For the complete documentation index, see llms.txt. This page is also available as Markdown.

How to test answers in the playground

This guide shows you how to test how the AI answers a guest question in the playground, without sending anything to a guest.

Before you start

  • A role with permission to use the playground.

Open the playground

1

Go to the playground

Go to AI Center → Playground.

2

Start a conversation

Click New conversation, and choose the Conversation scope: the whole organization, a complex, a multi-unit, a listing or a reservation. To test another scope later, start a new conversation.

Ask a guest question

1

Write as the guest

Type a message the way a real guest would, for example "when can we get in??", and send it.

2

Check the reply

Read the AI reply. Check the knowledge it used, the fact-check result and the label Would send or Would not send.

If the reply used the wrong entry or no entry, fix the knowledge, then click Re-run Auto-Pilot from here.

Troubleshooting

What do I do when the AI gives a wrong answer for one property?
  • Start a conversation with that listing, or a reservation there, as the scope, and ask the same question.

  • Check which knowledge the reply used. If it is the wrong entry, its scope is probably too wide. Narrow it with Edit scope in Browse.

  • If no entry was used, the answer is missing. Add a Q&A for that listing, and re-run.

  • Also check the listing details, such as amenities and custom fields, in Properties. The AI uses those too.