AI Agent Hub
Back to plugins
🖥️

dsh-file-upload-local

Client Updated 2026.09.05

Run the following command in DeepSeek Harness:

dsh plugin install lubenweimeiyoukaig/dsh-file-upload-local

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

Run dsh plugin install lubenweimeiyoukaig/dsh-file-upload-local in your terminal to install this plugin; the full source is available at https://github.com/lubenweimeiyoukaig/dsh-file-upload-local

About this plugin

DeepSeek Harness conversations had no native way to hand local files to the agent; users were limited to pasting text or bypassing the UI via API calls. dsh-file-upload-local closes that gap with a paperclip button beside the composer or a full-viewport drag-and-drop overlay, dropping documents, screenshots, and photos straight into the current session. Uploaded images render as 64-pixel thumbnails and other files as type glyphs with names, each with a hover-remove button. On send, file references are appended invisibly to the message text so the agent can act on them via the read_document tool, which pages through text files with line numbers (offset/limit) and runs tesseract.js OCR (chi_sim + eng) on images. Everything is stored under .dsh-uploads// in the session workspace, and a TTL sweep prevents unbounded disk growth.

The plugin suits anyone who wants a local, privacy-preserving way to share documents, screenshots, or photos with DeepSeek. A text-only model works out of the box thanks to the OCR path, and if the active model already supports image input the official read_image tool remains fully available as an alternative. With zero backend dependencies and an MIT license, it drops easily into self-hosted workflows or internal team distributions.

Use Cases

  • Drag a contract or report into the conversation for a text-only model to page through and summarize key clauses
  • Screenshot an error message or ID photo and let the model read it via the built-in OCR pipeline
  • Isolate uploaded files per session in a shared workspace to prevent cross-task contamination

Best For

  • Developers who rely on text-only models but still need to read screenshots or photos
  • Teams that prefer zero-backend-dependency, fully local file handling
  • Users who frequently exchange documents with agents inside DeepSeek Harness