chinng-inta/dsh-web-search-searxng

SearXNG-backed web search for DSH: your own metasearch instance, no API key and no model turn spent per search.

Moves web search from DeepSeek's own provider to a self-hosted SearXNG instance by calling that instance's /search?format=json endpoint. The README's stated difference is cost shape: the shipped provider issues a full Messages request with the native web_search tool for every search, paying latency and generated tokens, while this one is a plain retrieval call with no API key and no model turn.

Tools & Capabilities ★ 5
View on GitHub ↗

Install

dsh plugin --profile web add dsh-web-search-searxng

The npm line, and the README argues for it over the github: form it also lists: a git dependency ships no lib/, so it builds itself through a prepare script, which pulls the dev toolchain and takes roughly twenty seconds instead of two. Remaining steps are pointing the provider row at your instance and restarting.

Compatibility

SearXNG does not enable its JSON API by default — the instance's settings.yml has to be changed, and the README warns against pointing at a public instance that rate-limits aggressively. This is an implementation package: it registers a WebSearchProvider into ctx.web and registers no model-facing tool, because @deepseek-ai/dsh-tool-web keeps owning web_search, its schema and its result card. The config keys are declared volatile, so from dsh 0.1.7-rc.2 an edit updates the running plugin in place without a restart, but is not persisted across boots.

Details

Alternatives

Dingpenghui-good/dsh-web-search-serper · mcxianyujun/dsh-web-search-anysearch · gilsaia/dsh-web-search-doubao

More plugins in Tools & Capabilities

Browse more in Tools & Capabilities

Guides for Tools & Capabilities plugins