Run and Track

After scenario design, execution tells you whether your agent actually performs as expected. This page explains how to run simulations and monitor execution status confidently.

Step-by-Step: Run a Scenario

1) Open an existing conversation simulation

1. Go to Simulation -> Conversation.
2. Open a scenario from the list.

2) Start execution

1. Click Run on the scenario form page.
2. Confirm when prompted.
3. You will be redirected to the simulation result detail page.

3) Track run state

Watch the execution status while it processes:

• running: simulation is still processing
• completed: all steps finished
• failed: one or more steps failed
• stopped: execution was manually stopped

4) Use controls during and after execution

• Refresh while running to fetch latest status
• Stop if you need to cancel the run
• Run Again after completion for regression checks
• Delete old runs you no longer need

Use Results List Efficiently

Open Simulation -> Results to manage many executions.

Use filters and sorting to:

1. Find runs by date range.
2. Search by execution code.
3. Filter by status.
4. Sort by latest or oldest.

This helps you quickly inspect relevant runs after major changes.

Execution Best Practices

1. Run critical scenarios first after every major update.
2. Keep a consistent run cadence (daily or weekly depending on traffic risk).
3. Re-run the same scenario set before and after model changes.
4. Use execution codes in team discussions for traceability.
5. Stop and investigate immediately when high-priority scenarios fail.

Model-Switch Regression Mini-Process

Use this lightweight process whenever you change model:

1. Run baseline scenarios on current model.
2. Switch model in draft settings.
3. Run the exact same scenarios.
4. Compare criteria outcomes.
5. Publish only after critical scenarios pass.

Next Step

Now interpret criteria quality and convert findings into improvements:

Analyze and Improve