DSH Marketplace

DSH-vison

hisence999/DSH-vison

DSH-vison is a DeepSeek Harness plugin that adds image understanding for text-only models by converting images into descriptions from a configured vision-capable model.

Install

Add DSH-vison to DeepSeek Harness

Install into

via GitHub · GitHub source

Installing from GitHub runs the project's build script, which pnpm blocks until you allowlist it — run the command once and pnpm prints the exact key to add under `allowBuilds` in ~/.dsh/profiles/web/pnpm-workspace.yaml.

What happened when we ran it

Installed cleanly when we ran it

Every command here is run in a throwaway container against a clean profile, and the result is whatever the harness recorded — not a guess from the source. Last run 10d agoon DeepSeek Harness 0.2.2-rc.1.

Show it in your README

install verified — dshmarketplace

For maintainers: the badge serves this listing's latest sandbox verdict, so a re-run updates it on its own — and it links readers to the full result here.

Due diligence

Before you install DSH-vison

  • Source of record: hisence999/DSH-vison — added by hand from a community post; not in the curated registry.
  • Licensed under MIT.

Security

What installing DSH-vison runs

Build scripts

No build script was blocked in our run.

Source: our sandbox install on DeepSeek Harness 0.2.2-rc.1, 10d ago.

Install hooks

package.json declares no preinstall, install or postinstall script.

Source: package.json on GitHub, read at our last sync. prepare is not recorded; it runs only on a GitHub install, and shows up under build scripts when pnpm blocks it.

README mentions

shell commands (bash, exec, spawn …)

Source: a keyword match on the README — a hint about what the plugin touches, not an analysis of its code.

Not an audit. These are the record of one install and two text scans; nobody has read this plugin's code on your behalf. Plugins run with your agent's permissions.

The AI take

What it is — DSH-vison provides image understanding for text-only models by converting images into descriptions from a configured vision model.

Who it is for — When you use a text-only model to process user messages containing images, you need image understanding capability. If you use a multimodal model, you can skip installing this plugin.

Watch out — Sandbox test passed: it was installed in a fresh profile and registered by harness. No obvious pitfalls found.

The verdict — I would install it because the sandbox test passed and it enables global image handling for text models.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.

What DSH-vison does

Image tiles split between preserved records and a description path before converging in a model chamber.

DSH-vison is a DeepSeek Harness plugin that adds image understanding for text-only models by converting images into descriptions from a configured vision-capable model. It wraps llm.resolveModelInfo so image admission and tool checks can be allowed for text-only models, then listens to agent/pre-step to process image-bearing user messages. The image remains in session history for UI display, while a session-surface replacement and deriveMessages wrapper present the generated description to the model. tools/post-execute applies the same conversion to read_image results, and system-prompt/assemble identifies the model active for the current step.

The plugin is intended for installations where an agent may receive images but the selected model is not multimodal. Multimodal models are detected and passed through without interception. Configuration is stored in the dsh-image-vision settings namespace; the profile patch uses enabled: true and patchAdmission: true as fallback settings.

DSH-vison requires at least one configured image-capable model and does not cache descriptions, so images are sent for recognition again in later turns. Its terminal surface includes the Windows PowerShell and Unix shell installation and removal scripts, which copy files and modify $DSH_HOME profiles, cordis.patch.yml, and the DSH dsh-host-apiproxy whitelist. It is a poor fit where a multimodal model already handles all images directly, or where modifying those host files is not acceptable.

DSH-vison documentation

Configuration

The plugin stores settings in the dsh-image-vision namespace. The settings page exposes an enable switch, the recognition model, and the prompt. Its profile patch can provide these fallback values:

Key Value
enabled true
patchAdmission true

patchAdmission controls the reversible wrapper around the shared llm service. The plugin also requires dsh-image-vision to be added to the dsh-host-apiproxy setting namespace whitelist; otherwise the settings page cannot find its namespace.

