Skip to main content
You can ask Athena to list and read files embedded in an uploaded Word document or attached inside a PDF. Use this when a report or document package contains supporting material that is not part of its main text. You can inspect those files without first extracting and uploading each one yourself.

How to ask Athena to read a file attached inside a PDF

  1. Ask for the attached files
  2. Athena reads the file
  3. See what is attached
  4. Ask for one file
  5. Athena reads the attachment
  6. Check what was read

​
Discover supporting files

  1. Open a session and reference the uploaded Word document or PDF you want to inspect.
  2. Ask Athena to read the document or list its attached files. There is no separate list action: the first plain read of an uploaded PDF or Word file also finds the files attached inside it. For example: “List the attached files in this PDF before summarizing it.”
  3. Review the returned names and paths. The tool card does not show the listing, so you see the names in Athena’s reply. Athena lists up to 50 files. The main document read identifies attachments separately; it does not include their contents automatically.
  4. Ask Athena to read the supporting file you need by its returned name or path. For example: “Read the embedded workbook and summarize its assumptions.”
  5. Ask Athena to distinguish the main document’s statements from the supporting file’s content in its response.
Use the returned attachment path when names are ambiguous. The reader uses the containing file and that path to locate the attachment. You do not need a separate asset for every readable attachment.

​
Read longer supporting files

  1. Ask Athena whether the returned attachment content is complete.
  2. If the text is paged, ask it to continue reading before drawing conclusions.
  3. If an embedded workbook is truncated, upload that workbook as its own asset to read beyond the embedded-file window. Embedded sheets stop at 10,000 rows by 200 columns, and each attached file can be up to 32 MB.

​
Limits

You need access to the containing asset. Word attachment discovery requires a stored uploaded Word package. Naming a document created in the editor with a .docx extension does not give it embedded files. Readable attachment types include PDF, Word, Excel, PowerPoint, email, and plain text. Unsupported formats, password protection, unreadable packages, file-size limits, or upload-security policy can prevent a read. Returned content is text rather than an editable attachment view. A parent read lists up to 50 embedded files. An embedded workbook read covers up to 10,000 rows and 200 columns per sheet. For an embedded PDF, OCR covers at most five pages without a text layer. Check partial-read notices before relying on a summary.