulk action has been added to the Organization settings under the Users tab. This allows administrators to select multiple users and add them to several projects simultaneously, streamlining user management for organizations with numerous projects.
• Improved Table of Content Generation Workflow: The generation of tables of content now utilizes a more advanced workflow for splitting text into chunks, improving the accuracy and detail of document analysis.
• System Worker for Temporal Tasks: A dedicated system worker has been implemented within Temporal to prioritize system-related tasks. This ensures that critical background processes, such as rate limiting and internal routing, are not delayed by other ongoing tasks, improving the responsiveness and reliability of core functionalities.
  
• Updated Onboarding Models: The default models for new projects have been updated. We now default to Claude 4 on both AWS Bedrock and Google Vertex AI, and have updated the available Gemini models to the newer 2.5 series.
  
• Enhanced Runs Page UI and Filtering: The Runs page has been updated to display the Workflow Run ID for each execution. We've also added new filter options for both Run ID and Workflow Run ID, making it easier to find specific runs.

Bug Fixes

• Renaming Organization: Fixed a bug that prevented changes to an organization's name from being saved to the database. The system now correctly persists organization name updates.
• ID and Name Filtering for Content Objects: Fixed an issue where ID and name filtering for content objects was not working correctly. The filters are now split in the UI and handle both ID and name filtering