Tracing Actions

The following actions are available within Tracing. These actions are also accessible from the Sessions and Users pages under Observability.

1. Add to Dataset

Adds the selected trace's input and output data to a dataset.

Datasets are collections of examples used for:

• Testing AI applications.
• Evaluating model performance.
• Benchmarking application behavior.
• Validating AI workflows using real-world interactions.

2. Annotate

Adds annotations to a trace or observation using predefined Score Configurations.

Annotations can be used to assign:

• Scores.
• Labels.
• Evaluation metrics.

based on configured scoring criteria.

Annotations can be applied at the following levels:

• Trace level – Applies evaluation results to the complete trace.
• Observation level – Applies evaluation results to individual observations within a trace.

3. Add Comment

Adds comments to a trace for documentation, review, and collaboration purposes.

• Enter comments to provide additional context about the trace.
• Use comments for review and collaboration among users.
• Comments remain associated with the selected trace.

Comments remain associated with the trace and are visible to users with the appropriate permissions.

4. Usage Breakdown

Displays token usage details for the selected trace.

The usage breakdown includes the following metrics:

• Monitor token consumption for individual traces.
• Analyze usage patterns and model interactions.
• Identify requests with high token utilization.

5. Download

Downloads the selected trace, session, or user data as a JSON file.

• Select the required trace, session, or user data.
• Click the Download option from the actions menu.
• The system prepares the selected data in JSON format.
• Download the generated JSON file for further analysis.
• Use the downloaded file for auditing, troubleshooting, or record keeping.

The downloaded file contains the corresponding execution details and metadata for further analysis or record keeping.