MIT-licensed DeepSeek Harness plugin (TypeScript) that passes images natively to vision-capable models and bridges a configured vision model for text-only models, with a macOS Vision / Tesseract fallback — one command via the plugin manager built into Harness.
DSH integration
Native runtime
Author-claimed
Safety audit
Unaudited
Last verified
2026-08-27
License
MIT
01What can it help you accomplish?
Make a text-only DeepSeek Harness model understand images
A configured vision model observes the original images and its output is injected as untrusted attachment context; the original DeepSeek model still produces the final answer
Users who select `deepseek-official` or another text-only model in Harness but still want to paste or drag images into the composer
Analyze several images together in one conversation
Multiple image attachments analyzed together so comparisons and combined evidence work naturally; the user's task is forwarded unchanged instead of being wrapped in a fixed report template
Users who want to compare screenshots or combine evidence across several image attachments
Keep image support when cloud vision is unavailable
Observation falls back to macOS Vision or Tesseract, and the DeepSeek model still produces the final answer
Users without access to cloud vision who still want image understanding in Harness
02How to install into DeepSeek Harness
Prerequisites
- DeepSeek Harness — the README introduces dsh-vision as a plugin for DeepSeek Harness and installs it through the plugin manager built into Harness
Installation steps
- 01
Run `npx @deepseek-ai/dsh plugin --profile web add github:oil-oil/dsh-vision`
$ npx @deepseek-ai/dsh plugin --profile web add github:oil-oil/dsh-vision
- 02
Restart Harness, then paste or drag images into the composer as usual
Verify the integration
Not specified by the author
03DSH integration and capability boundaries
Installed as a native dsh plugin through the plugin manager built into DeepSeek Harness (`npx @deepseek-ai/dsh plugin --profile web add github:oil-oil/dsh-vision`)
Native image passthrough
images for a vision-capable main model→original images sent directly to the current model without preprocessing or OCR
Vision bridge for text-only models
images when the main model is `deepseek-official` or another text-only model→a configured vision model observes the original images and its output is injected as untrusted attachment context; DeepSeek produces the final answer
the vision model's output is injected into the conversation as untrusted attachment contextLocal vision fallback
images when cloud vision is unavailable→observation falls back to macOS Vision or Tesseract, and DeepSeek still produces the final answer
Multi-image analysis with task passthrough
multiple image attachments→attachments analyzed together so comparisons and combined evidence work naturally; the task is forwarded unchanged
the plugin does not replace the main model selected in Harness
04Who is it for? When not to use it?
Good for
- Users who select `deepseek-official` or another text-only model in Harness but still want to paste or drag images into the composer
- Users who want to compare screenshots or combine evidence across several image attachments
- Users without access to cloud vision who still want image understanding in Harness
Not for
- When a text-only main model is in use, the vision model's observation output is injected into the conversation as untrusted attachment context.
05Compatibility, maintenance and safety notes
- Image understanding depends on the selected main model: vision-capable models receive images natively, while text-only models require a separate configured vision model to observe the images.
- When a text-only main model is in use, the vision model's observation output is injected into the conversation as untrusted attachment context.
- When cloud vision is unavailable, the fallback relies on local tools — macOS Vision or Tesseract.
MIT · actively maintained (last push 2026-08-18)
06Frequently asked questions
How do I install dsh-vision?
Use the plugin manager built into DeepSeek Harness: run `npx @deepseek-ai/dsh plugin --profile web add github:oil-oil/dsh-vision`, restart Harness, then paste or drag images into the composer as usual.
My main model supports images — does the plugin change anything?
No. Vision-capable models keep receiving images natively; original images are sent directly without preprocessing or OCR, and the current model produces the final answer.
Can a text-only model like `deepseek-official` understand images with this plugin?
Yes — when the main model is text-only, the plugin asks a configured vision model to observe the original images and injects its output as untrusted attachment context; the original DeepSeek model still produces the final answer.
What happens if cloud vision is unavailable?
The plugin falls back to macOS Vision or Tesseract for local image observation.
Does it replace my selected main model or wrap my task in a template?
No. The plugin does not replace the main model selected in Harness, and your task is forwarded unchanged instead of being wrapped in a fixed report template.
07Related DSH workflows
modlens
by liustack
Vision bridge for text-only models: paste an image, get structured JSON evidence (OCR, layout, semantics).
agent-vision-toolkit
by anionex
Vision toolkit and skills that give text-only LLMs eyes — multi-image understanding, image Q&A, OCR, frontend UI restoration and GUI automation, with optional agent integration.
dsh-vision-router
by ysr666
Eyes for text-only DeepSeek Harness agents: built-in free vision chain (no key) + pixel-level vision tools (Q&A, grounding, crop, pixel diff, colors, OCR, SVG trace, cutout, screenshots). One-command install, no Python, image turns work like ordinary tool-calling turns.
dsh-vision-toolkit
by anionex
为纯文本 DSH Agent 提供 10 个结构化视觉工具：意图感知图片问答、长截图 OCR、原始像素 grounding、UI 还原、像素 diff 等
08Data and sources
`dsh-vision` is a plugin for DeepSeek Harness. Vision-capable models keep receiving images natively.
Use the plugin manager built into DeepSeek Harness:
npx @deepseek-ai/dsh plugin --profile web add github:oil-oil/dsh-vision
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-08-27. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.