dsh-image-describe
Run the following command in DeepSeek Harness:
dsh plugin install sd1g1/dsh-image-describe
Paste the following prompt into your AI chat to install this plugin:
To install this plugin in DeepSeek Harness (source: https://github.com/sd1g1/dsh-image-describe), run dsh plugin install sd1g1/dsh-image-describe in your terminal.
About this plugin
Many efficient text-only models lack native visual understanding. When a user sends a screenshot, a chart, or a photo, the main model only sees an opaque attachment marker and cannot interpret the image at all. dsh-image-describe is a host plugin designed to close that gap: it equips a text-only main model with a describe_image tool that delegates to an external vision model, which reads the image and returns a structured text description so the conversation can proceed seamlessly.
Under the hood, the plugin accepts two input modes: attachment (reusing an existing attachment ID from the current or a prior session, even across compressed history) and path (pointing to a local image file in png, jpeg, webp, or gif format). Before dispatching to the vision model, images are auto-resized to a 2000x2000 cap and re-encoded to the smallest viable PNG or JPEG tier, balancing quality against bandwidth. Configuration changes take effect immediately without restarting the host process.
Who benefits most? If your main model is text-only by design yet you regularly need to handle user-submitted screenshots, diagrams, or scanned documents, this plugin plugs the visual gap with zero intrusion to your existing pipeline, turning an unrecognizable image into actionable context for reasoning and response.
Use Cases
- A text-only model receives a user screenshot, calls describe_image, and responds with the description
- A local PNG chart is pasted and the model reads it via path mode to understand its content
- After session history is compressed, the model re-describes the image using the same attachmentId
- A text-only conversation uses image descriptions for data analysis or content summarization
Best For
- Developers using text-only main models who need to handle user-submitted images
- Harness host users looking to add visual capability at low cost to their existing setup
- Conversation app builders needing image-based reasoning without native vision support
Related Plugins
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.
An armor-breaking plugin for DeepSeek: rewrites instruction priority, expands refusal-blocked phrases, persists across sessions, and shows a green active indicator.