In this article, we explain why the answers generated by the chatbot in the Interactive tester and the live environment may differ and how to address these differences.
It’s possible for answers in the Interactive tester and the live environment (via the chat widget, WhatsApp, Facebook, or Instagram) to vary. Several factors can cause this. Below, we outline why this happens and what you can do about it.
The chatbot generates answers based on current knowledge
Every time a question is asked, the chatbot generates an answer based on the most recent knowledge stored in it. This applies to both the Interactive Tester and the live environment. However, some factors can affect the consistency of answers:
- Conflicting knowledge: If conflicting information is present in the chatbot, it may result in different answers. For example, discrepancies between information retrieved through the Web Crawler and instructions added manually in the Instructions section can cause inconsistencies.
- Incomplete knowledge: Missing information can prevent the chatbot from providing a consistent response.
- Insufficiently specific instructions: If instructions in the chatbot are not clear enough, this can lead to variation in the answers provided.
To minimize these differences, ensure that the chatbot’s knowledge and instructions are clear and complete.
Cookies and cache in the browser
When using the chat widget to interact with the chatbot, the browser stores cookies and cache. These stored files can affect how recent updates to the chatbot are displayed. If you’ve made changes to the chatbot’s settings and then test further in the chat widget, it’s possible you may see older answers because the browser hasn’t refreshed its stored data.
Solution:
Clear your browser’s cache before continuing to test. This ensures the latest version of the chatbot is loaded. You can find a step-by-step guide on how to clear your cache in this Help Center article.
Still experiencing issues?
If you’re unable to determine why the answers remain inconsistent after following the above steps, feel free to send your examples to our support team at support@watermelon.ai. We’ll be happy to assist you!