AI Agent Hub
Back to plugins
🧰

dsh-local-search

Web Tools Updated 2026.09.07

Run the following command in DeepSeek Harness:

dsh plugin install fangqian616/dsh-local-search

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install fangqian616/dsh-local-search in your terminal, or find dsh-local-search in the DeepSeek Harness plugin marketplace (source: https://github.com/fangqian616/dsh-local-search) to install.

About this plugin

In DeepSeek Harness, the moment a model's web-search tool hits a firecrawl keyless 403—or no search API key is configured at all—the whole capability silently dies. dsh-local-search was built to fill that gap: it needs no third-party search service, no registration, and no API key. You simply get a ready-to-use bing_search tool that the model can call on demand.

The mechanism is deliberately plain. It fetches bing.com/search through DSH's built-in web service, strips titles, links, and snippets out of the raw HTML with regular expressions, decodes entities, and filters out Bing/Microsoft self-promotion. The result comes back as an array of { title, url, snippet } objects. Beyond the DSH core runtime, there are zero npm dependencies—install and go.

It is aimed at developers who do not have a search API key and would rather not sign up for firecrawl or SerpAPI just to give their model a way to look things up. One caveat: the parser is regex-based, so a major Bing UI change or a CAPTCHA interstitial will surface as an error field; and Bing was chosen over DuckDuckGo partly for broader regional reachability.

Use Cases

  • The model needs web search but no search API key is configured
  • firecrawl keyless rejects the request with a 403 and web_search is down
  • Rapid prototyping without registering for any third-party search service

Best For

  • Individual developers without a search API key
  • Small teams that want to avoid extra dependencies for web search
  • Experimenters who want their model to search out of the box