Ingesting documents

Adding a document starts work that takes minutes, so the call returns a job rather than a result.

Not written yet

This page is a placeholder. The site's build and hosting are in place; the prose is not. What belongs here:

• The three sources -- text, url, file_url -- and exactly what each accepts
• That file_url is not any public URL on the internet; state the real constraint
• That raw file upload (multipart) is not supported in v1, and what to do instead
• What a job_id is, how to poll it, and what each document status means
• That a 202 means accepted, not done -- the mistake integrations make most often
• The batch ceiling of 500 documents per call, and that a file reference cannot contain a comma