Pdf Reader

• Open with pdfplumber: if page.extract_text() returns non-empty string → digital PDF.
• If empty or near-empty on multiple pages → likely scanned; warn user before continuing.
• Count pages, identify approximate layout (single-column, two-column, mixed).
• Load references/pdf-formats.md for encoding issues and text layer structure.

Reference Guide

Bundled Scripts

Dependencies: pip install pdfplumber pymupdf

Constraints

MUST DO

• Detect scanned vs. digital PDFs before extraction — warn the user clearly if no text layer is found.
• Preserve table structure: output tables as GFM markdown or CSV, never as raw whitespace-padded text.
• Include page numbers in table and image output file names to maintain traceability.
• Validate that required Python packages are installed before running; print clear pip install instructions if missing.
• Process all pages unless --pages range is specified.

MUST NOT DO

• Silently output garbage text from scanned pages — always detect and warn.
• Flatten table cells into prose or collapse multi-row headers into a single row.
• Overwrite existing output files without --overwrite flag.
• Discard images smaller than --min-width/--min-height without reporting them in the summary.
• Ignore page-level extraction errors — log failures and continue with remaining pages.

Output Template

For PDF extraction tasks, report:

1. Classification: digital / scanned / mixed; page count; layout type (single-column / two-column / mixed).
2. Text extraction: pages processed, characters extracted, any pages with empty or low-confidence output.
3. Tables found: count, page numbers, dimensions (rows × columns) for each, output file paths.
4. Images found: count, page numbers, dimensions (W×H px), color space, output file paths.
5. Errors / warnings: encoding failures, corrupt embedded objects, scanned pages detected.
6. Recommended next step: e.g., "Send extracted images to datasheet-intelligence for curve digitization."

Primary References

• pdfplumber — https://github.com/jsvine/pdfplumber
• PyMuPDF (fitz) — https://pymupdf.readthedocs.io
• PDF Reference, ISO 32000-1 — Adobe Systems