PDF Processing Skill

Automated workflow for converting the OXI ONE MKII PDF manual into Next.js application data.

⚠️ CRITICAL: Process Adherence

When using this skill, Claude Code MUST follow the documented process EXACTLY as written in .claude/commands/pdf-process.md.

Mandatory Rules:

1. NEVER deviate from the documented translation process
2. NEVER use MCP Codex for translation - Only use Task tool with manual-translator
3. NEVER implement "improvements" mid-process - Report them after completion
4. ALWAYS use the worker pool pattern as documented in the command file

See .claude/commands/pdf-process.md for detailed execution instructions.

Overview

This skill guides the complete PDF processing pipeline:

1. Split - Divide PDF into manageable parts (30 pages each)
2. Render - Convert pages to PNG images at 150 DPI
3. Extract - Extract text from PDFs
4. Translate - Translate English to Japanese using Claude API
5. Build - Combine data into Next.js JSON structure
6. Manifest - Generate manifest.json with metadata
7. Verify - Run dev server and confirm page 1 displays correctly

Quick Reference

Run Full Pipeline

# Run all steps
pnpm run pdf:all

Run Individual Steps

pnpm run pdf:split       # Step 1: Split PDF
pnpm run pdf:render      # Step 2: Render pages
pnpm run pdf:extract     # Step 3: Extract text
pnpm run pdf:translate   # Step 4: Translate (requires API key)
pnpm run pdf:build       # Step 5: Build JSON
pnpm run pdf:manifest    # Step 6: Create manifest
pnpm dev                 # Step 7: Run dev server for verification

Verification Step

After completing the pipeline, verify the output:

# Start dev server
pnpm dev

# Open browser and verify
# http://localhost:3100/manuals/oxi-one-mk2/page/1
# Confirm page 1 displays with translation

Requirements

• PDF file in manual-pdf/ directory
• Claude Code CLI installed (for translation subagents)
• pnpm package manager

Output Structure

manual-pdf/parts/            # Split PDF files
public/manual/pages/         # Rendered PNG images (150 DPI)
data/extracted/              # Extracted text
data/translations-draft/     # Translation drafts
data/translations/           # Final JSON files
  ├── manifest.json
  ├── part-01.json
  └── ...

Configuration

Edit pdf-config.json to customize:

• Pages per part (default: 30)
• Image DPI (default: 150)
• Translation model
• Max retry attempts

Error Handling

• Error reports saved to __inbox/
• Scripts can resume from failed step
• Retry logic for API failures (3 attempts)

Performance

Estimated time (280-page manual):

• Total: 15-30 minutes
• Translation: 10-20 minutes (most time-consuming)

Estimated cost:

• Translation: $5-10 per full manual (Claude Sonnet 4.5)

Detailed Documentation

See WORKFLOWS.md for step-by-step details.

See scripts/README-PDF-PROCESSING.md for comprehensive documentation.