视觉、语音与多模态

Automatic tier-based model routing for DeepSeek Harness: a virtual `smart` model classifies each request by difficulty (hard / normal / easy) and by vision need, then delegates it to the models you already configured. 三级难度 + 视觉自动路由，虚拟 smart 模型零配置接入。

DSH voice input plugin: browser realtime + VAD dictation + local/cloud ASR + DeepSeek AI polish

System-audio driven UI visualizer for DeepSeek Harness: a draggable 48-band spectrum chip plus a frame-wide bass glow (memory-only FFT).

Adds a file-picker 'attach image' button to the composer toolbar, since DSH's own web client only supports paste and drag-and-drop for image attachments (v1 scope, per its own upstream design notes) - a real gap for touch devices with no drag-and-drop and

DSH (DeepSeek Harness) ComfyUI 桥接插件：让 Agent 直接驱动本地/局域网 ComfyUI 生图生视频 — 8 个全局模型工具、预设工作流模板(txt2img/img2img/Wan/SVD/H3)、任意 API 工作流逃生舱、设备能力守卫、模型注册表断点续传下载、ComfyUI 一键安装评估。ComfyUI bridge plugin for DeepSeek Harness: image & video generation tools driving a local

DeepSeek Harness Web plugin that keeps image-bearing Tool results visible after Compact transcript folding

Xiaomi MiMo-powered voice for DeepSeek Harness: browser 🎤/🧠/🔊 UI, voice_transcribe/voice_understand/voice_speak tools, configurable voice map (preset/voicedesign/voiceclone). Fork of zhuiyueya/dsh-voice (MIT), Settings pattern from Anionex/dsh-vision-toolkit (MIT).

在准入前把提示词里的图片换成视觉模型的文字分析，因此任何模型（包括纯文本模型）都能读图且会话不切换模型；另提供 describe_image 工具处理图片路径。

发送前把请求中的历史图片裁剪至保留张数，并从上游 400 错误解析单次图片上限、按更少张数降级重试，保证含图会话不被图片数量限制打断。

Three tools for DeepSeek Harness: typeset LaTeX formulas into images, draw mathematical figures from a declarative spec, and convert a formula or SVG into an image ready to embed in a document.

Voice input plugin (dual-face): mic button beside the composer send action, live recording bubble streaming PCM to a local Qwen3-ASR python service managed by the host, plus an ASR environment settings page (clone runtime/model repos, create venv, run ser

AI 回复流式 TTS 朗读插件，思考等待期有文字与语音同步的趣味短语反馈。内置 Edge TTS，支持任意 OpenAI 兼容云端引擎。

在设置中配置多个生图供应商，对话通过 image_generate 调用当前选中的模型；图片保存到工作区 generate/image，并在对话中内联显示。

DeepSeek Harness plugin: turn lines of narration into a finished vertical reel. Free neural voice-over, burned-in captions, no API keys.

Minimal English-only voice input for the DeepSeek Harness Web UI: a mic button in the composer that transcribes speech into the draft via the browser's native SpeechRecognition API. No dependencies, no subprocess, no network calls beyond whatever the brow

DSH 对话结束提示音：内置网络热梗原声（你干嘛哎呦、鸡你太美、神鹰哥等），设置面板可试听/切换/随机，支持自定义音频目录。

Voice for the DeepSeek Harness: speak to the agent, hear it back, and see what it is doing from a floating window that stays on top of whatever you are running.

DeepSeek Harness (DSH) 会话内图片渲染插件：全局 show_image 工具 + 点击放大 lightbox。 | Inline image rendering plugin for DSH: global show_image tool + click-to-enlarge lightbox.

DeepSeek Harness plugin: a model-facing show_image tool that presents a local image to the human viewer in the conversation.

DSH 网页端语音控制：按住说话将语音转文字输入发送框（可自动发送），用浏览器语音合成朗读助手回复；右键麦克风打开设置面板，支持 Ctrl+M 全局快捷键。

视觉桥：贴图先经 deepseek-v4-flash-vision-exp 转写为文字，再交给纯文本的 deepseek-v4-pro 回答，零第三方依赖。

苹果设备端 Vision 工具：本地 OCR（含中文）、图像分类、人脸检测、文档版面分析与综合描述，100% 离线、无需 API key，支持长截图切片。

面向模型的图像生成工具，提供可配置通道和标准化图像参数。

通过已配置的 DSH Provider 或 OpenAI 兼容端点提供图片输入与识别。