dsh-multimodal-bridge

About

multimodal-bridge is a multimodal capability bridge: it brings Qwen's visual understanding (Qwen-VL) and image generation (Qwen-Image) to text-only models (like DeepSeek) that lack native multimodal capabilities. It features two forms with a unified backend: MCP Server (qwenvision / qwengenerate tools) for direct mounting by any MCP-compliant host (Claude Code, Kimi Code, etc.); and DSH plugin (npm package dsh-multimodal-bridge, DeepSeek Harness bundle) installable with one line via dsh plugin add, featuring automatic model fallback, responsive sizing, and image result cards.

Description machine translated from the original.

Topics