dsh-kb-rag
Indexes your PDFs or Zotero library into one local SQLite knowledge base and returns passages with section, page and DOI; embedding and reranking run locally.
Install
- Click Plugins in the sidebar, then Add plugin at the top right
- Enter the package name dsh-kb-rag, press Install
- Restart when asked
dsh plugin --profile web add dsh-kb-rag- Restart
dsh web - No dsh yet?
npm i -g @deepseek-ai/dsh, needs Node 22.19+
What the code shows
10 tools and a /kb command. kb_ingest and kb_zotero index PDF, TXT, MD and DOCX files into <workspace>/.kb; kb_search and kb_rag return cited passages; kb_scope, kb_stats and kb_status manage scope and jobs. It needs Python with PyMuPDF, faiss, sentence-transformers and torch, and downloads about 1.2 GB of bge models from Hugging Face (or hf-mirror.com) on first use. No key is needed. Your documents stay local. The one outbound feature is kb_fetch, which downloads PDFs by DOI or arXiv ID from arxiv.org, doi.org, publisher sites, Unpaywall and Crossref into ~/.kb-rag/downloads. Two things to know: kb_fetch copies your proxy settings into its output, so a proxy password would reach the conversation, and it passes identifiers to its download script unchecked. No install scripts; an optional installer runs pip and installs pnpm. It reads papers you already have and does not search databases, so it pairs with dsh-ai4scholar. MIT. Maintained by Breeze136, not by us.
Listing is not endorsement. A plugin runs with your permissions: look at its source before installing.