Any vision model automatically converts images to plain text LLM
poiuyjie/dsh-vision-opencode · skills
DSH plugin: Automatically convert images to plain text LLM (DeepSeek, etc.) by any vision model.
★ 14GitHub stars
0Forks
2026-08-15Last updated
JavaScriptLanguage
MITLicense
Key features
- I just want to turn off "automatic conversion of images" (keep the tools and selectors): vision-opencode.autoConvert: false in settings.yaml, restart.
- "Image recognition model" drop-down selector on the right side of the input box (automatically lists models from all suppliers that declare image input)
- Runtime skill vision-image-analysis: guides the main model to call vision_read_image for OCR/chart/scene understanding on demand
- When the DSH skill service is not enabled, it will automatically degrade to tool and system prompt words.
Requirements
- The front-end selector is a handwritten DSH client bundle (window.__ModuleLoader__ format), and the client interface may change between rc versions
- When the selector does not appear, open the browser console and post an issue.
Install command
dsh plugin --profile web add github:poiuyjie/dsh-vision-opencode