AI Agent Hub
Back to plugins
🖥️

dsh-paperclip

Client Updated 2026.08.24

Run the following command in DeepSeek Harness:

dsh plugin install runfali/dsh-paperclip

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install runfali/dsh-paperclip in your terminal to install; source code at https://github.com/runfali/dsh-paperclip , then restart dsh web to apply.

About this plugin

DeepSeek Harness ships without a native file entry point, and the model has no way to touch the documents you want it to read. dsh-paperclip adds a borderless 📎 button and an upward popover to the composer: picked files are persisted immediately, server paths are appended to the message on send, and a small ▴ badge lets you expand the pending list or remove entries with ×. Supported formats cover txt, json, conf, md, ini, plus pdf, docx, and xlsx (the last three activate after installing their optional dependencies). All uploads land in a content-addressed, deduplicated directory out of the box—no extra folder config needed.

Beyond uploading, the plugin registers a read_document tool so the model can page through any of those formats using offset and limit parameters. The default window is 2 000 lines with a 64 MB per-file cap, and each xlsx sheet keeps at most 200 rows. Text formats work zero-dependency out of the box; pdf, docx, and xlsx light up once pdf-parse, mammoth, and xlsx are added inside the plugin directory, and a friendly hint is returned when they are missing rather than a hard error.

Every behavior is governed by a single toggle under Settings → General: switch it off and the button disappears, uploads are blocked, no extra tabs to manage. The UI consumes DSH design tokens so it renders correctly in both light and dark themes, and the whole bundle loads as a standard Cordis plugin with zero changes to DSH source code—uninstall restores the stock experience instantly. It is built for users who regularly feed configs, spreadsheets, or reports into a conversation and want to keep the core codebase untouched.

Use Cases

  • Feeding config files, reports, or spreadsheets into the model for analysis
  • Paging through large docx or xlsx files via the read_document tool
  • Uploading multiple files mid-conversation and having the model process them at once

Best For

  • Harness users who frequently feed documents into conversations
  • Plugin developers who want to extend features without touching core source
  • Business teams that need the model to read multiple file formats