Monitor, Visualize, and Provide Feedback on Agent Runtime

Agents Visualization is now in Public Preview

Agent Visualization now allows you to visually understand your deployed agent definition, correlation of agentic loop to the trace, and replay jobs to better understand failure/error patterns, etc. This new style of agent representation is based on user feedback to better help understand what is happening with their agent to make them more reliable. This will be a new authoring style that we introduce into lowcode agent designtime over the next few months as well.

New Refinements to Agent Monitoring/Observability in Agent Instance Management:

The instances section is now organized into 3 distinct tabs:

  • All Agents for a comprehensive view of all agents
  • Early adopter Agents for agents created as part of the early adopter program
  • Errors (formerly incidents)

The Management section now includes the following to provide a unified interface:

  • Templates for accessing Marketplace
  • Context Grounding Indexes: create new indexes, sync existing, and monitor the utilization of indexes across your organization

Provide Feedback to Help Improve your agent
You can now provide feedback on agent traces to indicate if the output and trajectory (or agent plan execution) was acceptable. This will help you aggregate meaningful data to inform agent improvements and will ultimately be used programmatically to help agents learn from previous interactions. Give it a try - stay tuned for some additional product updates coming soon on this!


Documentation
Release Notes,
Managing UiPath agents,
Traces

1 Like