web-multi-search

Web Multi-Search

Search the web across multiple search engines at once using async-search-scraper. Collects results from Bing, Yahoo, Startpage, Aol, and Ask, iterating through multiple result pages.

Setup

cd skills/web-multi-search
python3 -m pip install -r requirements.txt
python3 -m pip install git+https://github.com/soxoj/async-search-scraper.git --no-deps

Note: The library must be installed from the GitHub URL, not PyPI. The --no-deps flag is required because the library pins bs4 (wrong package name); the real dependencies are already in requirements.txt.

Linux (apt) fallback

If pip isn't available, install the system packages:

sudo apt-get update
sudo apt-get install -y python3-requests python3-aiohttp python3-aiohttp-socks python3-bs4

Usage

Run the search script with a query:

python3 web_multi_search.py "your search query"

Options

Flag	Default	Description
--pages	3	Number of result pages per engine
--engines	all working	Comma-separated list: bing,yahoo,startpage,aol,ask
--proxy	none	HTTP/SOCKS proxy URL
--timeout	10	HTTP timeout in seconds
--output	json	Output format: json, csv, text
--unique-urls	off	Deduplicate results by URL
--unique-domains	off	Deduplicate results by domain

Examples

# Basic search, 3 pages per engine, JSON output
python3 web_multi_search.py "python async tutorial"

# Search only Bing and Yahoo, 5 pages each
python3 web_multi_search.py "machine learning" --engines bing,yahoo --pages 5

# Unique URLs only, CSV output
python3 web_multi_search.py "OpenClaw skills" --unique-urls --output csv

# Use a proxy
python3 web_multi_search.py "privacy tools" --proxy socks5://127.0.0.1:9050

Output format

JSON output (default) returns an array of result objects:

[
  {
    "engine": "Bing",
    "host": "example.com",
    "link": "https://example.com/page",
    "title": "Page Title",
    "text": "Snippet of the page content..."
  }
]

How the agent should use this

When you need to search the web for information:

1. Run the script with the user's query.
2. Parse the JSON output to extract relevant links and snippets.
3. Use the results to answer the user or to fetch specific pages for deeper reading.

# Quick search and capture output
RESULTS=$(python3 /path/to/skills/web-multi-search/web_multi_search.py "query here" 2>/dev/null)
echo "$RESULTS" | python3 -c "import json,sys; data=json.load(sys.stdin); [print(r['link'], r['title']) for r in data[:10]]"

Available search engines

Engine	Status
Bing	Working
Yahoo	Working
Startpage	Working
Aol	Working
Ask	Working
Google	Not working (requires JS)
DuckDuckGo	Not working (CAPTCHA)
Dogpile	Not working (HTTP 403)
Mojeek	Not working (HTTP 403)
Qwant	Not working (HTTP 403)
Torch	Requires TOR proxy

Troubleshooting

• Import errors: Make sure you installed the library from the GitHub URL with --no-deps.
• Empty results / bans: Search engines may rate-limit. Increase delay or use fewer pages.
• Torch engine: Only works with a running TOR proxy at socks5://127.0.0.1:9050.