dsh-tool-see-image
Run the following command in DeepSeek Harness:
dsh plugin install gugu123a/dsh-tool-see-image
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install gugu123a/dsh-tool-see-image in DeepSeek Harness to install this plugin; full source URL: https://github.com/gugu123a/dsh-tool-see-image
About this plugin
The default text-only model in DSH cannot see images. Send it a screenshot and it can only reply with a dry "I cannot view images." dsh-tool-see-image fixes exactly this: it gives the text model a pair of eyes by routing each image to a configurable vision model (Zhipu GLM-4V-Flash by default, free), then relaying the textual description back into your DSH session. No local GPU, no model swap, no extra infrastructure needed.
The core see_image tool takes an image path and a question, and returns a structured text description. The vision backend is swappable to any OpenAI-compatible /chat/completions endpoint such as SiliconFlow or Qwen2.5-VL. Files are read through ctx.fs so DSH sandbox and observation policies are always respected. A bonus patch script lets you paste an image straight into the chat box; it is auto-converted to a description via the vision model, with local caching and an 8-second graceful timeout.
Ideal for developers who work with a text-only model in DSH but regularly need to look at screenshots or document photos.
Use Cases
- Let a text-only model read UI screenshots and answer questions about the interface
- Convert document photos into structured text descriptions for downstream reasoning
- Paste images directly into the chat box and auto-convert them to text via a vision model
Best For
- Developers using a text-only model in DSH who occasionally need to inspect screenshots or photos
- Lightweight workflow users who want image understanding without deploying a local GPU
- Users who prefer to extend multimodal capabilities via a plugin without changing their existing DSH model setup
Related Plugins
A unified suite combining hot runtime injection, task-aware thinking-mode routing, and a graded session protocol with red-team gates to sustain model diligence across long-horizon inference.
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.