Back to directory

moki-library-for-dsh

In a curated listMaintenance: Active

jkzeng-cell/moki-library-for-dsh

Manage and preview local files inside DSH (images, PDF, Office, archives, 3D), edit Word/Excel/PowerPoint drafts, and send an exact document selection to the agent.

View on GitHubHomepage
$ dsh plugin --profile desktop add "dsh-moki-library@file:<absolute-archive-path>" --allow-build=sharp

33

stars

1

forks

JavaScript

Language

MIT

License

2026-11-11

Created

2026-11-15

Last push

A MIT-licensed community DSH plugin that puts a local file library, multi-format previews and Office draft editing beside the DeepSeek Harness conversation and passes exact selections to the agent.

DSH integration

Native runtime

Author-claimed

Safety audit

Unaudited

Last verified

2026-11-15

License

MIT

01What can it help you accomplish?

  • Keep project documents in a local file library beside the DSH conversation

    A local file library with virtual folders, tags, search and list/card views, plus built-in previews for images, audio/video, PDF, Office, text, archives, fonts and common 3D formats

    DSH users who want their project documents and AI-generated files in the same window as the chat

  • Edit Word, Excel and PowerPoint from a precise selection handed to the agent

    Isolated drafts shown in the editor, reviewed change by change, then saved back to the original file or as a copy; selection references carry bounded content and exact locations

    Analysts, PMs and engineers who want an AI edit scoped to one passage, cell range or slide object instead of a whole document

  • File AI-generated task files without moving them

    New DSH task files mapped under “DSH task files” with identical paths deduplicated, plus Library buttons on generated-file cards and Add to MOKI Library on outputs

    DSH users who generate documents inside a conversation and need them catalogued in the library

02How to install into DeepSeek Harness

Prerequisites

  • DSH Desktop `>=0.2.2-rc.1 <0.3.0` — tested on Windows with DSH Desktop 0.2.2-rc.1; other desktop versions and operating systems need further verification
  • Node.js 22.19+ on the runtime that DSH Desktop uses
  • Network access for the first installation, to fetch runtime dependencies
  • A locally installed Edge or Chrome for native preview windows; FFmpeg is optional for additional audio/video decoding

Installation steps

  1. 01

    Option 1 (agent task): open DSH Desktop → New conversation with a model whose conversation can run local terminal commands, paste the whole install prompt from the README, and answer DSH’s download/terminal prompts

  2. 02

    Option 2 (manual): open DSH Desktop once to initialise its desktop profile, quit it fully, and download dsh-moki-library-0.1.73.tgz from the v0.1.73 release

  3. 03

    Run `dsh plugin --profile desktop add "dsh-moki-library@file:<absolute-archive-path>" --allow-build=sharp` with the desktop’s own CLI (resources/runtime/cli/bin/dsh.cmd on Windows) and the archive’s absolute path; keep the desktop’s existing DSH_HOME and existing plugins, configuration and library data

    $ dsh plugin --profile desktop add "dsh-moki-library@file:<absolute-archive-path>" --allow-build=sharp

  4. 04

    Fully quit DSH and reopen it, confirm MOKI Library is enabled in the plugin list, then open it from the sidebar

Verify the integration

  • `dsh plugin --profile desktop list dsh-moki-library --depth 0` — confirm that desktop lists version 0.1.73 (or a newer retained version)
  • Preview a sample file: select text or cells in a supported editor and press Alt+D to check that the conversation entry works

03DSH integration and capability boundaries

DSH integrationNative runtime

Native DSH Desktop plugin: installed into the DSH desktop profile with the official `dsh plugin` CLI (`dsh plugin --profile desktop add … --allow-build=sharp`) and opened from DSH's own sidebar

  • Multi-format preview

    mapped local files and library items→image, audio/video, PDF, Office, text, archive, font and common 3D previews — some formats need conversion or have documented limitations

  • Office draft editing with review

    a selected passage, cell range, slide text or slide object→an isolated draft with highlighted AI edits, Accept / Undo latest edit / Discard all, then save back to the original file or save a copy

    Accept saves the current result to the original file immediatelyhighlights are temporary decorations and are never written into document formattingDiscard all restores the draft without overwriting the original file
  • Exact selection hand-off to DSH

    text, PDF page or region, spreadsheet range, slide object or image region→a selection reference with bounded content and exact location, sent as `context → edit_selection` for supported edits

    default context text is bounded to 4,000 characters and composer quotes to 1,000selection references expire after ten minutes and reject stale or reused edits
  • Local-first library storage

    library catalog, generated files and editing drafts→files kept under `DSH_HOME/moki-library` or the user’s `.dsh/moki-library` directory

    mapping an existing file keeps its original path — nothing is moved or copied

04Who is it for? When not to use it?

