시각 도구 · DeepSeek Harness 플러그인 — 27 페이지

High-performance, streamlined Antigravity OAuth LLM routing and Nano Banana 2 image generation capability bundle for DeepSeek Harness

image_recognize tool for DSH: recognize image content via a configurable vision model (OpenAI-compatible API), with auto/ocr/describe modes, magic-byte content validation, and sandbox-compliant file reads through ctx.fs.

Image generation for DeepSeek Harness via OpenRouter's unified Image API: text-to-image and image-to-image with capability-gated parameters, workspace file output, an inline chat card and a settings card

Eight DeepSeek Harness plugins: persona, language guard, vision fallback, python workdir guard, windows encoding guard, cross-agent memory, image generation, and skill shell injection.

One-click DSH Web background switcher: built-in gradient/solid presets plus a local-image background with opacity, blur, and fill controls.

DSH web UI 定制插件：全屏壁纸背景（设置中可切换图片、透明度与比例，持久化到 $DSH_HOME/uwsp-settings.json）

驯兽师（驯兽场）：把人类语言翻译给智能体的 DSH 插件——注入「开智」内核（第一性原理 + 需求剖析 + 极简沟通 + 最少 token），左上角悬浮萌宠，设置页可配内核与开关。

Wallpaper plugin for the DeepSeek Harness web GUI: translucent app surfaces over a user-picked local image, GIF, or video background

Unlimited-OCR long-document parsing with a native DeepSeek Harness tool and GUI configuration.

Give text-only DeepSeek Harness agents video understanding: scene-aware frame sampling + VLM + optional ASR transcript fused into timeline evidence. / 给纯文本模型的视频理解插件（场景感知抽帧 + VLM + 可选语音转录）

Vision bridge for text-only models: describe attached images via another vision model or system OCR

Model-facing analyze_image tool for the DeepSeek Harness: multi-modal image understanding via any OpenAI- or Anthropic-compatible vision API, with 8 analysis modes, local path / http(s) URL / data URL input, and a Web UI hint that guides image-incapable models to the reliable local-path route.

DeepSeek Harness plugin: auto-route image-bearing requests to deepseek-v4-flash-vision-exp, then fall back to the original model.

DSH 识图桥接插件：让 DeepSeek 等纯文本模型会话也能发送图片——图片在进入模型前自动交给外部视觉模型（阿里云百炼 qwen-vl）识别，识别结果以文字形式喂给当前模型。

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。

Give a text-only dsh model eyes: pasted images are recognized into text via an OpenAI-compatible vision endpoint (Zhipu free by default) before the message reaches the model.

DSH 视觉插件（Edge 豆包桥接）：通用识图 + 数学建模图专项（几何图形/流程图/图表/表格/公式）+ 不确定项澄清闭环。零成本，免 API Key。

DSH 静默视觉增强：主模型照常选择，图片自动交给固定视觉模型后以隐藏上下文返回主模型。

Deployment-level vision plugin for DeepSeek Harness: registers the vision_analyze tool backed by a configurable multimodal model, with a settings-page UI. Zero dependencies.

DeepSeek Harness plugin: (1) installs an image-input (input modalities) checkbox into the Models settings catalog so a custom model can declare input:[text,image] and DSH passes images straight to it — the core feature; (2) optionally registers an OpenAI-compatible vision relay provider + vision_analyze/vision_status/vision_test/vision_configure tools. Host-only bundle (no settings card).

dsh-vision-plugin: give DeepSeek Harness text-only models a pair of eyes — pasted images are transcribed by a vision model before they reach a text-only main model, plus the vision_analyze tool and a bilingual settings page.

Native interactive visual-reasoning plugin for DeepSeek Harness: precise pixel grounding (SOM grid, zoom, annotate, measure, diff, color, OCR) + MiMo V2.5 multimodal backend, with zero external MCP servers.

Eyes for text-only DeepSeek Harness agents: built-in free vision chain (no key) + pixel-level vision tools (Q&A, grounding, crop, pixel diff, colors, OCR, SVG trace, cutout, screenshots). One-command install, no Python, image turns work like ordinary tool-calling turns.

vision tool for DeepSeek Harness agents: describe local images via any OpenAI-compatible vision endpoint you configure (no built-in keys, optional multi-model cross-check)