DeepSeek Harness Plugin Hub

Publish and manage complete Harness Profiles. Discover Plugins for your next setup.

Explore

PluginsPresetsDocsNews

Community

Publish a pluginContactReport an issue

Resources

Plugin Hub on GitHubDeepSeek HarnessSystem statusPrivacy notice
© 2026 DeepSeek Harness Plugin HubPowered byPaxTech

Independent and unofficial. Not affiliated with, authorized by, or endorsed by DeepSeek.

Video Understand — DSH Plugin for DeepSeek Harness
DeepSeek Harness Plugin Hub
ProfilesPluginsCategoriesNewsDocsSign inManage Profiles
ProfilesPluginsCategoriesNewsDocsSign in
← Plugins

dsh-video-understand

Video Understand

Low-cost video understanding tool: Bilibili links/BV/local videos → information layer (ASR + scenes + object trajectories + YOLO) → summaries + Q&A. Question-driven dynamic routing across levels (L0/L1/L2), semantic-layer reuse, and budget caps. Self-contained engine with no external dependencies.

The plugin will be installed here. Keep web if you are unsure.

npx -y @deepseek-ai/dsh plugin --profile web add dsh-video-understand@0.5.2
READMECompatibilityVersions

Compatibility and provenance

Video Understand is published as dsh-video-understand and currently resolves to version 0.5.2. The Hub verifies its manifest and preserves the exact installation source for reproducible installs.

DSH compatibility
*
Runtime surfaces
any
Release source
npm
Registry updated
9/20/2026

Versions

0.5.2stable
8/20/2026
0.5.0stable
8/19/2026
0.2.0stable
8/16/2026
Show 2 more versionsCollapse versions
0.1.1stable
8/16/2026
0.1.0stable
8/16/2026

Related plugins

Loading related plugins…

Latest
0.5.2
DSH
*
HMR
Process restart
Tree shaking
Safe tree shaking not declared
Unpacked size
6.8 MB
Files
37
Surface
any
License
MIT
Source
npm
GitHub
★ 8
Weekly downloads
185
Security scan
✓ v0.5.2 scan passed
Last push
8/24/2026
View source ↗Project homepage ↗
README badge

Click the badge to copy Markdown for your README.

Do you maintain this Plugin?Claim benefit · Priority security scan

Verify the GitHub repository declared in package.json to manage this listing. After you claim it, Hub will prioritize a security scan of the current version and publish the result when it passes.

Claim this Plugin →
Report an issue

Related plugins

More verified plugins in vision-media.

Tool Describe Image@linxin666/dsh-tool-describe-imageModel-facing describe_image tool for the dsh web GUI: gives a text-only model image understanding by asking a vision-language model at an OpenAI-compatible endpoint to describe one image (local path, http(s) URL, or attachment reference). Hot-pluggable — Deepseek Ivideodeepseek-ivideoiPolloWork HyperFrames Video Studio and 27 editable video templates as a native DeepSeek Harness conversation view.Imagegen@dickpy/dsh-imagegenAI image generation plugin for the dsh web GUI: text-to-image and image-to-image through configurable provider channels (gpt-image-2 / grok-imagine-image / nanobanana series / seedream-5.0-pro / dall-e-3, with native xAI Grok Imagine, Google Nano Banana aVoice Modedsh-voice-modeFull-duplex voice plugin for DeepSeek Harness: local zipformer2 streaming ASR (no API key) → editable draft; Edge TTS or local VITS / Kokoro read-aloud with live captions; true barge-in; hardened HTTP surface + model SHA256 pinning; compatible with all ds

README

dsh-video-understand

低成本视频理解插件:给 dsh agent 注册 video_understand 工具——B站链接 / BV 号 / 本地视频 → AVIS 信息层(ASR 转写 + 场景结构 + 运动对象轨迹 + YOLO 语义)→ 摘要+问答。引擎自包含,无需外部依赖。

安装

# npm 安装
npm install dsh-video-understand

# 或 dsh plugin add dsh-video-understand

装完即用:首次调用 video_understand 时自动创建插件本地隔离环境(.venv)并安装核心依赖(优先 uv,回退 venv+pip 清华镜像)——无需手动执行任何命令,不污染系统 Python。

引擎已内含,无需额外克隆外部仓库。

环境自检

npx dsh-video-understand doctor        # 逐项检测 + 给出修复命令
npx dsh-video-understand doctor --fix  # 一键自动修复(建环境 + 装依赖)

