September '26 Release Notes

New Features and Models

• App Builder: We’ve brought coding agents together with our new app development and hosting infrastructure to build, deploy, and manage AI-powered apps. Create a new app and connect your favorite IDE and coding agent, or use Vertesia’s new suite of agents (in preview) to design, develop, test, and deploy custom AI apps.
• Intake Policies: Intake policies determine how content is processed, including classification, conversion, and extraction. Each project has a configurable intake policy, along with policies per content type. Enable text conversion for preserving the structure of long-form content, or the new grounded extraction for structured data with citations. 
• Grounded Extraction: Grounded extraction is now available, enabling AI-powered intelligent document processing. With Vertesia’s new grounded extraction, every extracted value is tied back to its visual location in the document, providing data provenance without the limitations of traditional IDP software.
• Attribute-Based Access Control (ABAC): Handle the most complex access control requirements with Attribute-Based Access Control (ABAC). Access to content objects and collections can be authorized based on a range of user, document, or collection attributes, offering flexible, rule-based security policies. 
• Content Views: Configurable search experiences with natural language queries. Use the Studio Assistant to quickly configure and preview new views of content and then embed the views in custom apps.
• Event Bus: Route events to agents, workflows, or external systems via webhooks. The Event Bus provides an event-driven architecture (EDA) for the entire platform including event subscriptions, indexing, and the audit trail.
• Collaborative Document Editing: Collaborate with an agent to review and edit documents — edit the document directly, ask the agent for changes, review the result, and decide when to publish changes back to the document.
• Agent Checkpoint Configuration: Agent checkpoint behavior can now be controlled by admins in the project Settings, giving teams control over larger context windows for agents and when they checkpoint.
• Project-Scoped User Groups: Organization administrators can now create user groups restricted to specific projects, providing group isolation in a multi-project organization.
• Custom Header Buttons in Composite App: Composite apps now support a customizable header button bar, letting admins configure custom action buttons directly in the app header.
• Bedrock Mantle: Run workloads with OpenAI-compatible models on Bedrock using Vertesia's new provider for Amazon Bedrock Mantle
• Gemini Omni: Gemini Omni video generation is now available through Google Agent Platform (Vertex AI), enabling AI-powered video creation. 
• Gemini 3.8 Flash: The latest version of the Gemini Flash family is supported in Vertesia on Google Agent Platform (Vertex AI) environments.
• xAI / Grok Model Support on Vertex AI: Grok models are now available through Vertex AI.
• GPT-6 Astra: OpenAI's new frontier model built for cybersecurity, professional work, software engineering, and science is supported on OpenAI-compatible environments.

Improvements

• Agent Panel Overflow: When an agent has more right-panel tabs than can fit on screen, the excess tabs now overflow into a "More" dropdown, keeping the panel clean and accessible.
• Agent Deep Links: Deep-link URLs now navigate directly to specific agent resources, making it easy to share or bookmark a particular agent or resource with a teammate.
• Slim Payloads for Conversation Workflows: Payload overhead for conversation workflows has been significantly reduced, improving throughput and performance for long-running, conversation-based agent sessions.
• Process Test Lab: A new Process Test Lab in Studio lets teams write and run deterministic tests against processes, making it easy to validate them before deploying changes.
• API Contracts: API contracts across the platform are now defined and validated at runtime using Zod schemas, providing clearer error messages and stronger validation.
• Prompt Caching with OpenAI and Gemini: Prompt caching now includes OpenAI and Gemini models for faster responses and lower costs.
• Model Thinking Output: Thinking and reasoning output from AI models is now preserved consistently across all providers, ensuring reasoning traces are always available regardless the model or provider.
• Standardized Delete Actions: Delete actions across all table components now follow a consistent pattern, making the interface more predictable and reducing the chance of accidental deletions.
• Bulk Content Object Export: Multiple content objects can now be exported in bulk, significantly reducing the time needed to extract large sets of content from Studio.
• Permission Settings: The permission settings have been improved, making it easier for administrators to control who has permission on what.

Deprecated

• Legacy Workflow Rules: The legacy workflow rules have been deprecated by the new Event Bus.