Release 1.56.0

Improvements & Enhancements

• Agent Cloning - Agents can now be downloaded and imported as .gwai files, making it easier to import and export configurations across environments. This simplifies workflows for teams that need to move Agents between development, staging, and production setups.

• Styling Updates - We’ve refreshed the styling across Agents, Evaluations, and Chunks to create a more seamless and consistent user experience. These updates improve visual clarity and make core workflows feel more unified across the platform.

• Data Retention - Projects now support configurable data retention policies that allow user inputs and Agent responses to be redacted. This gives teams more control over how conversational data is stored and managed.

• Session Handling - Browser session handling has been improved to provide a more reliable experience. These updates are designed to reduce session-related interruptions and improve overall stability during use.

• Agent Triggers - You can now add Agent Triggers that execute when conditions defined in natural language are met. This makes it easier to configure dynamic Agent behaviour without relying on rigid rule definitions.

• Smart Response Blocks - Smart Response Blocks are the first supported Trigger type. They allow Agents to generate explicit response blocks when specified conditions are met, enabling more structured and predictable responses in the right contexts.

• Evaluation Test Bench - The new Evaluation Test Bench allows teams to configure Test Suites and run them against Agents at scale. This helps detect behavioural drift over time and makes it faster to validate changes, including model updates and other Agent adjustments.

• Custom Metrics - Evaluations now support custom defined metrics, giving teams more flexibility in how Agent performance is measured. This makes it easier to track success criteria that are specific to your use case.