tiefeiyu/dsh-see-image
A see_image vision tool plugin for DeepSeek Harness — describe images through any OpenAI-compatible vision mod
Registers a see_image tool that routes image files to any OpenAI-compatible vision model and returns a text description, so text-only models can 'see' pictures without switching models or orchestrating subagents: reads the image through DSH's file service (sandbox/observation policy applies), base64-encodes it, POSTs to {baseURL}/chat/completions. Default backend GitHub Copilot gpt-4.1 (Free-plan eligible, built-in token auto-refresh). Requests are made by the plugin directly and don't carry DSH's tool list, so OpenAI-style gateway limits (128 tools) don't apply.
Install
git clone https://github.com/tiefeiyu/dsh-see-image && cp -r dsh-see-image "$DSH_HOME/profiles/web/plugins/dsh-see-image"Install = copy the folder into your DSH profile + one cordis.patch.yml entry (append the insert row for the see-image plugin), then restart dsh web if hot reload doesn't pick it up. Zero npm installs — reuses the host's @deepseek-ai/dsh-tools / schemastery. Default backend: GitHub Copilot gpt-4.1 (works on the Free plan) with built-in token auto-refresh — no key management. Any OpenAI-compatible VLM works (Zhipu/Bailian/Doubao/SiliconFlow/local Ollama/OpenRouter) by changing three config keys. Requires DSH 0.1.0-rc.6.
Compatibility
DSH 0.1.0-rc.6; any OpenAI-compatible vision endpoint. Zero npm dependencies; hot-reload friendly.
Details
- Repo: tiefeiyu/dsh-see-image
- Category: Coding & Development
- Stars: 2
- Version: GitHub source (folder-copy install; no npm package)
- Last push: 2026-08-14
- First seen: 2026-08-14
Recent updates
see_image tool to any OpenAI-compatible VLM; GitHub Copilot default with auto-refresh; zero npm deps; no tool-count limits.
FAQ
- How do I install it?
- Copy the folder into the DSH profile and add one cordis.patch.yml entry — no npm installs; restart dsh web if hot reload doesn't pick it up.
- Which vision backends work?
- Any OpenAI-compatible VLM: Zhipu, Bailian, Doubao, SiliconFlow, local Ollama, OpenRouter, or GitHub Copilot gpt-4.1 by default.
- Does it need an API key?
- The default GitHub Copilot backend includes token auto-refresh (Free plan works); other backends need their own key in config.
Alternatives
niyongsheng/free-vision-skill · FuzzySoul/dsh-free-vision · 54xkeee/dsh-youreyes