Sqhao-O/dsh-docs
dsh-docs gives a DeepSeek Harness agent on-machine document intelligence: hand it a PDF, Word, Excel, PowerPoint, Markdown, HTML, CSV or image and get clean Markdown, plain text or structured JSON, with a fully offline OCR pipeline so no document ever leaves your disk.
dsh-docs gives a DeepSeek Harness agent real document intelligence entirely on your own machine. Hand it a PDF, Word, Excel, PowerPoint, Markdown, HTML or CSV file and get back clean Markdown, plain text or structured JSON; hand it a scanned page or image and a fully offline OCR pipeline reads it. There is no Docker, no HTTP service, no API keys, and no document ever leaves your disk. The README provides a one-prompt install that the Harness agent runs end to end - installing the published npm package, downloading the pinned offline OCR runtime (Windows x64) with SHA-256 and manifest verification, editing cordis.patch.yml to add the plugin config, then verifying with dsh --profile web --dump-config - and the session workspace is readable automatically with extra persistent directories added through allowedLocalRoots.
Install
dsh plugin --profile web add dsh-docCopied verbatim from the README's one-prompt install (step 1) and its plain install block; the README names the published package and plugin id dsh-doc whose tools use the dshdoc_* prefix.
Compatibility
Prerequisite is a working dsh CLI on Node ^22.19 or >= 24. It ships a pinned, self-contained Python + Xberg runtime with offline Tesseract language data (English and Simplified Chinese) delivering full PDF/Office/OCR coverage on Windows x64 out of the box; the native Xberg Node binding is a lightweight non-OCR fallback on any platform, selected with engine: node and defaultOcr: false. File reads stay confined to folders you explicitly authorize. English README with a Simplified Chinese translation.
Details
- Repo: Sqhao-O/dsh-docs
- Category: Files & Data
- Stars: 16
- Version: GitHub repo Sqhao-O/dsh-docs (16★, MIT); npm package dsh-doc at 0.1.1 (registry check 2026-10-13; the registry repository field resolves back to this repo). The published package and plugin id use the dsh-doc spelling, renamed from the initial dsh-docling / docling_* release.
Alternatives
Aidenwu0209/dsh-Unlimited-OCR-Skill · amethystluna/logicprobe · bill9109/dsh-101