I tried the following first five chess moves: e4 e5 Nf3 Nc6 Bb5 a6 Ba4 Nf6 O-O Be7 and your chatbot gave the correct answer (picture). Now it uses GPT-4 Turbo via API, just you put your own API key in it. However, yesterday I also managed to get wrong answers from GPT-4, so test it carefully, and the best option would definitely be to put all possible (best) chess first five opening moves in the Knowledge Base documents or tables and retrieve the correct answers from there. Also when you look at this copy-chatbot files Execute code card, when you give better instructions in the content section (system and user), you will definitely get the chatbot to give better answers with Query Knowledge Bases, AI Task and GPT-4.