The community members are discussing ways to make the pandas query engine more contextual, such as by creating a chat engine that can handle follow-up questions. They also discuss issues with plotting and saving plots from the query engine, as well as passing column descriptions as metadata to help the engine handle similar column names. The community members also discuss ways to improve the inference time of a context chat engine without using a GPU, and how to make the query engine work with multiple CSV files. However, there is no explicitly marked answer in the comments.
You can try creating a chat_engine from the pandas query engine.
There are different chat modes you can use for chat engines, the following example creates a "condense_question" one which creates a standalone query from the chat history.
Will try this . Also pandas query engine is able to to plot few times inline in jupyter notebook, but I am unable to see any .jpg or plot instance in return response from pandas query engine. How can I save those plot and show in any other chatbot
it doesn't really support multiple CSV files on its on -- each query engine using a single dataframe. You could combine with an agent or sub-question query engine to work with multiple
