The community member is asking if there is a way to maintain context in their query_engine.query() function, so that the question can refer to previous questions and answers. The comments suggest that the community member can either use a chat engine or manually manipulate the prompt templates before calling the query function each time. However, the per-LLM context limit would still apply in the latter approach, whereas the chat engine would read the chat history and write its own query to the index.
Sadly yes, although the condense chat engine reads the chat after a user asks a question, and then writes its own query to the index. Whereas the latter approach would just have the chat history in the prompt for the index
