AI Agent Hub
Back to plugins
🤖

DSH-Multimodal

Model Inference Updated 2026.08.25

Run the following command in DeepSeek Harness:

dsh plugin install yauntyour/DSH-Multimodal

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

To install the DSH-Multimodal plugin in DeepSeek Harness, use the command `dsh plugin install yauntyour/DSH-Multimodal` with source code at https://github.com/yauntyour/DSH-Multimodal.

About this plugin

Many powerful session models, such as DeepSeek, support only text input. Directly attaching images triggers errors, while video and audio files are completely unsupported. DSH-Multimodal addresses this by intercepting file inputs during transmission, using configured vision models to convert images, videos, and audio into detailed text descriptions. It offers a flexible settings page for independent configuration of model chains per file type, featuring intelligent fallback mechanisms. It is perfect for users looking to enhance text-only models to understand visual content or process media files without altering the native session model settings. Whether analyzing screenshots, handling media files, or building complex multimodal workflows, this plugin provides the necessary support, empowering pure text models to "see" and "hear".

Use Cases

  • Enable text-only models like DeepSeek to accept images
  • Analyze screenshots, charts, or video/audio file content
  • Configure processing workflows for locally deployed vision models

Best For

  • Users relying on text-only models like DeepSeek
  • Developers needing vision capabilities for local LLMs
  • Researchers wanting to extend AI assistant functionalities