Multimodal bridge for any LLM: qwen_vision (Qwen-VL) + qwen_generate (Qwen-Image) — an MCP server for Kimi Code & Claude Code, and a DeepSeek Harness plugin (dsh-multimodal-bridge). 给任何 LLM 补上 Qwen 视觉与出图能力:MCP server(Kimi Code / Claude Code)+ DeepSeek Harness 插件双形态。
DeepSeek Harness plugin: local llama.cpp (llama-server) as a first-class model provider — process/router management, catalog sync with mmproj vision pairing, auto-start, stop-then-load switching, and a sidebar terminal monitor panel.
OpenAI-compatible HTTP surface for DeepSeek Harness (dsh): POST /v1/chat/completions, POST /v1/responses, GET /v1/models over the harness webServer + llm runtime
Open-source MCP vision server that gives text-only LLMs and AI agents image understanding, OCR, visual analysis, UI inspection, and multimodal capabilities.
DeepSeek Harness plugin: role-based model routing — planner (root agent) on deepseek-v4-pro, delegated executor subagents on deepseek-v4-flash; ships a prompt section and a pro-flash-routing skill.
DeepSeek Harness subagent provider: delegate reasoning to Tabbit-served models through a local gateway. Ships its own minimal preset, gateway management, and delegation guidance.