Playground
Test and optimize your AI agent before deployment. Compare models, adjust settings like temperature and system prompts, and evaluate responses.
The Playground is your testing environment. Test your agent, compare different configurations, and refine responses before deploying to production.
System Prompt (Instructions)
Your system prompt defines your agent's personality, tone, knowledge scope, and response style.
Key Benefits
- Brand alignment — Match your agent's voice to your company's tone
- Consistent behavior — Predictable, reliable responses
- Focused knowledge — Guide the agent to specific topics
- Professional tone — Set appropriate formality for your audience
Best Practices
- Be specific — State the agent's role clearly (e.g., "customer support specialist")
- Include examples — Show desired response styles
- Set boundaries — Define what the agent should and shouldn't discuss
- Test variations — Find the most effective prompt for your use case
Start with a simple role description, then refine based on real test results and user feedback.
Creativity / Temperature
Temperature controls how creative and varied your agent's responses will be. Range: 0 to 1.
Temperature Settings
- Lower (close to 0) — Focused, consistent, predictable responses. Best for support, FAQs, and factual information.
- Higher (close to 1) — More creative, varied, less predictable responses. Best for creative writing or brainstorming.
Default: 0 (consistency). Adjust based on your use case and test in the Playground first.
Test & Compare
Use the preview panel to test your agent's responses interactively.
Testing Process
- Test variations — Ask the same question different ways
- Evaluate quality — Check accuracy, relevance, and tone
- Adjust settings — Change temperature, model, or instructions
- Retrain — Update training data based on findings
- Verify — Confirm improvements before deploying
What to Look For
- Accuracy — Is the response factually correct?
- Relevance — Does it answer the specific question?
- Length — Is it appropriately concise?
- Tone — Does it match your brand voice?
Best Practices for AI Agents
Learn how to train high-performing AI agents, improve response quality, and structure training data for accurate answers.
Data Sources
Train your AI agent with websites, documents, text snippets, Q&A pairs, and Notion. Learn how to structure training data for optimal agent performance.