Playground
Test your agent in real-time
Playground
The Playground is your safe space to test and debug your agent before going live. It simulates a real chat environment but with added developer tools to help you understand why the AI answers the way it does.
Key Features
- Real-Time Simulation: Test exactly how your agent will behave on your website or WhatsApp.
- Debug Mode: When enabled, the AI displays the "Sources" for every answer. You can see the exact snippets of text from your PDFs or website that were used to ground the response.
- Model Switcher: Test how different brain models (e.g., GPT-4o vs Claude 3.5 Sonnet) perform with your data. Some models are better at technical reasoning, while others excel at creative writing.
- Variable Viewer: Track captured data (like Name, Email, or Order ID) that the AI extracts during the conversation.
- System Instruction Testing: Instantly see how changes to your "Personality" settings in the dashboard affect the AI's tone and behavior.
Testing Strategy
- Ask Edge Cases: Don't just ask easy questions. Try to confuse the bot to see where its knowledge ends.
- Verify Citations: Ensure the AI is citing the correct documents. If it's citing the wrong page, your data might be too redundant.
- Reset Session: Use the "Clear Chat" button to start fresh and test lead capture flows from the beginning.