Step 1
Simulate
Generate simulated conversations between virtual users and your agent.
Need scenarios? Use to generate them, or provide your own scenario.json below.
Agent Config
How arksim connects to your agent.
Type
Name
Endpoint
Loaded from config YAML. Select a different config in the sidebar to change.
Scenario Input
Directory:
Simulation Parameters
Requires env:
Run
Status:
No logs yet.
Step 2
Evaluate
Score simulated conversations on goal completion, helpfulness, and coherence.
Model
Requires env:
Input Source
Run Simulate first, or switch to Load from Disk.
Configuration
Metrics
Run
Status:
No logs yet.
Step 3
Results
Review evaluation scores and the full HTML report.
Input Source
Run Evaluate first, or switch to Load from Disk.
Overall Score
Pass Rate
Avg Turns
Conversations
Conversation Scores
| Conversation | Goal Completion | Final Score | Status |
|---|---|---|---|
HTML Report
The full report is best viewed in its own tab.
No results yet
Run Simulate then Evaluate to generate results, or switch to Load from Disk.
Step 0
Build Scenarios
Create test scenarios that define how simulated users interact with your agent.
Auto-generate Scenarios
PRO
Automatically generate realistic test scenarios from your agent's knowledge base.
Load Existing
Scenarios
No scenarios yet. Add one manually or load an existing file.