Skip to main content

Agent Executions Feed

Monitoring agent execution provides valuable insights into agent performance, usage, and potential areas for improvement. The Agent Execution Feed is designed to deliver detailed metrics on these operations, including key information about the model, project, and agent activities.

Key Metrics in the Agent Execution Feed​

The Agent Execution Feed provides a variety of critical metrics, allowing users to track and analyze agent activity:

  • Model Provider: Indicates the provider of the model being used within the agent. Knowing the model provider helps assess the performance and resource consumption of different models within the system.
  • Model Name: Specifies the name of the model that is being used within the agent. This helps identify which specific model was used in the execution, which is useful for debugging or optimization purposes.
  • Project: Identifies the project the agent was built in. By tracking agent execution within specific projects, users can analyze performance and resource usage for each project individually.
  • Source: Shows the origin of the request or the trigger that initiated the agent execution, giving context to the agent's activity.
  • Agent Name: Displays the specific agent that was executed.
  • Response Time: Captures the time it took for the agent to respond to a request. A higher response time could indicate potential issues in the agent’s performance or inefficiencies in processing that need to be addressed.
  • Total Tokens: Reflects the number of tokens consumed by the agent during the execution, helping track the cost and efficiency of agent operations.

Execution Details​

Users can select a row within the agent execution feed to get further details on the execution. The details provide insgiht into the agent results, highlighting:

  • User Input: Details the input that the agent was given during execution.
  • Returned Chunks: Provides insight into the chucks returned that the model used to provide an answer
  • Agent Output: The ouput that the model provided as part of the agent execution.

This information can be used as an important audit tool to review any agent execution for performance and accuracy.

Filtering Options​

To enhance the usefulness of the data, the Agent Execution Feed allows for filtering based on:

  • Date: Narrow down the report to specific timeframes to analyze agent performance over time or to track spikes in execution activities.
  • Project: Filter the feed to focus on specific projects, providing a more detailed view of agent execution within a certain context.
  • Source: Use this filter to focus on specific sources of execution requests, such as particular data inputs, API triggers, or user-generated actions.

These filtering options allow for targeted analysis, helping you to drill down into specific aspects of agent performance and optimize where necessary.

Report Refresh and Export Features​

The Agent Execution Feed can be refreshed at any time to ensure that the latest agent execution data is always available. This allows for real-time monitoring and quick identification of performance issues or inefficiencies.

Additionally, the report can be exported to CSV for offline review or sharing. This feature is valuable for teams needing to analyze agent execution data outside of the platform or for reporting purposes. The CSV export provides a straightforward way to track long-term trends or integrate the data with other tools for deeper analysis.