2026-05-28

Date: 2026-05-28

Added reasoning controls in assistants, so teams can enable it only when needed and better manage token usage.
Added support for deep research and reasoning-capable models in generator workflows.
Added support for custom OpenAI-compatible provider URLs, making it easier to connect more LLM services.
Added a portal warning when a user has no assigned groups, so access issues are clear before using the platform.
Added guidance in the portal when web search requires a compatible model, helping users choose the right setup.

Fixed the evaluator list actions menu so all available options are shown correctly.