📄 Reference appBuilt in
Document reader
Reading PDFs and office files a piece at a time.
What you'll learn
- The four reading abilities.
- How to read a long file safely.
- Read with the reader, make with Python.
Tallgeese reads files out of the box, so a skill can say "read the PDF" and know it works.
| Ability | Use it for |
|---|---|
document_info | First, on a long file: its pages, sections or sheets, an outline, and how many reads it will take. |
read_document | The text of a file, a part at a time. Each result ends by saying how to continue. Choose pages like "5", "3-8" or "10-". |
read_sheet | Spreadsheet cells as a table. Narrow it with sheet and range such as "A1:F200". |
render_pdf_page | One PDF page as an image, for a scan, a chart or a layout that text doesn't capture. |
Formats: PDF, Word, PowerPoint, Excel and CSV, OpenDocument, Hangul, RTF, EPUB, and text or Markdown in any encoding. It works out the real format, not just the file's extension.
1. Call document_info on the file first. Tell the person how long it is.
2. Read with read_document, passing the cursor each result gives,
until it says the file has ended.
3. If a page is a scan or a chart, use render_pdf_page for that page only.
4. For a spreadsheet, use read_sheet with a range. Never read a whole
large sheet at once.
Read with the reader, make with Python
The Python skill tells the assistant: "to read a Word, PowerPoint, PDF or old Office file, the document reader is better than a script". Use Python to make files, and the reader to read them.