The Testing Playground allows you to test AI responses before they go live. It helps you simulate real customer interactions and verify how the assistant responds to various types of queries—such as product searches, store information, and more—across different store pages like the homepage, product pages, and collection pages.
Left panel options
Manifest model
Select from a list of available Manifest models to test and compare responses. This helps identify which model provides the most accurate and helpful answers.
Assistant's personality
Define the tone of your assistant’s replies using default personalities or configure a custom tone. This lets you test how your AI communicates with users.
Response length
Choose how detailed the AI’s replies should be. Options include:Short – brief responses
Normal – balanced detail
Long – more comprehensive answers
Test on
Simulate queries on specific page types:Homepage
Product pages (select specific products)
Collection pages (select specific collections)
Language
If your store supports multiple languages, switch between them to test the AI’s multilingual capabilities.
Playground window
Testing on
Displays the current page context being used for testing.Reset
Clears the current chat and starts a new session for testing.Contextual options
Access:Reset Chat
Get Chat Transcript
Message console
Use this input area to ask questions and test AI responses in real time.Message source
For every AI response, the system shows the data source. Clicking this reveals the source panel on the right, where you can verify or update the source content.Correct response
If the AI response is incorrect or unsatisfactory, you can correct it directly. A corrected response is saved as a private note and added as an FAQ. The correction modal appears on the right, pre-filled with the original response, allowing easy edits.
Right panel
Message source viewer
This panel shows the underlying source of each AI response. You can access it by clicking "Message Source" from any AI-generated message.
For more details, refer to the Message Source & Correct Response article.
How to use the testing playground
Go to the Home page in your Manifest dashboard.
Click "Open Testing Playground" at the top.
Select the page type from the "Test On" dropdown.
If testing on a product or collection, use the "Select Product/Collection" button to choose a specific item.
Configure the testing parameters:
Model
Personality
Response Length
Language
Click Apply changes after selecting these settings.
Begin testing by entering queries in the message console.
Click Reset to start a new test session.
Use Correct Response to fix any unsatisfactory answers and automatically create FAQs for future reference.
FAQs
1) What is the purpose of the Testing Playground?
It allows you to simulate user interactions with your AI assistant before they go live, helping you validate and refine the assistant’s responses.
2) Can I see the data source of an AI response?
Yes. Click on “Message Source” below to the AI’s response to view and verify the exact source used to generate it.
3) Is it possible to test specific products or collections?
Absolutely. Select the desired product or collection via the "Test On" dropdown to test how the AI responds to queries about them.
4) Will the corrected responses in the playground affect the AI’s live responses?
Yes, once a response is corrected and saved, it helps improve the AI’s future replies and becomes part of the assistant’s training data.
5) Can I re-correct an already corrected response?
Yes, you can modify previously corrected responses through the same "Correct Response" workflow.
For further assistance or to raise feature requests related to Testing Playground, please contact [email protected].