June 10

Agents

Full Response Mode for the Agent Component

The Full Response Mode is a new option available in the Agent Component that, when enabled, returns extra information about Agent Component execution in the output, including both the LLM-generated output and detailed metadata about the AI interaction, such as sources and toolExecutions.

This feature provides better visibility into the Agent Component execution, allowing improved error handling and evaluation of intermediate steps.

Read the Agent Component documentation for more information.

Token usage in the Agent Component’s Output Details

In the Output Details section of the Agent Component, token usage is now displayed alongside the Time and Duration metrics.

This improvement allows you to quickly see the number of tokens used in a specific execution, providing greater governance and control over your spending.

Learn how to run tests in your Agent Component and analyze the results.

Platform Improvements

Simplified API pipeline creation experience

The experience for creating pipelines with the API Trigger has been redesigned so you can build faster, without opening the Canvas from the start.

You configure endpoints, authentication, rate limiting, and CORS through a dedicated interface on the Pipelines page. The platform then generates the Canvas structure automatically, with all routes already configured and following best practices.

Create your first API pipeline

Run and Monitor updates

• New alert in Run: When you configure a deployment — whether deploying, redeploying, or promoting a pipeline — with more than 10 replicas, an alert will be displayed so you can evaluate whether to continue with this configuration. No action is blocked. Learn more in How to deploy or redeploy a pipeline.

• New 1 day quick filter in Monitor pages: You can now use the 1 day (1 d) quick filter on the Monitor pages to view data from the past day.

Usability improvements

• Latest Agent Component version in Capsules: Capsules now include the latest Agent Component version, matching the behavior in pipelines.

• Documentation icon in the Agent Component: The component now includes a book icon () for quick access to the configuration documentation, opening it in a new tab directly from the component.

Bug Fixes

• Persistent Alias: Fixed a bug where editing the Alias field of a copied connector also changed the Alias of the original connector. Each connector now maintains its own Alias independently.

• Paste not triggering search in Monitor pages: Fixed a bug where pasting a pipeline name in the Pipeline field did not trigger a search; it only worked when selecting from the dropdown.

• Copy and paste issue: Fixed a bug in the Agent Component's Datasets configuration where pasting content into one Evaluation field unexpectedly cleared data from other fields.

• Paste issue in Capsules: Fixed a bug where an Agent Component pasted into a Capsule retained its original account configuration, preventing the Capsule from being saved.