Playground
Playground is an interactive environment where you can try the chatbot exactly as it appears when deployed on the web. It is used to test responses after changes in configuration, data sources, or the system prompt.
1. How to do it
- In the left menu, click Playground.
- Type a test query in the chat field and send it.
- Check the agent's response.
Changes in Configuration or Knowledge base appear in Playground immediately after saving — no refresh is needed.
2. What you will find in Playground
- Chat — a full chatbot window, identical to what the user will see on the web.
- Agent type — switch between AI assistant and E-commerce agent (if active).
- Known sources — an overview of URLs and files the agent draws from. Shows the total number of sources and their status:
- Your chatbot has already learned your sources! — all sources are loaded.
- Your chatbot has already learned most of your sources! — some sources are still being processed.
- If no status is displayed, sources have not been added yet or are still loading.
- Add sources — a link directly to the agent's Knowledge base.
3. Testing tips
- Test queries that your real users would ask — not ideally formulated questions.
- If the agent does not respond correctly, first check the source status in the Known sources section. If sources are not loaded yet, wait and try again.
- After each change to the system prompt or sources, test at least 3–5 queries from different areas.
- Playground can be shared with colleagues using just the URL. No login to the application is required.
Frequently asked questions
-
Are Playground conversations saved to Messages?
Yes, Playground conversations appear in the Messages menu the same way as conversations from real users.
-
How do I tell if the agent is drawing from the wrong source?
Enable source display in Configuration → Advanced settings → Source display mode. The agent will then show a link to the source it drew from below the answer.