Documents
When a page fails to read
Total loss, a few missing pages, and the difference between the two kinds of retry.
Last updated 1 min readFor team members
Three states you might see
The document is marked failed and has no text at all
- What it means
- No page was read. Rather than showing an empty success, the job is failed with the error recorded.
- What to do
- Hit retry. The whole document runs again and you do not need to re-upload the file.
The document finished but says pages were lost
- What it means
- Some pages read and some did not. The exact page numbers that were lost are recorded.
- What to do
- Hit retry. Only the missing pages run this time; the pages that already succeeded are not re-read and cost nothing again.
The document is stuck on "processing" forever
- What it means
- The service restarted mid-run.
- What to do
- Nothing to do — interrupted jobs are marked failed on that same restart so they can be retried. If it is still processing, wait a few minutes and refresh.
Why an incomplete document is not indexed
Indexing — what makes a document usable by the assistant — only happens after a clean run. A run that lost pages is not indexed.
The reason is simple: an assistant answering from an incomplete document gives a confident wrong answer, and you have no way to tell.
Two runs at once is not possible
Retrying a document that is currently running is refused. The same holds for re-indexing.
That prevents both double spend and contradictory results.
Read this next
- When something does not workA document did not readThree failure states, and which of them is one click away from fixed.
- DocumentsReview a document page by pageSee the extracted text beside the original image, and correct it.
- DocumentsFormats and limitsWhat is accepted, what is not, and where the caps are set.
Still stuck?
If the answer wasn't here, write to us. It is a real inbox and it gets answered.