Open Training → Knowledge → Add source → Files.
Select multiple files from your computer, or paste public document links, one per line. Accepted files start processing automatically. Upload progress is shown per file; retry failed uploads without resending successful ones.
Supported inputs include PDF, Word, spreadsheets, presentations, text and Markdown. The upload screen shows the current size limit. A public file link must be accessible without signing in.
You can also paste a supported document URL into Website. The backend chooses the document reader, including for document URLs found in a sitemap. Those entries remain inside their website source.
Files appear directly in Knowledge. Open one to inspect its status and extracted content. If processing has not produced readable text yet, the panel explains that rather than showing an empty document as successful training.
Use the file's available actions to replace it with a newer file, retry processing or delete it. Use ··· for deletion and Tags when audience tagging is enabled. Deletion removes the associated knowledge; replacing content keeps existing usable knowledge until its replacement succeeds.
File cost is based on extracted characters, not PDF page count or upload size. Expand Training space to see usage and the largest sources. Files that cannot fit need space freed or an upgrade. See Training credits.
Use clear headings, self-contained sections and readable source documents. A scan or complex spreadsheet can extract differently from the original layout, so inspect the result and test a few answers.