dsh-web-polysearch

When you reach for a web search inside DeepSeek Harness, you are usually limited to whatever one engine happens to return — a single index, a single region, a single quality bar. dsh-web-polysearch removes that ceiling. It fires requests at five backends simultaneously (DeepSeek, DuckDuckGo, Exa, Google Custom Search, and Bing Web Search) via Promise.allSettled, merges the responses, deduplicates them by URL, and hands the model one consolidated, structured answer. Three of the five engines are keyless out of the box, so there is no account hunting before your first query; Google and Bing are optional add-ons for stricter precision.

Beyond multi-engine search, the plugin supports a direct-fetch mode: pass a URL instead of a query and you get the full page body back. An optional top-N content fetch pulls the body text of the best results in parallel (up to five), giving the model real context rather than a snippet. All of this is configurable from the bilingual settings panel in the DSH Web GUI — toggle sources as cards, slide maxResults, flip fetchContent on or off — with no config-file edits required.

This tool fits anyone who leans on web search in DSH and wants cross-engine coverage without betting on a single provider. It is lightweight, serverless, and drops straight into your DSH profile. No extra infrastructure, no dependency tree beyond the plugin itself — just broader, more reliable search results whenever your workflow demands them.