시각 도구 · DeepSeek Harness 플러그인 — 57 페이지

DSH vision plugin: image recognition (Codex + Zhipu fallback) and generation (GPT Image + CogView fallback), text-only-model safe

DeepSeek Harness 识图插件：为 Agent 注册 image_analyze 工具，把本地图片或 URL 发送给 OpenAI 兼容视觉模型（默认 OpenRouter google/gemini-3.7-flash）并返回文字分析

Vision Bridge plugin for the DeepSeek Harness: routes attached images to a vision-capable model (default xiaomi/mimo-v2-omni) and feeds the text description back to text-only routed models such as DeepSeek.

dsh-vision — a DeepSeek Harness plugin (tool template).

Give DeepSeek Harness eyes: a vision_analyze tool that sends an image (local path or URL) to an OpenAI-compatible vision model and returns a text description.

DSH Computer Use (vision + desktop control, auto-detects Windows native or WSL): see the Windows screen through the agent vision channel, drive mouse/keyboard with a Codex-style blue overlay (Esc to abort), plus a composer status pill.

Local reverse proxy plugin for DeepSeek Harness that caps images per request to 1, for self-hosted vLLM backends that reject prompts with more than one image

给 DeepSeek Harness 加视觉能力：可视化设置页选视觉厂商（Kimi/智谱/通义/OpenAI/Gemini/豆包/MiniMax/阶跃星辰）并粘贴 API Key，聊天里即可拖图识别——图片走视觉模型、文字走 DeepSeek 推理。

Vision + Android UI automation plugin for DeepSeek Harness: gpt-5.6-luna screenshot understanding and real-tap mobile app testing via adb

DSH 插件：可接入识图模型 API（本地大模型识图工具 + 设置界面）。配置 OpenAI 兼容的识图接口（LM Studio / vLLM / Ollama 等），接口留空即不启用识图模型。

DeepSeek Harness 视觉助手插件：给没有视觉能力的模型配一个可随时切换的多模态识别模型。输入框图片自动落盘并改写为文本提示，主模型调用 vision_recognize 工具即可完成看图；识别模型在 settings 的 vision-assist 命名空间热更新切换。

Vision 台架：会话区调试/上位机点表（按段添加地址），共享任务与时间线，并安装「台架模式」Agent 预设。

DSH 视觉桥接插件：复用 DSH Provider 或 OpenAI 兼容直连，支持可视化配置与图片粘贴

Vision Bridge — a persistent DeepSeek Harness plugin that gives the text-only DeepSeek model image understanding by bridging to user-configured vision model APIs (OpenAI-compatible / Google Gemini).

Give text-only DeepSeek Harness agents vision — multi-backend (OpenAI Vision API + CDP bridge), cross-platform, cached, configurable.

DeepSeek Harness 多模态视觉桥：贴图自动转文字描述（llm/stream 代理）+ view_image/ocr_image 主动视觉工具 + 原生多模态路由自动跳过（rc.7 适配），让 text-only 的 DeepSeek 模型看见图片

Minimal DSH vision bridge: capability-aware paste routing — image-capable models paste natively, text-only models get cached pastes as path markers read via vision_glance, so text-model sessions stay switchable

明眸 VisionBridge - 自研视觉桥：瞎子模型收图时自动调用视觉模型识别，把识别文字喂回主模型，无感、可配置、升级不丢

Vision bridge: 拦截粘贴/拖入/导入的图片，交给 OpenAI 兼容视觉模型识别为文字并自动发送进对话；主模型可通过 vision_ask 工具复核追问。

DeepSeek Harness plugin: when the main (text-only) model receives images, automatically caption them with a Qwen VL model and let DeepSeek continue the task.

Vision bridge for text-only models (DeepSeek and any route without image input): a pluggable vision model (Gemini, Groq, or any OpenAI-compatible endpoint) describes attached images at the llm/stream seam and provides the describe_image tool.

视力屈光度

Transparent image guard + vision analysis for DeepSeek Harness: text-only models read pasted images without the 400 session deadlock.

DeepSeek Harness 视觉集大成插件：路由 + 工具 + 降级链 + 免费兜底。像素级视觉工具（Q&A/定位/裁剪/像素对比/取色/OCR/SVG 矢量化/抠图/HTML 截图/长截图 OCR）、五级供应商降级链（用户模型 → Ollama → LM Studio → 自定义 HTTP → OVH 匿名免费兜底）、自动识图模型组、图片桥与桥接内联预览，一个插件全包。