January '26 Release Notes

Audio File Support: Go beyond basic speech-to-text and expand your multimodal capabilities with audio understanding, by using audio files as media inputs with interactions.
Custom Headers and Data for Webhook Notifications: You can now include custom headers and data in your webhook notifications, providing greater security and flexibility when integrating with external systems.
xAI (Grok): Create new environments in Vertesia using xAI (Grok) as a provider to access xAI models with tool usage.
OpenAI Compatible Providers: Connect to OpenAI-compatible API endpoints, expanding options for new inference providers and Vertesia environments.

Agent Conversation Management: Agents work faster and more reliably thanks to enhanced streaming support with multi-turn conversations and tool use tracking, plus image stripping and text truncation utilities to optimize agent conversations.
New Analytics: New analytics, designed for scalability, allow you to monitor total runs, view trends over time, track token usage, and analyze performance by status, environment, and interaction.
Monaco Editor in Prompt Studio: The code editor has been replaced in the Prompt Studio to provide superior syntax highlighting, autocompletion based on your prompt's schema, and inline error validation, preventing saves on invalid templates.
Input List Component: The multi-value input component (used for Tags and Agent Tools) now supports pasting a list of comma- or space-separated values and clearer UX for removing tags.
Role Name Change: The "project_admin" role has been renamed "manager" to provide clearer role distinction from users assigned the "admin" role on a project.
Smarter Text Chunking Logic: We've improved our text chunking algorithm to be more efficient, resulting in more balanced and meaningful text chunks.

Agent Runner UI: The Agent Runner creation modal now has proper scrolling to handle long forms, and proper handling of checkbox parameters. The Agent Runner detail view also features independent scrolling for the payload and conversation panels, improving usability on smaller screens.
Agent Topic Length Limit: Prevented workflow termination errors by truncating the conversation topic for agent runs to fit within the search attribute size limit.
Run View Stability: Resolved an issue on the Run Detail page where the UI could not load runs with missing environment data.
Search Robustness: Improved the robustness of vector search by allowing it to continue even if fetching an embedding for one of the configured models fails.
Sorted Dropdown Lists: We've fixed an issue where items in certain dropdown lists were not sorted alphabetically.
Handling Blank Pages in PDFs: The system now correctly identifies and handles blank pages in PDFs, improving the reliability of document intake.

Product release notes for January 2026