dsh-docs

Got a stack of scans, legacy PDFs, or text buried in images that you need to extract, but sending sensitive files to a cloud service feels wrong? dsh-docs exists for exactly that gap. As a document-inference plugin for the DeepSeek Harness ecosystem, it turns any document into clean, searchable text with a simple drag-and-drop, and every byte of processing stays on your local machine. Nothing is ever uploaded.

Capabilities span PDFs (including old versions where text cannot be copied), Word, Excel, and PowerPoint files, JPG and PNG images, and paper scan originals. A built-in offline OCR engine recognises letters, receipts, or even a phone photo of a restaurant menu and outputs selectable, searchable text in seconds. In-document keyword search and batch multi-file processing are included, all without an internet connection, sign-up, or subscription.

It is a natural fit for three audiences: individuals or small teams handling sensitive contracts and financial reports who insist nothing leaves their hardware; office workers digitising stacks of paper for archival; and developers who need a clean local text-extraction step before feeding a document into a RAG pipeline for an LLM. Drop the file in, and the plugin hands you the words out.