Skip to main content
The Playground is a private chat where you can talk to your assistant exactly as a customer would, without affecting your live website. Use it to verify answers, tone, and product recommendations before you publish.

Open the Playground

1

Open your assistant

Go to Assistants in the homie platform and select the assistant you want to test.
2

Switch to the Playground

Open the Playground tab to start a fresh conversation with the assistant.
3

Send a message

Type a question the way a real shopper would and review how the assistant responds.
Bildschirmfoto 2026 07 01 Um 10 24 22

Simulate the environment

By default the Playground talks to the assistant with no page context. To test realistic scenarios, select Simulate environment in the top-right corner to open the Test parameters dialog and adjust:
  • Simulate website — a page URL (for example https://example.com/product/123). The assistant responds as if it were opened on that specific page, so you can test page- and product-aware answers.
  • Store ID — for multi-location stores, simulate location-specific responses for a given store (for example store-berlin-001).
  • Language — override the language used in the Playground. By default it uses the assistant’s default language.
Select Save to apply. The conversation restarts with the simulated context.

What to look for

When testing, pay attention to whether the assistant:
  • Answers questions accurately and stays on topic.
  • Uses the right tone and personality for your brand.
  • Recommends relevant products at the right moment.
  • Handles questions it does not know gracefully, instead of inventing answers.
  • Responds in the expected language.
Test the Playground again after every change to your knowledge or fine-tuning. Updated content and tone settings change how the assistant answers, so a quick re-test confirms the change behaves as you intended.
The Playground reflects your saved configuration and current knowledge. If an answer looks wrong, check your knowledge sources and fine-tuning before publishing.

Next steps

Fine-tune responses

Adjust tone and product-recommendation behavior.

Add knowledge

Improve answers by expanding what your assistant knows.