AI Agent Hub
Back to plugins
🤖

dsh-image-to-path

Model Inference Updated 2026.08.15

Run the following command in DeepSeek Harness:

dsh plugin install cesaryike/dsh-image-to-path

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

Run dsh plugin install cesaryike/dsh-image-to-path in the DeepSeek Harness terminal to install; source repository: https://github.com/cesaryike/dsh-image-to-path

About this plugin

When you drag an image into DSH WebUI, it is sent as a base64 attachment that text-only models reject outright, even if vision tools are installed. dsh-image-to-path intercepts drop and paste events, saves the image into the current session workspace, and inserts a plain file-path token into the input box, so the model can hand that path to any tool for processing without needing built-in vision.

The plugin auto-detects the active model capabilities: multimodal models keep the native attachment channel, and only text-only models fall through to the path scheme. Content-addressed deduplication guarantees one copy per unique image, and existing files in the workspace are referenced in place. Security checks cover magic-byte validation, server-side naming, atomic writes, a 10 MiB cap, and same-origin enforcement, with zero outbound requests.

Ideal for developers who primarily converse with text-only models, want to drop images into a session without switching models, or have not yet deployed vision tools but still need images to flow into the workflow. Just drag and paste, no prompt rewriting or model swap required.

Use Cases

  • Dropping screenshots into a text-only model chat for content analysis
  • Letting a model read workspace images via file path without a vision tool
  • Unifying drag-and-drop image interaction across multimodal and text-only models

Best For

  • Developers whose daily workflow centers on text-only models
  • Teams that need an image pipeline but have not deployed vision tools
  • Users who want images in DSH sessions without switching to a multimodal model