dsh-vision-autoswitch
Run the following command in DeepSeek Harness:
dsh plugin install CultOfLuna/dsh-vision-autoswitch
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install CultOfLuna/dsh-vision-autoswitch in DeepSeek Harness to install; the source repository is at https://github.com/CultOfLuna/dsh-vision-autoswitch .
About this plugin
Pasting a screenshot or an image from tool output into a DeepSeek chat is perfectly natural, but Pro and Flash models have no vision capability, so images are simply ignored. dsh-vision-autoswitch solves exactly this: when a turn contains an image it routes to deepseek-v4-flash-vision-exp, and the moment the turn ends it falls back to the model you originally picked. Zero clicks required.
Three things make it work. Detection and switching: at the start of every turn the plugin sniffs for images in the request; if found, routing is rewritten to the vision model so the image is actually read. Immediate fallback: it listens for turn/end and restores your previous model selection right away, so long conversations never get stuck on the vision model. WYSIWYG: the model bar always mirrors the model actually in use, so you are never confused about which model is answering.
Built for anyone who frequently deals with screenshots, image attachments, or images inside tool output while staying on DeepSeek Pro or Flash. Pure-text turns are still handled by your high-capability model of choice, and the fallback target always follows your current manual selection rather than being hardcoded to a fixed model.
Use Cases
- Pasting a screenshot into a chat that requires visual understanding
- Tool output contains images that need to be properly processed
- Occasionally needing image comprehension without manually switching models back and forth
Best For
- DeepSeek everyday users who frequently handle screenshots, attachments, or tool image output
- Developers using toolchains where output often contains images
- Users who want zero-click vision while keeping a high-capability model for all text-only turns
Related Plugins
A unified suite combining hot runtime injection, task-aware thinking-mode routing, and a graded session protocol with red-team gates to sustain model diligence across long-horizon inference.
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.