시각 도구 · DeepSeek Harness 플러그인 — 20 페이지

DSH 对话内生成图片：Agent 调用 generate_image 工具，经自定义 OpenAI 兼容端点调用图片模型（gpt-image-1 / gpt-image-2 等），图片保存到工作区

Let the model render an image into the DSH web chat flow (QQ/WeChat style) via a show_image tool: the message content keeps only the path text, the image never enters model context, and the browser shows the picture inline, scrollable with history.

Multimodal plugin for DeepSeek Harness: understand images and generate images through configurable OpenAI-compatible or DashScope endpoints.

Give your DSH agent eyes: search Wikimedia Commons and Openverse, download a real photo, and verify it is a real image before you use it.

Multi-engine reverse image search aggregator for DeepSeek Harness. Turn one public image URL into Google Lens / Baidu / Yandex / TinEye / SauceNAO / IQDB / Ascii2d search links.

生图工作室：本地参考库（人物 / 服饰 / 场景 / 道具）+ 智能组合参考图 + 调用生图 API，模型可自行判断何时发图；设置页里可视化管理素材库。

Let DSH text-only-model conversations accept dragged/pasted images: save them into the session workspace and insert the file path as text, bypassing the webui attachment channel that otherwise rejects image parts.

对话内联图片:LLM 回复中输出的本地图片路径直接在消息正文内渲染为图片,支持 9 种格式、点击放大灯箱、可调尺寸

Labnana image generation for DeepSeek Harness: text-to-image / image-to-image / precise editing with credits estimation, subscription balance and web settings UI.

DeepSeek Harness bundle for a Windows desktop UI context picker and MCP server for Codex, DeepSeek Harness, and AI agents

DeepSeek Harness 插件：把手机端 Local Dream 的本地 NPU 生图接进对话，一句话出图、图生图、放大与分块重绘，全程离线不计费。

Generate images and videos in DSH chat via OpenAI-compatible providers configured in Model settings; dedicated Settings menu and workspace media_gen output.

Native MiniMax for DeepSeek Harness: chat with an M Plan subscription or a pay-as-you-go API key, image, video, speech and speech-to-text tools, and M Plan usage in the sidebar.

Per-model reasoning-effort and vision capability editor for DSH pi-ai provider profiles.

Choose which models appear in the chat composer's model list, with per-provider groups, per-model capability tags sourced from the adapter, the relay model list, a built-in catalog and vendor corrections, separate coloured labels for image/video generation models, name search and capability filters, and an image-input alignment (on by default) for relay routes.

Modellix Design, LLM, and Web providers for DeepSeek Harness

DeepSeek Harness tools for the local Moli headless browser: rendered fetches, markdown, semantic trees, screenshots, and PDFs.

WeChat Mini Program automated testing for DeepSeek Harness (dsh): selector-addressed actions, build-freshness gates, geometry-first assertions that work on text-only models, and screenshot image blocks on vision routes. 微信小程序自动化测试 dsh 插件：选择器寻址、构建新鲜度门、面向纯文本模型的几何断言、视觉路由附真实截图。

Register models, assist with portraits, and select the Agent model from a secret-free catalog for DeepSeek Harness.

Obsidian-style notes panel for DeepSeek Harness Web: vault directory tree, Markdown preview/edit/save, cite-to-chat, batch archive/delete of sessions and workspaces. DSH 内嵌 Obsidian 风格笔记面板：目录树 + Markdown 预览/编辑/保存 + 引用到对话 + 会话/工作区批量操作。

OCR via OpenAI/Anthropic vision APIs for dsh: ocr_recognize tool (percent coords + two-diagonal-point boxes, auto region re-read on low confidence), a collapsible settings card in Settings → Plugins (API format / URL / Key / Model, fetch model list, test connection). Hot-pluggable — mounted via ~/.dsh/profiles/web/cordis.patch.yml + a profile node_modules symlink, no dsh source changes.

泡泡（PaoPao）：给 DeepSeek Harness 的 QQ 通道 + 陪伴层 —— 按权重自己开口、自己找 B站 视频、扫码登录与私信监看。QQ channel plus a companionship layer for dsh.

A lightweight DeepSeek Harness vision delegation tool for text-only routes, with native OpenAI Responses, Chat Completions, and Anthropic Messages adapters.

Open Eyes for DeepSeek Harness: delegate images to a configurable multimodal model through OpenAI Responses, Chat Completions, or Anthropic Messages.