Files and host integration

The plugin package consists of index.js, client.js, and package.json. For a profile installation, these files are placed under $DSH_HOME/profiles/<profile>/node_modules/dsh-image-vision/; the README gives web as the usual Web profile and shows the corresponding Windows path under C:\Users\<user>\.dsh\profiles\web\.... The profile’s $DSH_HOME/profiles/<profile>/cordis.patch.yml contains an image-vision entry naming dsh-image-vision.

The setup scripts patch the WEB_SETTINGS_NAMESPACES list in dsh-host-apiproxy. Reinstalling or upgrading DSH can overwrite that host file, requiring the patch to be applied again.

Commands and platforms

The repository supplies install.ps1 and uninstall.ps1 for Windows PowerShell, plus install.sh and uninstall.sh for Linux and macOS. Uninstallation removes the package, the image-vision patch entry, and the API proxy whitelist change. Configuration remains unless the removal script receives -PurgeConfig on Windows or --purge-config on Unix. Restart DSH after installation or removal.

Known limits

At least one configured model that supports images is required. Descriptions are not cached; the same image is recognised once per turn, but later sends invoke the vision model again. Failed recognition retries other configured image-capable models at 0.6-second intervals. If all attempts fail, placeholder text is used instead of sending the image to a text-only model.

Written from the project's own documentation and kept in sync with it. Where the two disagree, the source is authoritative — read the README on GitHub

Same category

Alternatives to DSH-vison

10k

A `describe_image` vision tool for text-only models: images (local path, URL, attachment) go to a configurable OpenAI-compatible vision endpoint and only the returned text enters the session.

No one-line install — this plugin lives inside a larger repository and publishes no npm package.

GitHub sourceTypeScript2d ago

AI review

dsh-tool-describe-image

What it is — dsh-tool-describe-image is an image description tool for text-only models. It sends local paths, URLs or attachments to a configurable OpenAI-compatible vision endpoint and only the returned text enters the session.

Who it is for — When using text-only models in DeepSeek Harness on tasks involving image inputs, this plugin is suitable for installation. Users who rely on built-in vision capabilities in their models can choose not to install it, as they have native image support.

Watch out — Sandbox testing in a new profile showed successful installation and registration with no obvious pitfalls found. No issues were detected in static checks.

The verdict — I would install it because it allows text-only models to handle image inputs in the session.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
1Details

esp32-iphone-agent

onehumanbeing

4.4k

Let an AI model use a real iPhone with a $7 ESP32: AirPlay eyes, USB HID hands, any OpenAI-compatible vision model. No Mac, no developer account, no jailbreak.

Installed cleanly when we ran it

GitHub sourceSwift11d ago

AI review

esp32-iphone-agent

What it is — Connects an OpenAI-compatible vision model to a physical iPhone: ESP32-S3 AirPlay for the screen, USB HID for taps, swipes, and typing.

Who it is for — You are inside DeepSeek Harness with an OpenAI-compatible vision model and need it to finish a live iPhone flow such as setting a repeating alarm, without a Mac, a developer account, or a jailbreak. Skip it when the phone is below iOS 18 or you have no flashed ESP32-S3 — the receiver and the HID device are simply not there.

Watch out — The phone joins the board's hotspot for AirPlay while model calls go out over cellular; AssistiveTouch stays on, and you supply the vision API key or token. The pointer is relative, so the app wiggles it to re-find and correct, and the plugin runs shell commands. A clean profile accepted the install and registered it.

The verdict — A green harness install is not a working phone; I would wait until TestFlight already shows a stable mirror and pointer before adding this plugin to a profile.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
2Source

modlens

liustack

4.4k

Vision bridge for text-only models: paste an image, get structured JSON evidence (OCR, layout, semantics).

Installed cleanly when we ran it

npm packageTypeScript1mo ago

AI review

modlens

What it is — A DSH plugin that turns a pasted image into structured JSON evidence (OCR, reading-order layout, entities and relations) for text-only models.

Who it is for — You work in DSH on DeepSeek-V4-Flash, DeepSeek-V4-Pro, GLM-5.3, or MiMo Pro and need to inspect screenshots or scans those models cannot see. If your default route is already GLM-5.3-Flash or anything the host marks as native vision, skip it; those models are not wrapped and keep native paste.

