PDF Analyzer

Overview

Extract text, tables, and structured data from PDF files and convert them into usable formats. This skill handles text extraction, table detection, metadata reading, and output formatting for single or multi-page PDFs.

Instructions

When a user asks you to analyze, read, parse, or extract data from a PDF file, follow these steps:

Step 1: Identify the PDF and goal

Determine the file path and what the user wants extracted:

Full text: All readable text from every page
Tables: Structured tabular data
Metadata: Title, author, creation date, page count
Specific sections: Targeted content from certain pages
Summary: A condensed version of the document contents

Step 2: Choose the extraction method

Write a Python script using one of these libraries (prefer pdfplumber for tables, PyMuPDF for speed):

For text extraction:

import pdfplumber

def extract_text(pdf_path):
    text_by_page = []
    with pdfplumber.open(pdf_path) as pdf:
        for i, page in enumerate(pdf.pages):
            text = page.extract_text()
            if text:
                text_by_page.append({"page": i + 1, "text": text.strip()})
    return text_by_page

For table extraction:

import pdfplumber
import csv

def extract_tables(pdf_path, output_csv=None):
    all_tables = []
    with pdfplumber.open(pdf_path) as pdf:
        for i, page in enumerate(pdf.pages):
            tables = page.extract_tables()
            for table in tables:
                headers = table[0]
                rows = table[1:]
                all_tables.append({
                    "page": i + 1,
                    "headers": headers,
                    "rows": rows
                })
    if output_csv and all_tables:
        with open(output_csv, "w", newline="") as f:
            writer = csv.writer(f)
            writer.writerow(all_tables[0]["headers"])
            for table in all_tables:
                writer.writerows(table["rows"])
    return all_tables

For metadata:

import pdfplumber

def extract_metadata(pdf_path):
    with pdfplumber.open(pdf_path) as pdf:
        return {
            "pages": len(pdf.pages),
            "metadata": pdf.metadata
        }

Step 3: Run the script and format output

Execute the script, then present results in the format the user needs (plain text, JSON, CSV, markdown table, or summary).

Step 4: Handle errors gracefully

If extraction fails, try these fallback approaches:

Switch from pdfplumber to PyMuPDF (fitz)
For scanned PDFs, suggest OCR with pytesseract
For encrypted PDFs, inform the user a password is needed

Examples

Example 1: Extract all tables from an invoice PDF

User request: "Pull the line items table from invoice.pdf and save as CSV"

Actions taken:

Open invoice.pdf with pdfplumber
Detect tables on each page
Extract headers and rows
Write to invoice_items.csv

Output:

Found 1 table on page 1 with 5 columns: Item, Quantity, Unit Price, Tax, Total
Extracted 12 rows. Saved to invoice_items.csv.

Preview:
| Item            | Quantity | Unit Price | Tax   | Total   |
|-----------------|----------|------------|-------|---------|
| Widget A        | 100      | $2.50      | $25.00| $275.00 |
| Widget B        | 50       | $4.00      | $20.00| $220.00 |

Example 2: Summarize a research paper

User request: "Summarize the key findings from paper.pdf"