A common question about the API responses is:
What's the difference between embed_text
and raw_text
in the response?
It's simple! If you enable LLM table summaries, they're only added to the embed field. The raw text field is supposed to be ground truth. We've found that the summaries are helpful for retrieving the right chunks, but the table data in the raw text field is sufficient for Q/A once the right chunk is retrieved.