AI Agent Management
Playground
Test and optimize your chatbot before deployment.
Documentation Index
Fetch the complete documentation index at: /llms.txt
Use this file to discover all available QwryAI pages before exploring further.
Playground is the safest place to test how your chatbot will respond to customers. Use it before deployment, after every major source update, and whenever analytics or knowledge gaps show weak answers.
- Use realistic questions from customers.
- Validate the answer, tone, escalation behavior, and lead form timing.
What to Test
- Common customer questions
- Policy and product edge cases
- Requests for a human agent
- Lead form triggers
- Escalation behavior
- Prompt or model changes
- New or updated data sources
| Test type | What to verify |
|---|---|
| Common questions | Answers are accurate, direct, and grounded in source content. |
| Edge cases | The chatbot asks for missing details instead of assuming. |
| Human handoff | Escalation stops AI replies when agent judgement is required. |
| Lead forms | The form appears only when the visitor is qualified or needs follow-up. |
Run a Test Conversation
- Open the workspace and select the chatbot.
- Open Playground.
- Ask a real customer question.
- Read the answer for accuracy, tone, and confidence.
- Update data sources or instructions when the answer is weak.
- Ask the same question again after each change.
- A good answer should be specific, grounded, and clear.
- If the chatbot cannot answer, it should ask for context or escalate.
Deployment Checklist
- Common questions are answered correctly.
- The chatbot does not invent unavailable details.
- Escalation works when human support is needed.
- Lead forms appear at the right moment.
- Brand voice and tone feel consistent.
- The selected chatbot matches the site where it will be deployed.