jina-web-search-dsh-plugin

DeepSeek Harness includes a dependable web_search for routine lookups, yet it struggles once a query leans academic. arXiv hits frequently resolve to mirror domains rather than official papers, SSRN social-science literature is completely unsupported, and the platform offers no built-in route to text embedding, relevance reranking, or PDF formula extraction. jina-web-search-dsh-plugin closes that gap by exposing all twelve jina_* model tools natively inside dsh, letting an agent invoke them in the middle of a conversation without the user juggling a separate CLI or writing bespoke API calls.

In research-heavy workflows the plugin carves arXiv and SSRN into two purpose-built tools whose names correspond directly to their target domain, so the model routes queries straight to arxiv.org or papers.ssrn.com official links. General web search layers on time-range filtering, region and language hints, and first-rank prioritisation for authoritative Chinese-language sources such as government portals and corporate sites. Beyond retrieval the toolkit spans web-page-to-Markdown conversion, full-page screenshot capture, query expansion, text vectorisation with jina-embeddings-v5, relevance reranking via jina-reranker-v3.5, text classification, PDF chart and formula extraction, and a jina_primer tool that surfaces the host clock, public IP, and Jina account balance on demand.

The plugin suits graduate students and researchers conducting literature reviews within DSH, Chinese-language content creators who need credible, time-filtered search results, and developers assembling headless NLP pipelines that chain search through embedding and reranking in a single pass. Mainland-China network users additionally gain automatic system-proxy discovery on every call, self-healing when a VPN shifts its port, and descriptive error messages when no tunnel is active.