dsh-literature-search

Running a literature search inside DSH means dealing with two very different upstreams: PubMed's official E-utilities chain (esearch, esummary, efetch, elink) and Google Scholar, which offers no official API and punishes raw HTML scraping with 429 responses. Stitching together the calls, parsing heterogeneous responses, and babysitting rate limits are engineering chores that steal time from the actual reading and synthesis.

dsh-literature-search collapses both paths into five tools — pubmed_search, pubmed_paper, pubmed_related, scholar_search, and scholar_cite — that return a unified paper shape covering title, authors, journal, DOI, MeSH terms, abstract, citation count, and PDF link. PubMed calls go through the free, stable E-utilities API; Scholar calls prefer SerpApi and fall back to HTML parsing only when no key is present, failing loudly rather than silently returning empty results. The plugin ships with zero runtime dependencies, relying solely on Node's built-in fetch.

A dedicated settings card in the DSH panel lets you manage NCBI and SerpApi keys, switch the Scholar backend, tune rate limits and result caps, and run a one-click connectivity probe. Credentials are written through the DSH credential service, keeping them out of committed config files. The plugin is well suited for researchers, grad students, and anyone embedding literature search, abstract extraction, related-paper discovery, and citation formatting (MLA, APA, Chicago, Harvard, Vancouver, BibTeX) into a DSH conversational workflow.