Live Preview
Live preview lets you watch tests execute in real-time with a browser window showing exactly what the AI sees and does.
Starting a Live Preview
- Open a test case in the editor
- Click the Run button
- A split view opens:
- Left panel - Steps and status
- Right panel - Live browser preview
The Preview Interface
Steps Panel (Left)
Shows all test steps with status indicators:
| Icon | Status |
|---|---|
| ○ | Pending |
| ◔ | In progress |
| ✓ | Completed |
| ✗ | Failed |
Click a step to see AI reasoning for that step.
Browser Preview (Right)
A live view of the browser session:
- See the page as AI sees it
- Watch clicks and typing in real-time
- Observe navigation and page loads
- View the same state AI uses for decisions
The preview is a real browser session running in the cloud. You can see everything the AI sees.
During Execution
Step Execution
For each step, the AI:
- Takes a page snapshot
- Identifies the target element
- Performs the action (click, type, etc.)
- Waits for any specified condition
- Moves to next step
AI Reasoning
As steps execute, you can see AI reasoning:
Step: Click the "Sign In" button
Reasoning: Found a button element with text "Sign In"
in the header navigation. Clicking this button to
proceed with the login flow.
This helps debug when things go wrong.
Wait Conditions
When a step has a waitUntil condition:
- AI takes screenshot after action
- Evaluates condition against page state
- Retries with exponential backoff if not met
- Times out after 30 seconds if never met
After Completion
Successful Run
All steps show green checkmarks:
✓ Step 1: Navigate to login page
✓ Step 2: Enter email
✓ Step 3: Enter password
✓ Step 4: Click login button
✓ Step 5: Wait for dashboard
Failed Run
Failed step shows red X with error:
✓ Step 1: Navigate to login page
✓ Step 2: Enter email
✗ Step 3: Enter password
Error: Could not find password field
Review AI reasoning to understand the failure.
Rerunning Tests
Click Run again to restart the test:
- Fresh browser session created
- Steps execute from the beginning
- New dynamic data generated
Debugging Tips
Step Fails to Find Element
- Check AI reasoning for what it tried
- Make step description more specific
- Verify element exists on page in preview
- Check if element loads after a delay
Wrong Element Clicked
- Add more context to description
- Specify location ("in the header", "below the form")
- Use unique text or labels
Wait Condition Times Out
- Condition might never be true
- Page might load differently than expected
- Extend the wait or adjust condition text
Test Works Manually But Fails in Bug0
- Element might need scroll to be visible
- Animation or transition timing
- Dynamic content loading
- Try adding wait conditions
Best Practices
During Development
- Run after each step change
- Watch AI reasoning for insights
- Check preview matches expectations
- Iterate until steps are reliable
Before Publishing
- Run complete test multiple times
- Verify consistency across runs
- Check assertions pass reliably
- Test with different data
Debugging Failures
- Identify which step failed
- Read the AI reasoning
- Check browser preview state
- Adjust step description or add waits
- Rerun to verify fix
Technical Details
Browser Sessions
- Cloud-hosted Chromium browser
- Each run gets a fresh session
- Sessions are region-optimized for speed
- Cookies/storage applied from login suites
Execution Environment
- Browser controlled via Playwright
- AI interacts through accessibility tree
- Screenshots captured at key moments
- All actions logged for debugging