@irony mentioned some jobs might need paragraphs (the raw PDF content) to handle retries, but in the later steps it could be enough to work with the result from earlier steps.
Before making this change, we need to find which steps need the paragraphs as context and which ones can run without it.
@irony mentioned some jobs might need paragraphs (the raw PDF content) to handle retries, but in the later steps it could be enough to work with the result from earlier steps.
Before making this change, we need to find which steps need the paragraphs as context and which ones can run without it.