Watch out — A clean-profile install registered with the harness. Reads reuse a Claude Code, Codex, OpenCode, or Pi login, or Antigravity CLI / Gemini / OpenAI-compatible keys; a free Gemini key is quoted at 5–10 seconds per read, with comma-separated keys rotating on auth, rate-limit, or quota failures. It executes shell commands. Capture hotkey is a separate plugin, dsh-screenshot.

The verdict — I'd keep it for the auto-wrapped (modlens vision) entries on every eligible provider route, including extras like opencode-go or zai, not because it ships any local vision runtime.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
1Details

为纯文本模型"看图“设计更好的视觉工具箱和技能,支持多图理解,图片问答,前端UI还原、GUI 自动化等,并可选无缝接入多个主流agent,直接识别粘贴图片| A vision toolkit and skill designed for text-only llms — image Q&A, long-screenshot OCR, frontend UI restoration, and GUI automation, with optional seamless integration for Codex, Claude Code, Pi, Oh My Pi, and OpenCode

Installed cleanly when we ran it

GitHub sourcePython1mo ago

AI review

agent-vision-toolkit

What it is — CLI tools and a skill that let a text-only agent do multi-image Q&A, long-screenshot OCR, frontend restoration, and GUI automation.

Who it is for — You run DeepSeek in DeepSeek Harness, rebuilding UIs from mockups or driving a GUI, and the model never sees the pixels. If the agent you actually work in is already multimodal and can read a pasted screenshot, this CLI layer does not enter that loop.

Watch out — It installed on a fresh profile and the harness registered it. You pull it from GitHub source and the agent must be able to run shell CLIs. Seamless paste-in images need an extra local proxy; the ten structured DSH tools live in the dsh-vision-toolkit submodule and are absent until you init it.

The verdict — I would add it so vision sits in the harness and DeepSeek can stay the model I work with.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
1Source

Free vision for text-only agents: built-in keyless vision chain plus pixel tools (Q&A, grounding, crop, pixel diff, colors, OCR, SVG trace, cutout, screenshots); paste an image to use it.

Installed cleanly when we ran it

npm packageJavaScript24d ago

AI review

dsh-vision-router

What it is — Gives text-only DeepSeek Harness agents a vision chain and pixel-level tools; enable Vision beside the composer, then attach an image.

Who it is for — You run a text-only model in DeepSeek Harness on error screenshots, UI coordinates, OCR, or mockup diffs, and the chat model cannot take pixels. Skip it if images must stay on the machine and you will not wire a local vision backend.

Watch out — In a clean profile it installed and harness registered it. A fresh install ships an anonymous OVHcloud fallback, so images or crops, the vision prompt, and request metadata leave the machine by default — no key is not offline. It also runs shell commands.

The verdict — I'd add it to the profile I already live in: seeing as a repeatable tool turn, not another chat window that accepts pictures.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
1Details

Vision for text-only models: paste an image and the model switches to a Vision Toolkit variant for image Q&A, multi-image comparison, long-screenshot OCR, screenshot-to-UI reproduction, element grounding, and pixel diff. No API key by default — images are processed by the author-hosted free service, 100 per machine per day; configurable to your own provider.

DeepSeek Harness 0.2.2-rc.1 refuses the latest release

npm packageTypeScript23d ago

AI review

dsh-vision-toolkit

What it is — A DeepSeek Harness plugin that, on image paste, switches a text-only model to a Vision Toolkit variant for Q&A, multi-image comparison, long-screenshot OCR, screenshot-to-UI, element grounding, and pixel diff.

Who it is for — You are on a text-only model in DSH Web, cloning a UI or tracing a bug from a screenshot dropped into the session. Skip it if you already stay on vision-capable models and will not drop back to text-only.

Watch out — DeepSeek Harness 0.2.2-rc.1 refuses every published release before install: none declare a DSH range that includes 0.2.2-rc.1. Images default to the author's free host at 100 calls per machine per day, or you point it at your own provider.

The verdict — I'll keep it on the shortlist until the range is widened; I will not pin latest onto 0.2.2-rc.1.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
Details