Hey All, Here is our use case, we are loading daily event logs per user as documents via pandas. The goal is to generate and persistantly store individual daily summaries, 7 day rolling summaries, monthly summaries and lifetime profile summaries. We also plan on using these indices in a chatbot.
Is it possible to define individual summary templates for the timeframes when constructing the tree index so that they can be stored? if so how? should we be using a tree index or a graph?
Why involve a tree or graph? You could just generate a summary of the logs in the rolling windows no? And using a list index for generating each rolling summary with response_mode="tree_summarize"?
(i.e. daily summaries, last week of daily summaries go into creating a weekly summary, etc.)