x-Xin23/dsh-vision-bridge

Equip the DeepSeek Harness plain text model with native vision (Windows): paste and see the image - pre-inject description, the model is visible in the first round, no need to select a model, no need to adjust tools; see_image precise inspection; custom visual backend (any OpenAI compatible model) +

Web UI Enhancements ★ 2 updated 2026-08-15 — untested
View on GitHub ↗

Install

dsh plugin --profile web add github:x-Xin23/dsh-vision-bridge

Details

More plugins in Web UI Enhancements