Image generation and image reading for DSH: an Images page in the sidebar, labelled 图像 in Chinese (one row per model, 10 ratios, 1K–4K sizes, a contact sheet with detail cards, zoom, reuse and delete), three preset providers plus custom endpoints — one needs no key at all — and a vision-review skill. A pasted key is written to ~/.dsh/secrets/media-tools.env (mode 600) and only sent to the provider you picked. The page follows the harness language (Chinese or English). Requires Python 3.9+ (generation runs through scripts/generate.py); the page needs desktop DSH 0.2.1+.
Install
# from npm (prebuilt)
dsh plugin --profile web add dsh-media-skills
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:MJorgin/dsh-media-skills
Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).
README
dsh-media-skills
An image workspace for DeepSeek Harness. Generate images, revisit your work, and ask your agent to review a screenshot.
简体中文 · Quick start · Image review · Privacy · Troubleshooting
One plugin, three ways to work
| What you want | Where to start | What it needs |
|---|---|---|
| Generate an image | Images → Generate in the sidebar | Desktop 0.2.1+; a provider. Flux via Pollinations is configured as keyless. |
| Find or reuse a result | Images → Works | The page stores up to 60 recent entries in local storage |
| Describe or inspect an image | Ask for vision-review, or use a compatible image model in chat |
A configured vision provider for the skill, or an image-capable chat model for native attachments |
The sidebar page and media-tools use the same generation script. vision-review is a separate image review skill.
Quick start
In DeepSeek Harness Desktop, open Plugins → Add plugin, enter
dsh-media-skills, and install it. Or use the desktop profile:dsh plugin --profile desktop add dsh-media-skillsRestart DSH. On Desktop 0.2.1+, open Images in the sidebar.
Enter a prompt, select Flux (Pollinations), and generate. This path does not require a key; availability, image size and rate limits depend on the provider.
Open Works to revisit a result. To use another provider, configure it under Image key.
The two skills are registered with the plugin. Image review through vision-review requires a configured vision provider; a keyless image-generation option does not automatically supply image understanding.
First run: this repository does not control third-party availability or pricing. If Flux is unavailable, check the provider status or configure another supported provider.
The Images page
Interface illustration. The sample picture is a placeholder, not a generated result; the actual page and available models can differ.
Choose a prompt, model, aspect ratio and canvas size. The result card shows the image and its settings. The page also has a Works tab for revisiting images, copying prompts, reusing settings and deleting files.
Gallery illustration with placeholder images.
Generated examples from the repository. Prompts and provider details: examples.
Image review
Ask the agent to use vision-review to describe a local image, inspect a UI screenshot, or produce structured observations. The script supports configured vision providers and a fallback chain; see vision-review/SKILL.md.
Native chat attachments are a separate DSH feature: the currently selected model must accept image input. If a text-only model rejects an attachment, switch to an image-capable model or use a local file path with a configured vision-review provider. Installing this plugin does not turn a text-only model into a vision model.
Privacy & providers
- The plugin does not bundle API keys. Panel-managed keys are stored locally in
~/.dsh/secrets/media-tools.env; the code requests file mode600where supported. - The prompt and relevant request parameters go to the provider you select. Image review sends the image to the configured vision provider. Custom endpoints are chosen by you.
- Generated files are saved in the local workspace. The page keeps up to 60 recent entries in local storage. Deleting an entry attempts to delete its file; if file deletion fails, the entry is still removed and the file can remain. Provider data-handling rules still apply.
- Free tiers, limits, watermarks, models and terms can change. Consult the provider before sending sensitive content or depending on a free service.
Details: generation skill · vision setup · ModLens coexistence.
Troubleshooting
- No Images page: the page requires Desktop 0.2.1+ and a restart after installation. The skills can be used separately on supported older hosts; see compatibility notes.
- Generation fails: check the selected provider, network, account balance and rate limits. The panel reports common HTTP errors where available.
- An image attachment is rejected: verify the current chat model supports image input, or use the file-path review workflow with a configured vision provider.
- Need a diagnostic for image review: run
python3 skills/vision-review/scripts/vision.py --doctorfrom the repository. Some checks may contact configured providers.
Languages
English · 简体中文 · 繁體中文 · 日本語 · 한국어 · Español · Deutsch · Português · Русский
English and Simplified Chinese describe the current Images page. Other translations are retained for reference and may describe earlier versions.
Development
Requires Node.js 22.19+ and Python 3.9+; image review uses Pillow. Run the repository checks with:
npm test
The package contains client.js, index.js, the DSH bundle manifest and the two skill directories. Examples · Issues · MIT license
Links
More in this category
liustack/modlens★ 4961
Vision bridge for text-only models: paste an image, get structured JSON evidence (OCR, layout, semantics).
ysr666/dsh-vision-router★ 1268
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.
Anionex/dsh-vision-toolkit★ 945
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.
SAOS-3624/dsh-see-now★ 630
Lets a text-only agent watch a live screen or camera feed: every frame is scored by how much the picture changed, only the ones that matter are packed into one numbered sheet with an audio envelope, so the model looks at what actually changed instead of a screenshot every few seconds.
SAOS-3624/dsh-video-storyboard★ 536
Turns a video (a local file, or a Bilibili/yt-dlp link) into one addressable storyboard grid — numbered cells with exact timestamps, per-cell transcript and an audio envelope on the same timeline — and video_zoom re-reads any window at higher resolution, so a text-only model can look at the video instead of reading a summary.
dickpy/dsh-imagegen★ 188
AI image generation for the DSH Web GUI: text-to-image and image-to-image through a configurable OpenAI-compatible endpoint (gpt-image-2 / gpt-image-1 / dall-e-3), with an api_url/api_key settings card and a sidebar split-pane generation studio.
Community comments
Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.