To train our ML engine, we require a recent dataset of API calls related to the target API endpoint(s) with at least with three months of historical data. This helps us build models for the product's analytic process.