This article is derived from the plugin's already-indexed fields (wiki / faq / compatibility_json / readme), not freshly generated by AI. Source field is noted at the end of each section.

Quick start

dsh-minimal-vision

— source: plugin_wiki.wiki_content

Install & verify

dsh plugin --profile web add dsh-tool-vision

Run the command above in your DSH Web Profile. Then enable the plugin in the plugin list.

— source: plugins.install

FAQ

Must it be configured immediately after installation to use?

Yes, it is required. You need to fill in the API address, vision model ID, and API Key in Settings → Vision for it to take effect; the test connection only sends a minimal text request and does not actually upload images (src/connection-test.js:31-35 / src/client.js:287-306).

What image and document formats are supported?

Images support PNG/JPEG/WebP/GIF, documents support PDF/DOC/DOCX/PPT/PPTX. Documents are not directly sent to the vision interface; they need to be first exported as images using Bash and then passed to dsh-vision (src/image.js:8-14 / src/file-upload-api.js:8 / src/injector.js:15).

How many images can be uploaded at once?

A single dsh-vision call accepts a maximum of 4 images, each not exceeding 20 MB (src/analyze.js:6,60 / src/upload-api.js:8); documents have a single file limit of 100 MB (src/file-upload-api.js:7).

Where are attachments stored after upload? Are they uploaded to the cloud?

Both images and documents are stored in the local cache directory $DSH_HOME/cache/dsh-tool-vision/. Images are only sent to the configured vision service when the Agent explicitly calls dsh-vision analyze, and documents always remain local (README.md:43-46).

Can it run on native Windows PowerShell?

Not supported in the current version. Tested successfully on Ubuntu and WSL2; Windows users need to install and run DSH and this plugin within WSL2 first, native PowerShell support is not yet available (README.md:49).

What is the difference from the official vision capabilities?

This plugin does not register vision models as new chat models, nor does it rewrite the system prompt in minimalist mode; vision descriptions and attachment paths are included as folded context alongside user messages, the two original tools persistent-bash and str_replace_editor remain unchanged (README.md:21-31 / presets/vision-assist/agent.cordis.yml:1-2).

How to check if the environment is normal?

The plugin comes with a dsh-vision-doctor script that checks whether the profile installation package, launcher, Node runtime, CLI paths, and preset files exist and are executable (scripts/doctor.mjs:12-58).

— source: plugin_wiki.faq_json

Compatibility

• DSH: >=0.1.0-rc.6
• Node: >=20

— source: plugin_wiki.compatibility_json

Pitfalls

Review the upstream repo before installing. This guide is auto-derived from indexed fields and may lag the latest release. If anything contradicts the official docs, treat the upstream source as authoritative.

— source: general rule