June '26 Release Notes

New Features and Models

• Google Cloud Storage URI Support for Gemini (Vertex AI): You can now reference files stored in Google Cloud Storage, enabling support for significantly larger files and reducing unnecessary data transfer when using Gemini models through Google Agent Platform (Vertex AI). To get started, configure the Bucket Access Principal in your Vertesia environment settings.

Bug Fixes

• Default Prompt Example: The default prompt example reflects the expected format when creating new prompts.
• Prompt Rendering in JavaScript Templates: A syntax and rendering error has been resolved with newly created JS prompt templates.
• Image Links in Markdown: Images in documents converted to Markdown are now referenced using the correct standard attachment path, preventing broken image links from appearing in converted content.
• Bulk Imports Routed to Low-Priority Queue: Large bulk import jobs are sent to the low-priority processing queue, so they don't compete for resources with interactive workflows.
• Extended Workflow Execution Timeouts: Workflow execution timeouts have been extended so that longer-running jobs complete successfully instead of failing prematurely on complex, multi-step processing tasks.
• Office Conversion and Timeouts: Error handling in the Office-to-PDF conversion workflow has been hardened so unexpected file types, edge-case inputs, and timeouts no longer cause conversion jobs to fail. 
• Video Processing Timeout Extended to 30 Minutes: The video preparation step now allows up to 30 minutes to complete, resolving timeout failures that occurred when processing larger video files.
• DSL Workflow Execution Timeout Extended to 4 Hours: Complex DSL workflows can now run for up to 4 hours without timing out, resolving failures that previously occurred midway through long-running multi-step jobs.
• Conversation Lesson Resources Restored: An incorrect file path preventing access to conversation lesson content has been corrected.
• OpenAI Driver Updates and Parsing Fixes: Several issues with the OpenAI and Azure Foundry integrations have been resolved, including Azure Foundry request rejections and structured output streaming errors with GPT-5 family models.