ddgo-search 1.0.5

A resilient, token-efficient CLI wrapper around the DuckDuckGo Search (ddgs) Python library.

ddgo-search

A highly resilient, token-efficient, and feature-rich Command Line Interface (CLI) wrapper around the DuckDuckGo Search (ddgs) Python library. It features built-in proxy rotation, rate-limiting, custom token-saving ASCII rendering, webpage extraction, and direct content fetching.

---

✨ Features

• 🌐 Comprehensive Query Support: Subcommands for text, images, videos, news, books, and web page extraction/fetching.
• 🔄 Resilient Proxy Rotation: Accepts single proxy URLs, comma-separated lists, or files containing lists of proxies. Automatically rotates proxy servers sequentially on failure.
• ⏱️ Process-Safe Rate Limiting: Randomised delays (between 1.5s to 3.0s) are tracked globally using a system-level temporary file to ensure safe execution even when multiple commands are run concurrently.
• ⚡ Direct Web Fetching (fetch): Inspired by Charmbracelet's crush tool. Directly fetches and converts webpages using httpx, BeautifulSoup, and markdownify into beautiful plain text, markdown, or HTML, with auto-truncation limits (e.g., 100KB) to preserve context windows.
• 📊 Token-Efficient ASCII Layouts: Formats console lists using space-padded ASCII dividers rather than heavy Unicode grids to save context window tokens when using LLM agents.

---

🚀 Installation

It is recommended to install and manage ddgo-search using uv.

1. Global Installation (As a CLI Tool)

To install it globally so that the ddgo-search command is available from anywhere:

# Install directly from PyPI
uv tool install ddgo-search

# Or install from GitHub
uv tool install git+https://github.com/twn39/ddgo-search.git

2. Local Development Installation

If you want to clone the repository and run it locally:

# Clone the repository
git clone https://github.com/twn39/ddgo-search.git
cd ddgo-search

# Install dependencies and sync virtual environment
uv sync

---

📖 CLI Usage

Invoke ddgo-search directly using uv:

uv run ddgo-search [GLOBAL-OPTIONS] COMMAND [ARGS]...

Global Options

These options must be passed before any subcommand:

• -p, --proxy TEXT: Proxy URL, comma-separated list of proxy URLs, or file path containing proxies (one per line). Falls back to the DDGS_PROXY environment variable.
• -t, --timeout INTEGER: Request timeout in seconds (default: 10).
• --verify / --no-verify: Enable/disable SSL certification verification (default: --verify).
• -r, --max-retries INTEGER: Maximum retries upon server failures or timeouts (default: 3).

---

Commands

1. Text Search (text)

Search the web for text results with custom formatting.

uv run ddgo-search text "artificial intelligence" --format plain
uv run ddgo-search text "python programming" --format table --max-results 5

2. Image Search (images)

Query and filter DuckDuckGo images.

uv run ddgo-search images "cute kittens" --size Large --color Monochrome
uv run ddgo-search images "space nebula" --layout Wide --format json

3. Video Search (videos)

Search for videos with specific duration, resolution, or license filters.

uv run ddgo-search videos "golang tutorial" --resolution high --duration short

Note: The third-party library standard resolution uses the spelling "standart". The CLI enum automatically handles this mapping.

4. News Search (news)

Query recent news.

uv run ddgo-search news "climate change" --timelimit w --format csv

5. Book Search (books)

Search DuckDuckGo books.

uv run ddgo-search books "machine learning" --max-results 10

6. Web Page Extract (extract)

Fetch and extract webpage content using DuckDuckGo's internal extraction backend.

uv run ddgo-search extract "https://example.com" --format text_markdown

7. Direct Web Page Fetch (fetch)

Directly fetch a URL via httpx and convert its HTML content locally to Markdown, clean text (excluding scripts, styles, headers, footers), or HTML. Includes auto-truncation.

# Direct fetch and convert to Markdown
uv run ddgo-search fetch "https://example.com" --format markdown

# Direct fetch and extract readable plain text
uv run ddgo-search fetch "https://example.com" --format text

# Direct fetch and write to file
uv run ddgo-search fetch "https://example.com" --format markdown --output doc.md

# Set custom truncation limit (e.g. 5KB)
uv run ddgo-search fetch "https://example.com" --max-size 5120

---

🤖 Codex Subagent & Skill Integration

You can integrate ddgo-search as a custom subagent and skill in Codex (or other supported agents like Antigravity, Crush, and Claude Code) to handle all web search and page fetching tasks.

1. Global Installation (Recommended)

To allow Codex to automatically call the ddgo-search subagent across all your projects:

1. Install the CLI globally so it is available from any workspace directory:

   # Install globally using uv
   uv tool install .
   

2. Install the Skills and Subagent Configurations using the CLI installer:

   # Install skills globally (for Codex, Antigravity, Crush, and Claude Code)
   ddgo-search skills install
   
   # Install the subagent configuration globally (for Codex)
   ddgo-search agents install --target codex
   

2. Project-level Installation

If you only want this subagent and skill available inside this project directory:

1. Ensure the project is trusted in your ~/.codex/config.toml:

   [projects."/absolute/path/to/ddgo-search"]
   trust_level = "trusted"
   

2. Install the skills and agents locally:

   # Install skills locally
   ddgo-search skills install --local
   
   # Install the subagent configuration locally
   ddgo-search agents install --target codex --local
   

3. The packaged configurations are located at:
   • Skill: SKILL.md
   • Subagent Config: ddgo-search.toml

3. Usage

Once the skill and subagent are installed globally, Codex can delegate searches automatically when prompted. You can trigger it explicitly by prompting:

"使用网络查找关于..."

---

🧪 Development & Testing

Run the comprehensive unit test suite:

uv run pytest

Our tests mock the ddgs.DDGS library as well as network activities to ensure the test suite runs instantly, robustly, and offline.