視覚ツール · DeepSeek Harness プラグイン — 19 ページ

Immutable command-line handoff from a dsh launcher to any app plugin that injects cmdlineArgs

Worker-thread implementation of the DeepSeek Harness code-execution seam

GUI-grounding vision model seam (ctx.guiModel): calls an OpenAI-compatible vision chat-completions endpoint with a screenshot and a task, and returns one structured next action

The dsh one-shot bundle: a direct core Agent/Session runner over dsh-base with no Host, HTTP, or browser layer

Vision-augmented DeepSeek adapter plugin for DeepSeek Harness: a vision-capable model describes image input, then a text-only DeepSeek model reasons over the description

本地 LM Studio 视觉识别插件：连接 LM Studio OpenAI 兼容接口（默认 http://127.0.0.1:1234），用视觉模型（自动使用最后加载的模型，不预设具体模型）识别本地图片与剪贴板图片。agent 工具 lmstudio_vision / lmstudio_clipboard / lmstudio_models + 设置页「LM Studio 视觉识别」面板；配置持久化到 ~/.dsh/dsh-lmstudio-vision.json。Hot-pluggable — 挂载方式同 dsh-ssh：profile node_modules link + dsh.profile.bundles，无需修改 dsh 源码。

DSH Pet Corner: a floating pet, keyless pet-image proxy, favorites, and plugin-owned settings API

Model-facing consumer of the ctx.macosUse / ctx.guiModel / ctx.browserUse capability seams: bounded desktop and browser primitives plus vision-model-driven computer_use_task and browser_use_task tools

DSH plugin: vision_read — route image reading to a dedicated vision model (e.g. Kimi K3) so text-only agents can see images

Codex-style image offload for DeepSeek Harness: routes chat-box images to a user-configured vision model and feeds the text description back to a text-only primary model (e.g. deepseek-v4-pro).

The dsh browser-surface bundle: the web patch layer over dsh-base plus the runtime glue plugin (frontend dist serving, web-surface prompt, bash runtime variables, URL line)

DeepSeek Harness plugin — a live, stealth-capable Chrome inside the conversation. Tiered frame transport, Patchright engine seam, four-tier CAPTCHA pipeline with interactive human handoff, and a page-owned browser panel the user can grab the mouse in. Tested against DSH 0.1.5-rc.2.

Chat with, monitor, and approve your DSH agents from WeChat — an iLink gateway + conversation node bundle for DeepSeek Harness

On-demand image generation for DeepSeek Harness (DSH): a generate_image tool that calls your own OpenAI-compatible image URL + API key and delivers the image into the session

UI customizer for the dsh web GUI: a sidebar menu above Settings that sets a full-page background image with adjustable opacity

ComfyUI 驱动器（AI-first 合同制）：模型只见合同槽、不见工作流 JSON。文生图/图生图/文生视频/图生视频模板，双重预检防幻觉，官方/自定义节点活体判定。

DeepSeek Harness 文档拖拽读取插件：捕获阶段接管文档拖拽（PDF/Word/Excel/文本），双轨（Inline/Mounted）上下文注入与 doc_read 自主检索

Drag, drop or paste images / PDFs / office docs / zips / media into DeepSeek Harness - they reach a text-only model as workspace file paths. | 中文：把图片、PDF、Office 文档、压缩包、音视频拖入或粘贴到 DSH，转成工作区文件路径交给纯文本模型。

Lift the 'model does not support images' gate in the DeepSeek Harness web GUI: declares image input for every llm-pi-ai model that does not state its own input modalities, so attachments are admitted, and textifies image blocks for the text-only opencode-go upstream (pair with a vision skill for actual recognition). Toggleable by the claude-vision-skill directory's presence: skill enabled = vision-skill flow, skill disabled = native vision model.

DeepSeek Harness plugin: click-to-zoom lightbox for markdown/chat images

通过 QQ 消息远程控制 DeepSeek Harness：命令执行(/exec)、任务派发(/ask)、阶段进度汇报、屏幕截图发送(/screenshot)、AI 聊天(/chat)与聊天特化会话(/chatbind)、会话标题化管理(/newsession /title /session)。OneBot 11(NapCat/Lagrange/go-cqhttp)，零 npm 依赖。

Vision for text-only DeepSeek: a view_image tool that answers questions about images via any OpenAI-compatible VLM endpoint (defaults to the free Zhipu glm-4.6v-flash; DashScope, Ollama and others supported)

主题换肤插件：修改 Harness 界面主题颜色（品牌色/底色/文字色等），并设置背景图片（支持 GIF/APNG/动态 WebP 等动态图）与背景视频（MP4/WebM 等）

让不支持图片输入的主模型通过已配置的视觉模型理解图片：自动把消息中的图片替换为带像素坐标系的文字描述，并提供 vision_ask 追问工具与设置页