Hey, i am trying to find a way to improve the context awareness or context knowledge of my chatbots. I noticed that when using Knowledgebases tha chatbot can only retrieve answeres that are explicitly in the KB. For example i have a list of shops, each with opening hours and some product categories they have. For example H&M Products: T-Shirts, Sweatshirts, Accessoires, Hoodies, Shoes Opening hours: 8am -9pm if a user asks "Where can i buy a T-Shirt" the Bot would answere something like "you can buy a Tshirt at H&M" But if a user asks "where can i buy jeans", the bot cant find an answer, because "jeans" isnt explicitly stated in the KB. When using context injection instead of a KB the AI is able to "think" about it and come to the conclusion that H&M also has jeans, although it isnt explicitly stated in the context i have given, leading to a better answer quality. Does anyone know a solution to this, or ways on how to improve the behaviour?

Hey This is a really good question and also a bit of a tricky one. I would recommend you to try and add context to the KB agent in its settings. You should be able to find an input field where you can give it some context. The first context you give it there will probably not be the best so I would recommend you to try giving it all kinds of context in different sizes and with different information. Most importantly, it can only be as good as you make it. Having a KB with less redundant information but with more important information will let it answer more diverse questions.

Thanks for the advice!

You are very welcome