Plugin directory / Developer / dsh-vision-LMstudio
dsh-vision-LMstudio
Verified · install-tested on dsh TiankunDai
What it does
dsh-vision-LMstudio — DSH plugin (llm)
Works — verified, early-stage project
dsh-vision-LMstudio — DSH plugin (llm) It installs cleanly and boots without issues in our testing. It's early-stage but functional.
“Verified” means our automated CI actually ran dsh plugin add in a clean profile and it booted — nothing more. Feature descriptions and version compatibility are the author’s claims. This is not a security audit and not an endorsement of third-party code.
README
dsh-vision-LMstudio
Local LM Studio vision recognition plugin for DSH (DeepSeek Harness) Web GUI —
recognize, describe, and OCR local images or clipboard screenshots with a local
vision model. All inference happens on your own machine; images never leave it.
本插件为 DSH(DeepSeek Harness)Web GUI 提供本地 LM Studio 视觉识别能力:连接
LM Studio 的 OpenAI 兼容接口(默认 http://127.0.0.1:1234),
调用其视觉模型对本地图片与 Windows 剪贴板图片做识别、描述与 OCR,图片不出本机。
Features
- Agent tools (registered host-side, available in every session)
lmstudio_vision— recognize a local image file by absolute pathlmstudio_clipboard— recognize the current clipboard image (screenshot / copied image), no file path neededlmstudio_models— list loaded models with vision-capability labels
- Settings panel:「设置 → LM Studio 视觉识别」— configure server URL / model / default
prompt / temperature / max tokens, refresh the model list, and one-click recognize a
local image or the clipboard - Automatic model selection — without an explicit
model, picks
gemma-3/4 → other vision models → other non-embedding models in that order
(prefersgoogle/gemma-4-31b-qatfor accuracy; minicpm-class models are not recommended) - Persistent config —
~/.dsh/dsh-lmstudio-vision.json(atomic writes), shared by the
panel and the agent tools
Requirements
- DSH (DeepSeek Harness) with the web profile
- LM Studio with a vision model loaded and the local server started
(Developer → Start Server, default port 1234) - Windows for the clipboard tool (uses Windows PowerShell 5.1
Get-Clipboard -Format Image)
Install
pnpm install
pnpm build
dsh plugin --profile web add link:<path-to-this-repo>/packages/dsh-lmstudio-vision
Then restart DSH (dsh web) and refresh the page. dsh plugin appends the package
to dsh.profile.bundles automatically (the package declares dsh.bundle).
安装后重启 DSH 并刷新页面:设置页出现「LM Studio 视觉识别」面板,三个 agent 工具随宿主
常驻,任何会话都可用。
Configuration
Configurable from the settings panel or by editing ~/.dsh/dsh-lmstudio-vision.json:
{
"baseUrl": "http://127.0.0.1:1234",
"model": "",
"prompt": "请详细描述这张图片的内容。",
"temperature": 0.2,
"maxTokens": 1024
}
Repository layout
packages/dsh-lmstudio-vision/src/index.ts— host half: LM Studio engine, API routes,
agent tools, system-prompt announcementpackages/dsh-lmstudio-vision/src/client/— browser half: the settings panelpackages/dsh-lmstudio-vision/cordis.patch.yml— bundle patch plugin rowshared/tsdown.client.ts— build preset (follows the dsh-web-uiclientBundleconvention)
License
Install
Install the catalog once, then DeepSeek Harness can find and install any plugin from this site automatically:
dsh plugin add dshbase-catalog Then say "install dsh-vision-LMstudio for me" — your agent finds it in the directory and installs it. Docs: dshbase-catalog · verified packs.
This plugin is GitHub source (not published to npm) — install it straight from the repo:
Web profile:
dsh plugin --profile web add github:TiankunDai/dsh-vision-LMstudio Headless (CLI) profile:
dsh plugin --profile headless add github:TiankunDai/dsh-vision-LMstudio Test report
Verified: L1 install + L2 load + L3 runtime from GitHub source on dsh 0.1.0-rc.6.
When to use it
Extend the agent's coding surface — give it a new tool, workflow, or integration so it handles a dev task it couldn't before.
Who it's for
Developers who want dsh to behave like a teammate on real codebases — editing, running, and verifying changes rather than just answering.
For developers — extending it
The tool/command surface is the seam: expose more of the SDK, add smarter context wiring, or tighten the loop between code changes and verification.