Docshot

docshot

**Convert documentation into terminal-style images for efficient Claude context usage**

Reduce your Claude context token usage by 56-73% while maintaining 89-97% accuracy! Based on comprehensive experiments showing that documentation stored as images uses significantly fewer tokens than raw text.

Why Use This?

When working with Claude Code or Claude API, you often need to provide large documentation files as context. This can quickly consume your token budget. Our experiments show:

• Medium Density (recommended): 67% token reduction with 97% accuracy ⭐
• High Density: 73.5% token reduction with 91% accuracy
• Low Density: 56% token reduction with 89% accuracy

Installation

# Clone or copy this directory
cd docshot

# Install dependencies with bun (required)
bun install

**Note:** This project requires Bun runtime. If you don't have it installed:

curl -fsSL https://bun.sh/install | bash

Usage

Basic Workflow

**1. Convert documentation to images:**

# Single file
bun run dev convert path/to/documentation.md

# Multiple files with glob patterns
bun run dev convert "docs/**/*.md"

This creates a `docshot` folder with your documentation split into terminal-style images.

**2. Load into Claude Code:**

bun run dev load docshot

This shows you exactly what to paste into Claude Code to load the images.

**Or directly in Claude Code:**

Review all images in docshot/

Process Multiple Files

# Convert all markdown files in a directory
bun run dev convert "docs/**/*.md"

# Convert specific patterns
bun run dev convert "src/**/*.{md,txt}"

# Files are organized in subdirectories by name
# docshot/
#   ├── README/
#   │   ├── page_001.png
#   │   └── page_002.png
#   └── API_Guide/
#       ├── page_001.png
#       └── page_002.png

Specify Output Directory

bun run dev convert docs.md --output my-images

Choose Density

``