I am working on extracting the Profit & Loss Statement and the Balance Sheet from annual financial reports. Currently, I use two separate prompts to extract these statements from each annual report of a stock.
I process five years of annual reports per stock using the Batch API, where each PDF is uploaded once and two extractions are performed per file.
I would like to understand whether it is possible to reuse cached content (i.e., store the PDF content in cache) to extract both statements within Batch API requests. If this is supported:
-
Will the second extraction be billed at cached content rates, or
-
Will it be treated as a separate request with standard input token charges, even though the same report file is reused?