/pdf-to-page <pdf-path>
/pdf-to-page +Attachments/architecture-spec.pdf
/pdf-to-page +Attachments/meeting-presentation.pdf --title "Custom Title"

Question: "What level of analysis do you want for this PDF?"
Header: "Analysis"
Options:
  1. "Sonnet - Concise" (Recommended)
     Description: "Faster processing, concise overview. Best for straightforward documents."
  2. "Opus - Comprehensive"
     Description: "Deeper analysis with detailed YourOrg context. Best for complex technical documents."

from docling.document_converter import DocumentConverter
import json

# Initialize converter
converter = DocumentConverter()

# Process PDF
result = converter.convert(pdf_path)

# Extract structured outputs
markdown_content = result.document.export_to_markdown()
doc_dict = result.document.export_to_dict()
json_data = json.dumps(doc_dict, indent=2)

# Get statistics
page_count = len(result.document.pages)
table_count = len(result.document.tables) if hasattr(result.document, 'tables') else 0
image_count = len(result.document.images) if hasattr(result.document, 'images') else 0

# Save to temp directory for verification
output_dir = Path(pdf_path).parent / "docling_output"
output_dir.mkdir(exist_ok=True)

md_file = output_dir / f"{Path(pdf_path).stem}_docling.md"
json_file = output_dir / f"{Path(pdf_path).stem}_docling.json"

with open(md_file, 'w', encoding='utf-8') as f:
    f.write(markdown_content)

with open(json_file, 'w', encoding='utf-8') as f:
    f.write(json_data)

Task: Analyze PDF visually and extract organization-specific entities

1. VISUAL ANALYSIS:
   - Read the PDF file (Claude can view PDFs visually)
   - Identify all diagrams, charts, graphs, and images on each page
   - Create descriptive names for each image based on content
   - Note which images should be extracted as PNGs
   - Assess image types (architecture diagrams, flowcharts, screenshots, photos, logos)
   - Count pages with significant visual content

2. YourOrg ENTITY EXTRACTION:
   - Identify YourOrg projects mentioned (MyDataIntegration, ModernizationProject, NewProductLine, MaintenanceSystem, etc.)
   - Find YourOrg people and roles (architects, PMs, stakeholders)
   - Identify YourOrg systems and technologies (SAP, AWS, MROPlatform, etc.)
   - Extract YourOrg organizations (VendorA, SAP, VendorB, VendorC, etc.)
   - Find dates, deadlines, milestones
   - Identify action items, decisions, recommendations

3. CONTENT CLASSIFICATION:
   - Determine document type (spec, presentation, report, guide, training, compliance)
   - Identify key topics and themes
   - Note any sections that need special formatting
   - Assess document purpose and audience

Return:
- Image inventory with page numbers, descriptions, and suggested filenames
- YourOrg entity list (projects, people, systems, organizations)
- Document classification and topics
- Action items and key dates

---
type: Page
title: {title}
created: {YYYY-MM-DD}
modified: {YYYY-MM-DD}
tags: [pdf-import, {auto-generated tags from agent classification}]
source: "{PDF filename}"
sourceType: PDF
processedWith: docling
---

# {title}

> **Source**: [[{PDF filename}]]
> **Imported**: {YYYY-MM-DD}
> **Pages**: {page count from docling}
> **Tables**: {table count from docling}
> **Processed with**: [docling](https://github.com/docling-project/docling) + Claude {Sonnet|Opus}

## Overview

{2-3 paragraph summary from agent's classification}

## Document Information

| Property | Value |
|----------|-------|
| **Document Type** | {type from agent classification} |
| **Source File** | `{PDF filename}` |
| **Pages** | {count from docling} |
| **Tables Detected** | {table count from docling} |
| **Images Detected** | {image count from docling} |
| **Processing Time** | {docling processing time} seconds |
| **Creation Date** | {if found in PDF metadata} |
| **Author** | {if found in PDF metadata} |

## Content

{Use docling's markdown output directly - already has proper structure}

{The markdown_content from docling includes:}
- Heading hierarchy (##, ###)
- Bullet and numbered lists
- Paragraphs with preserved reading order
- Image markers (<!-- image -->) at correct positions
- Tables (if detected)
- Code blocks and formulas (if detected)

{Replace <!-- image --> markers with actual embedded images if available}

## Diagrams & Visual Content

{For each extracted PNG from Phase 4:}

### {Image description from agent} (Page {N})

![[{image filename}.png]]

{Brief description from agent's visual analysis}

---

## Tables

{If docling detected tables, they're already in the markdown content above}
{If tables are in images, note from agent's visual analysis}

## Key Information

### Projects Referenced

{List from agent's YourOrg entity extraction with [[Project]] links}

### People Mentioned

{List from agent's YourOrg entity extraction with [[Person]] links}

### Systems & Technologies

{List from agent's YourOrg entity extraction}

### Organizations Referenced

{List from agent's YourOrg entity extraction with [[Organisation]] links}

### Dates & Deadlines

| Date | Context |
|------|---------|
{Dates from agent's analysis}

## Action Items

{Action items from agent's analysis}

- [ ] {item 1}
- [ ] {item 2}

## Related

- {Auto-link to related Project, Meeting, or other notes from agent's entity extraction}
- [[{PDF filename}|Original PDF]]

## Processing Notes

**Extraction Method**: Docling v2.66.0 with MLX acceleration
**Processing Time**: {docling time} seconds
**Visual Analysis**: Claude {Sonnet|Opus} agent
**Analysis Depth**: {Concise|Comprehensive}
**Quality**: {any notes from agent about extraction quality}

{Any extraction notes, quality issues, or recommendations from agent}

## PDF Import Complete

**Created**: `Page - {title}.md`

**Processing Statistics**:
- Processing time: {docling time}s (docling) + {agent time}s (visual analysis) = {total}s
- Pages processed: {count}
- Tables detected: {count}
- Images detected: {count}

**Extracted Images**: {count} PNG files
- `{image1.png}` - {description}
- `{image2.png}` - {description}
- ...

**YourOrg Entities Identified**:
- Projects: {count} ({list})
- People: {count} ({list})
- Systems: {count} ({list})
- Organizations: {count} ({list})

**Content Classification**:
- Document type: {type}
- Key topics: {topics}
- Action items: {count}

**Suggested Actions**:
1. Review extracted content for accuracy
2. Verify YourOrg entity links are correct
3. Add additional tags if needed: {suggestions}
4. Create tasks from action items: {task suggestions}
5. Link to related projects: {suggestions}

pip install docling

brew install poppler  # Install if not available
pdfimages -png input.pdf output-prefix

pdftoppm -png -r 150 -f 3 -l 3 input.pdf output-prefix  # Page 3 only

# Docling doesn't handle password-protected PDFs
# Prompt user to unlock PDF first or use:
# qpdf --decrypt --password=<pwd> input.pdf output.pdf

# Docling handles OCR automatically with ocrmac (on macOS)
# Uses Vision framework for high-quality text extraction
# For better results, can use VLM pipeline:
# converter = DocumentConverter(
#     format_options={"pdf_options": {"use_vlm": True}}
# )

User: /pdf-to-page +Attachments/MROPlatform-Architecture-Overview.pdf