For tables and chart I don't know for sure. Embedchain is not multimodal so if your chart is an image it can't analyze and interpret it. For tables, the performance is similar to what you can get from LangChain or ChatGPT