Good for

  • DSH users who want their project documents and AI-generated files in the same window as the chat
  • Analysts, PMs and engineers who want an AI edit scoped to one passage, cell range or slide object instead of a whole document
  • DSH users who generate documents inside a conversation and need them catalogued in the library

Not for

  • Windows with DSH Desktop 0.2.2-rc.1 is the tested baseline; the declared compatibility range is >=0.2.2-rc.1 <0.3.0, and other desktop versions or operating systems still need verification.
  • Node.js 22.19+ is required, native preview windows need a locally installed Edge or Chrome, and FFmpeg is optional for additional audio/video decoding.
  • Content selected for AI collaboration or attached to a conversation is provided to DSH and its configured model/tools, so local storage is not a promise of offline AI processing.

05Compatibility, maintenance and safety notes

  • Windows with DSH Desktop 0.2.2-rc.1 is the tested baseline; the declared compatibility range is >=0.2.2-rc.1 <0.3.0, and other desktop versions or operating systems still need verification.
  • Node.js 22.19+ is required, native preview windows need a locally installed Edge or Chrome, and FFmpeg is optional for additional audio/video decoding.
  • The first installation needs network access to fetch runtime dependencies; the README ships a SHA256 checksum manifest alongside the .tgz release asset.
  • Content selected for AI collaboration or attached to a conversation is provided to DSH and its configured model/tools, so local storage is not a promise of offline AI processing.
  • Legacy XLS, macro-enabled and password-protected documents are not listed as supported embedded editing formats, and preview support does not imply editing support or complete fidelity.
  • Conversation attachments are capped at 32 MB per file, text over 5 MB is read-only paging, and whole-file decoding uses a 512 MiB guard (256 MiB expanded-data budget for Office structured previews).
2026-11-112026-11-13v0.1.73

MIT · latest release v0.1.73 (2026-11-13T11:03:12Z)

06Frequently asked questions

How do I install MOKI Library into DSH Desktop?

Two documented routes. The agent route: open DSH Desktop → New conversation with a model that can run local terminal commands, paste the whole install prompt from the README and answer its prompts. The manual route: open DSH Desktop once, quit it, download dsh-moki-library-0.1.73.tgz from the v0.1.73 release, then run `dsh plugin --profile desktop add "dsh-moki-library@file:<absolute-archive-path>" --allow-build=sharp` with the desktop’s own CLI. Afterwards fully quit and reopen DSH and confirm the plugin is enabled in the plugin list.

Which DSH and Node.js versions does the plugin support?

Declared compatibility is DSH `>=0.2.2-rc.1 <0.3.0` with Node.js 22.19+. The maintainer tested it on Windows with DSH Desktop 0.2.2-rc.1 and notes that other desktop versions and operating systems need further verification; native preview windows use a locally installed Edge or Chrome.

How does a selection reach the agent?

Select text, cells, a slide object or an image region and use the whale beside the selection, Alt+D, or Add selection to conversation. The reference carries the exact location (text offsets, sheet/range address, slide object, page number or region) and bounded content — 4,000 characters by default, 1,000 for composer quotes — and supported edits use `context → edit_selection`. References expire after ten minutes and stale or reused edits are rejected.

Where is my library data stored, and does the model see my files?

The catalog, generated files and editing drafts are stored locally under `DSH_HOME/moki-library` or the user’s `.dsh/moki-library` directory, and mapping a file keeps its original path. Content selected for AI collaboration or attached to a conversation is provided to DSH and its configured model/tools, so local storage is not a promise of offline processing. Attachments go through DSH’s native pipeline at up to 32 MB per file.

Can I edit DOCX / XLSX / PPTX without risking the original?

Editing happens on isolated drafts: AI edits are highlighted, and you Accept, Undo latest edit or Discard all before saving back or saving a copy. Accept saves to the original file immediately, Discard all does not overwrite the original, and highlights are temporary decorations that never become document formatting. Legacy XLS, macro-enabled and password-protected documents are not listed as supported embedded editing formats.

08Data and sources

  • Author-claimedgithub.com561e92bbf735…

    MOKI Library is a community plugin for **DeepSeek Harness (DSH)**.

  • Author-claimedgithub.com561e92bbf735…

    Supported edits use `context → edit_selection`; complex edits can use the Office tools.

This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-11-15. Found an error? Submit a correction.

🏆

Best DeepSeek Harness Plugins

Twelve plugins worth installing first — picked from the whole catalog, across every category.

DSH Plugins is an independent community directory of DeepSeek Harness plugins. Not affiliated with or endorsed by DeepSeek. Third-party plugins are not security-audited — review the source before installing.

New DeepSeek Harness plugins, weekly. No spam.