The community member is trying to obtain the llam_index to use as the completion prompt when using a model. They noticed that when they manually create a query_engine using llama_debug, the event_pairs is empty, but when they create the query_engine using index.as_query_engine(), the event_pairs is not empty and they can get the desired content. A comment suggests that the community member should try passing the callback manager into the response synthesizer as well.
I want to obtain the llam_index to input as the completion prompt when using the model. Why is it that when I manually create query_engine using llama_debug, the event_pairs in llama_debug is empty, but when query_engine = index.as_query_engine() is created, event_pairs in llama_debug is not empty and I can get the desired content?
