gemini-pdf

odysseus0 的头像作者 odysseus0

使用 Gemini CLI 处理多模式文档,利用 Gemini 卓越的多模式功能。用于 PDF、扫描文档、图像较多的文档或任何需要视觉理解的文件。非常适合从复杂的布局、表格、图表、手写笔记或混合文本/图像文档中提取内容。在 PDF 处理、文档提取、“为此使用 Gemini”或当文档具有受益于多模式理解的视觉复杂性时触发。

📖 中文导读:使用 Gemini CLI 处理多模式文档,利用 Gemini 卓越的多模式功能。用于 PDF、扫描文档、图像较多的文档或任何需要视觉理解的文件。非常适合从复杂的布局、表格、图表、手写笔记或混合文本/图像文档中提取内容。在 PDF 处理、文档提取、“为此使用 Gemini”或当文档具有受益于多模式理解的视觉复杂性时触发。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: gemini-pdf description: Process multimodal documents using Gemini CLI, leveraging Gemini's superior multimodal capabilities. Use for PDFs, scanned documents, image-heavy documents, or any file where visual understanding matters. Ideal for extracting content from complex layouts, tables, diagrams, handwritten notes, or mixed text/image documents. Triggers on PDF processing, document extraction, "use Gemini for this", or when document has visual complexity that benefits from multimodal understanding.

Gemini Document Processing

Delegate document processing to Gemini CLI for superior multimodal understanding. Use when documents have visual complexity - layouts, tables, diagrams, scans, mixed content.

Workspace Restriction

Gemini CLI sandboxes file access to the current working directory. For files outside the vault, run from the file's directory:

# Use subshell to preserve cwd
(cd /path/to/files && gemini "Summarize: ./document.pdf")

The --include-directories flag exists but doesn't work reliably. Running from the target directory is the workaround.

Basic Usage

Reference file paths directly in your prompt - Gemini reads them via its file system tools:

gemini "Convert this to markdown: /path/to/document.pdf"

# Save output
gemini "Convert to markdown: /path/to/doc.pdf" > output.md

Common Tasks

Faithful Conversion:

gemini "Convert this PDF to clean markdown. Preserve all content including headers, lists, tables. Output only markdown, no commentary: /path/to/document.pdf"

Table Extraction:

gemini "Extract all tables as markdown tables: /path/to/document.pdf"

Structured Extraction:

gemini "Extract and structure as markdown:
- [list the fields you want]
- [be specific about format]
File: /path/to/document.pdf"

Diagram/Image Description:

gemini "Describe the diagrams and figures in this document: /path/to/document.pdf"

When to Use Gemini vs Other Tools

Use Gemini:

  • Scanned documents / OCR needed
  • Complex layouts (multi-column, mixed content)
  • Tables, diagrams, charts
  • Handwritten content
  • Image-heavy documents

Use pypdf/pdfplumber:

  • Simple text-only PDFs
  • Programmatic batch processing
  • When you need raw text extraction without interpretation