dsh-highres-vision
Run the following command in DeepSeek Harness:
dsh plugin install azwosile/dsh-highres-vision
Paste the following prompt into your AI chat to install this plugin:
In DeepSeek Harness, install the plugin from GitHub using the command 'dsh plugin install azwosile/dsh-highres-vision', with the full source code at https://github.com/azwosile/dsh-highres-vision.
About this plugin
When working with high-resolution images, DeepSeek's vision model (deepseek-v4-flash-vision-exp) often faces size and dimension limits, leading to loss of details or inaccurate recognition. The dsh-highres-vision plugin is designed to address this by relaxing constraints like 32MiB per image and 8192px per side, enabling smoother handling of large visuals. Its core feature is high-resolution tiling: it automatically detects images in the session, and if they exceed 800×800, it intelligently splits them into overlapping blocks. These blocks, along with a full-image thumbnail, are injected into the model, preserving edge details and enhancing OCR and fine-grained recognition. The plugin runs out-of-the-box with Node.js (no Python needed) and is ideal for developers requiring precise image analysis—such as in document scanning, medical imaging, or remote sensing—where detail accuracy matters. However, note that tiling increases token usage, so it's recommended only when detailed inspection is essential to balance cost and performance.
Use Cases
- Processing large images beyond 800×800 pixels for detailed recognition.
- Scenarios requiring OCR or local detail analysis.
- Breaking image size limits when using the DeepSeek vision model.
Best For
- Developers using the DeepSeek vision model.
- Data scientists requiring high-precision image analysis.
- Professionals handling documents, medical, or remote sensing images.
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.