pdf-extraction

PDF Text Extraction

Extract text from a local PDF with the Asta remote OCR API. The command returns markdown that preserves document structure, headings, formatting, tables, lists, and emphasis.

Quick Start

# Print extracted markdown to stdout
asta pdf-extraction remote paper.pdf
 
# Save extracted markdown to a file
asta pdf-extraction remote paper.pdf

Arguments

• <pdf>: local PDF path (required; the file must exist and be readable)
• -o / --output: markdown output path (default: stdout)
• --start-page: first page to process, using zero-based page numbering (default: 0)
• --max-pages: maximum number of pages to process (default: 50)
• --images / --no-images: save embedded images alongside the markdown (default: --no-images)

The command requires an authenticated Asta session.

Procedure

1. Confirm the input and output

Identify the local PDF and where the user wants the markdown saved. Prefer an explicit -o path so the result is available for later work. The command creates missing parent directories for the output file.

2. Choose a page range

For PDFs of 50 pages or fewer, use the defaults:

For longer PDFs, process consecutive ranges in separate commands. Page numbers are zero-based, so the second 50-page range begins at page 50:

Keep the part numbers zero-padded so the files sort in page order.

3. Extract images when needed

Use --images when figures, diagrams, or other embedded images are important:

The images are saved in the output file's directory and referenced by filename in the markdown. Always provide -o with --images; without it, images are written to the current directory.

4. Verify the result

After extraction, check that:

• The output file exists and is not empty.
• The first and last requested pages are present.
• Headings, tables, equations, and multi-column text are readable.
• Image references resolve when --images was requested.

If the document was processed in parts, retain the original PDF and page ranges with the extracted files so their order and provenance remain clear.

Common Workflows

Extract a scanned PDF

Extract only a selected range

To extract pages 21-30 as displayed in a PDF viewer, start at zero-based page 20 and request 10 pages:

Extract text for an Asta document index

Save the markdown under the user's dataset or document directory, then use the asta-documents or local-paper-index skill to index it. Preserve a link to the source PDF in the index metadata when possible.

Troubleshooting

Authentication error

Confirm that the user is logged in to Asta, then retry the same command.

Remote OCR API error

The command reports the HTTP status and response from the service. Verify the network connection and retry. If a large request repeatedly fails, reduce --max-pages and process smaller page ranges.

Empty or incomplete output

• Confirm that --start-page is within the document.
• Check whether the requested range extends beyond the final page.
• Retry with fewer pages.
• For image-only or unusually complex pages, inspect the corresponding page in the source PDF before accepting the extraction.

Images were written to the wrong directory

Run the command with both --images and an explicit -o path. Images are saved next to the output file; if no output file is provided, they are saved in the current directory.

When to Use This Skill

Use PDF extraction when:

• The user wants text or markdown extracted from a PDF.
• The PDF is scanned or contains image-based text that requires optical character recognition (OCR).
• The document has tables, equations, multiple columns, or complex formatting.
• Extracted text is needed for indexing or downstream analysis.

Do not use it when:

• The input is an image rather than a PDF.
• The user only needs paper metadata or short searchable snippets.
• The user needs real-time or streaming extraction.