前置条件仅两个:ffmpeg(macOS brew install ffmpeg / Ubuntu sudo apt install ffmpeg)和一个 LLM API key(见下表)。语义层依赖(torch/CLIP/YOLO,约 2GB)为可选,仅建完整语义层时再装:pip install -r engine/requirements-layer.txt。

⚠️ 数据流披露

级别数据流向说明
L0(默认)完全本地ASR + 场景分类 + 运动检测,不上传任何数据
L1MiMo API视频帧发送至 MiMo 服务器进行 VLM 分析
L2MiMo API视频帧发送至 MiMo 服务器进行 VLM 分析
  • L0 级别(默认)仅使用本地 ASR + 场景分类 + 运动检测,不涉及云服务
  • L1/L2 级别会将视频帧(JPEG 编码)发送至 MiMo API 进行视觉理解
  • 帧数据仅用于单次 VLM 推理,不会被存储或用于训练

环境变量

变量必需说明
LLM_API_KEY✅*LLM API 密钥(默认 MiMo:sk-xxxxx)
LLM_API_URL❌API endpoint(默认 https://api.xiaomimimo.com/v1/chat/completions)
LLM_MODEL❌模型名(默认 mimo-v2.5)
VIDEO_UNDERSTAND_PYTHON❌Python 解释器路径(自动检测有依赖的 Python)
BILI_DOWNLOAD_SCRIPT❌bilibili-downloader 脚本路径(下载 B站视频用)

* 未设置 LLM_API_KEY 时,自动从 ~/.dsh/.credentials.yaml 读取(优先 XIAOMI_API_KEY,兼容 DEEPSEEK_API_KEY)。一个 key 搞定。

工具

video_understand(target, questions?, noDownload?, level?, window?)

参数类型说明
targetstringB站 URL / BV 号 / 本地视频绝对路径
questionsstring[]可选,自定义问题(默认 3 问)
noDownloadboolean本地文件置 true
levelstringl0(默认) / l1(+3-5帧VLM视觉摘要) / l2(+时间窗密集帧证据)
windowstringL2 时间窗,如 10-30 或秒数(auto=轨迹最活跃30s)
budgetCnynumber单次问题预算上限(元),视觉成本估算超预算自动降级(拦截 L2 用 L0/L1 回答)

返回 JSON:video / duration_s / token_compression_pct / cost_cny / answers[]。

结构

dsh-video-understand/
├── package.json
├── cordis.patch.yml
├── dsh/
│   └── index.js          # host 端:注册 video_understand 工具
├── engine/               # 自包含引擎(无需外部依赖)
│   ├── understand_video.py
│   ├── avis.py
│   ├── visual_level.py
│   ├── frame_prep.py
│   └── livestream-highlight/
│       └── asr.py
└── skills/
    └── video-understand/SKILL.md

分级实测(2026-08)

层级内容数据流向成本(估算*)
L0 信息层ASR+场景+轨迹+YOLO → 摘要/问答本地仅 LLM 文本成本
L1 视觉级3-5 帧 VLM → 颜色/姿态/衣着MiMo API+数帧 VLM 成本
L2 证据级时间窗密集帧 → 时间线MiMo API按窗长

* 成本按 engine/understand_video.py 的价格常量估算(可配置输入),实际随厂商定价变化,非测量承诺。详见 docs/blog-视频理解性价比实验(方法 + 实测)。

实测:电影解说 L1 补出「白色立领衬衫/神情凝重/暗色调诊室」(L0 完全给不出);舞蹈 L2 逐帧「头部转 15-20°→45°、口型开口→闭合→微笑」。

设计背景

本插件的核心目标是低成本视频理解:用信息层代替逐帧像素喂 LLM,单视频 LLM 调用仅需几千 token(具体成本取决于所选模型定价,见上)。

  • LLM 选型:最初选用 DeepSeek v4 Flash 因其性价比最优;近期调价后已迁移至 MiMo,持续追踪成本效益比
  • 视觉级(L1/L2):DeepSeek 暂不支持多模态输入,故 L1/L2 使用 MiMo API 进行视觉分析(此前曾使用 DashScope 作为过渡方案)。未来若 DeepSeek 支持多模态或出现更优的本地推理方案,可进一步优化成本

原理

引擎把视频压缩成信息层(ASR 转写 + 场景结构 + 运动对象轨迹 + YOLO 语义,约 1k token)再喂 LLM;同一视频重复理解时信息层缓存复用(内容哈希,二次提问跳过 ASR)。成本与 token 对比的具体测量见 docs/blog-视频理解性价比实验。

License

MIT