lijian-ui/dsh-vision-toggle

A model-vision toggle plugin for the DSH desktop: it lists every model declared in the llm-pi-ai provider routes grouped by provider and gives each one a Support image switch that rewrites the model's input modalities between ['text','image'] and ['text'] - no settings.yaml hand-editing and no restart. This fills a gap where custom OpenAI-compatible providers (LM Studio / vLLM / Ollama) whose models support vision otherwise fail with an unsupported-image error.

dsh-vision-toggle gives DeepSeek Harness desktop users control over whether each configured model accepts image input. It reads every model declared in the llm-pi-ai provider routes grouped by provider and gives each model a Support image switch that rewrites its input modalities between ['text','image'] and ['text'] in the official config, so llm-pi-ai re-registers and the change takes effect instantly with no restart. It addresses the gap where DSH's settings page offers no way to declare the input modalities of custom providers - so an OpenAI-compatible provider added via LM Studio, vLLM or Ollama whose model actually supports vision otherwise fails with a this-model-does-not-support-images error, today only fixable by hand-editing settings.yaml.

Vision & Multimodal ★ 1
View on GitHub ↗

Install

dsh plugin add @lijian-ui/dsh-vision-toggle

Copied verbatim from the README's install line. It writes into the official llm-pi-ai settings namespace, which re-registers on change, so the toggle takes effect immediately with no restart.

Compatibility

Targets the DSH desktop's llm-pi-ai provider routes; the UI follows the dsh desktop language setting (Chinese / English) and flips optimistically, rolling back if the host write fails. English README with a 中文 translation.

Details

Alternatives

54xkeee/dsh-vision · 54xkeee/dsh-youreyes · Cheng-cheng9669/dsh-deepseek-vision

More plugins in Vision & Multimodal

Browse more in Vision & Multimodal

Guides for Vision & Multimodal plugins