Hi! I use results from KB (kb.results) and user question as AI Task Input. My KB consists of question-answer pairs. It seems like even though the right pair is in the kb.results, the AI task only picks the first pair and gives me the answer from it as a result. I don't understand where the problem comes from. Maybe I'm doing something wrong?

Hi @eager-accountant-7961, are you daying that more than one fact was relevant and should have been returned as well^

@freezing-printer-49373 Hi! There are 5 facts that KB returns. One of them is the most suitable for the user question, but it may be placed as the second or third in the array. I turn this array into a long string, where facts are separated by special symbol sequence, eg $###$. But AI task alwasys return the first one (fact), no matter if it suits the question or not

yeah, prioritisation of answers within a KB or also between multiple KBs is not well explained. Please shed more light on this.