Generated application - how do I break up the loading

[kehh]

I'm finding that the ingestion pipeline is not able to deal with very large amounts of data as it appears to load all of the documents in memory before persisting them in a storage context. Is there a way to partition documents by loader or something similar so that we don't clear out all documents when we run a single loader. I'm specifically looking at the code in https://github.com/run-llama/create-llama/blob/main/templates/types/streaming/fastapi/app/engine/generate.py#L78