As projects grow so does the complexity. Asking AI to crawl backtest data, large sets of data, large files, is all really wastful and often leads ot ai errors. Instead, tag all your data using ontologies and taxonomies and then ai can make multi hop reasoning jumps across super large data sets.
If you are looking for that next big discovery from your data, you should be using a RAG such as Neo4j.
This is how you build your own Palantir, a project built on "tags" - ontologies, for connecting the dots in your in noval ways.
Example: Tag all your trade data as just that : Trade, Asset, Time, Outcome, Stategy.