Hi,
When I call RetrievalQA.from_chain_type(…) with parameter retriever = docsearch.vectorstore.as_retriever() where docsearch from CSVLoader - doesn’t it mean that all my file will be uploaded into OpenAPI (and use max 1K tokens)?
I’ve askes this stupid question because I load my data into CSVLoader and model uses only last portion of data and “lost” first set of rows. For example it’s set of items. I’ve asked model how many items I have and result is much less than I have from direct call len(loader.load()).
Pavel