Skip to main content

​
Overview

Use this stage for conversation simulation only with pre-built scenarios. When to use this:
  • Running conversations with pre-built scenarios
  • Re-testing your agent after making changes without regenerating scenarios
  • Testing specific scenario sets in isolation
What it does: Takes the generated scenarios and runs conversations with your agent, recording all exchanges. Output: Conversation logs are saved to output_dir (default: ./results/conversation/) for evaluation.

​
Running the Command

./run_arksim.sh simulate config_simulate.yaml

​
Configuration File

​
Understanding Key Settings

  • input_dir — Points to where your scenarios live (from the build step)
  • max_turns — Safety limit on conversation length. Prevents runaway conversations
  • display_time_delay — Adds realistic pauses when viewing conversations in real-time

​
YAML Configuration

# Path to agent setup directory
agent_setup_dir: ./examples/bank-insurance

# Path to scenario output directory (from build step)
input_dir: ./results/scenario

# Output directory for conversation results (default: conversation/)
output_dir: null

# Simulation settings
num_conversations: 5                      # Number of conversations to generate/simulate
max_turns: 5                              # Maximum turns per conversation
simulated_user_behavior_instructions: []  # Custom instructions for simulated user behavior
num_workers: auto                         # Number of parallel workers
display_time_delay: false                 # Add delays between turns when displaying conversations

# LLM settings
model: gpt-5.1                            # LLM model used by Arksim (not your agent)
provider: null                            # LLM provider (e.g., openai, azure)

​
Next Steps

After simulation completes, proceed to the Evaluate Conversation to analyze agent performance.