dsh-vision-all：为 DSH 里所有 LLM Provider 统一补上图片输入
介绍 GGBound-1229 维护的 DSH 插件 dsh-vision-all。当 OpenAI 兼容中转站未在元数据中声明视觉能力时，DSH 会关闭粘贴图片入口；该插件在 internal/config 阶段为所有 provider 合并 defaultInput，补上 image。说明安装步骤（pnpm/npm 从 GitHub 安装并写入 bundles）、日志验证、环境变量 DSH_VISION_ALL_LOG、关闭方式，以及上游必须真实支持视觉的前提。
Read Moresearch-baidu：基于腾讯 ADP 的多模态生成 Skill
search-baidu 是一个封装腾讯 ADP 平台能力的 Skill，支持文生图、图生图、文生视频、图生视频、文生 3D、图生 3D 全场景生成。它自动路由到混元 image.hy、GI 版 image.gi、可灵 video.kling、VIDU video.vidu、混元 3D 3d.hy 等模型，通过 submit+轮询 query 返回成品 URL。支持通过关键词自动识别用户意图，无需明确说“用 ADP”。
Read Moredsh-send-photos：让 DSH 接收未被声明为多模态的图片
让 DSH 接收未被声明为多模态的图片。该插件补全中转站缺少的模态声明，在 llm-pi-ai 的 modelOf() 方法外套一层，将 image 加入 input 列表，使 DSH 允许图片通过。支持空数组覆盖全部路由，也支持通过 providers 精确限定。安装为 bundle 包，卸载即可还原，不修改请求体。
Read Moredsh-vision-unlock：让未声明多模态能力的模型也能发图
dsh-vision-unlock 通过运行时接缝给未声明多模态能力的模型补上 image 能力声明，让 DSH 不再因缺少该字段而拒绝图片。不修改 app.asar 或配置树，适用于自定义网关路由、本地适配器等场景。
Read Moreomni-reader 多模态文档解析：把文件与 URL 统一转成 Markdown
omni-reader多模态文档解析（@org-y1eh5fyg/cue-omni-reader，v1.8.8）是 SkillHub office-efficiency 类企业 Skill，支持上传或 URL 将 PDF、Office、图片、扫描件、网页、压缩包及音视频等解析为 Markdown，强调目录与复杂表格还原、默认不落盘，音视频可转文字稿并具备画面理解。需 API Key（requires_api_key=true）。本文基于官方简介说明安装：skillhub CLI 或按 skillhub.md 安装至 Cursor 等 skills 目录，并概括典型用法与注意事项。
Read MorePPT图片内容识别：把幻灯片导出为 PNG 并用多模态模型读懂画面
PPT图片内容识别是 SkillHub 社区 Skill（@user_46974ee3/ppt-image-extract，v1.0.0，office-efficiency）。官方简介：读取 PPT 时先导出每页 PNG，再用多模态模型识别图片内容，基于 tiangong-wps-ppt-automation。标签 requires_api_key=false。文章说明定位、能力、skillhub 安装命令与各客户端 skills 目录，并提示以 SKILL.md 为准核对环境与识别结果。
Read More大模型视觉问答技能：在 Agent 里对图片做开放式 VQA
介绍 SkillHub 社区 Skill「Large Model Visual Question Answering Skill | 大模型视觉问答技能」（@user_bb47e3e3/visual-qa-analysis，v1.0.4）。说明其基于 CV 与 LLM 对本地或 URL 图片做开放式问答、通过 open-id 与可选 API 密钥调用云端服务，以及用 --list 查询历史记录。涵盖 skillhub 安装命令、Cursor/Claude Code/Codex 的 skills 目录、典型 python -m scripts.visual_qa_analysis 示例，并提醒 requires_api_key、禁止用本地 memory 替代云端历史等约束。
Read MoreBird Recognition Tool：在 Agent 里识别图片与视频中的鸟类
Bird Recognition Tool（鸟类识别工具）是 SkillHub 社区 Skill @user_bb47e3e3/bird-recognition-analysis（v1.0.8），用于识别目标区域图片与视频中的鸟类，支持不低于 500 种常见鸟种及定制化模型训练，标签 requires_api_key=false。本文基于 SkillHub 公开信息说明定位、能力与适用场景，并给出 skillhub CLI 安装到 Cursor、Claude Code、Codex 等 skills 目录的步骤，以及通过官方安装提示词让 Agent 加载的方式；用法遵循 SKILL.md，在对话中提供媒体路径即可驱动识别。
Read Moredsh-llm-multimodal：在 DSH 聊天里接入六类多模态生成工具
dsh-llm-multimodal 是 xiaokaizhou 维护的 DSH 社区插件（模型推理），在对话中提供 generate_text、generate_image、generate_video、generate_vision、generate_tts、generate_music 六个工具。除文本走 harness LlmRuntime 外，其余多基于 OpenAI 兼容 API，并从 llm-pi-ai 配置按模型 id 关键词自动发现能力。安装推荐 dsh plugin --profile web add xiaokaizhou/dsh-llm-multimodal；可与 dsh-media-studio、media-preview 组成生成与预览闭环。MIT 许可，当前资料版本 0.2.0。
Read MoreIntelligent Video Search & Retrieval Analysis Skill：面向目标与语义描述的视频检索
SkillHub 社区技能 @user_bb47e3e3/video-search-analysis（v1.0.8）面向长视频按目标与自然语言做智能搜索，支持常规目标检索、描述检索与向量化匹配，并可经脚本查询云端历史报告。本文基于官方页面与 SKILL.md 说明安装路径（skillhub install）、Cursor 等 skills 目录、以及在技能根目录运行 python -m scripts.video_search_analysis 的典型命令与注意事项。
Read More即梦 Seedance 2.0 视频提示词：为多模态视频生成整理可执行提示词
SkillHub 社区 Skill「即梦 Seedance 2.0 视频提示词」（@user_9cf3307c/seedance-video-prompt-zh，v1.0.0）面向即梦 Seedance 2.0，在用户以文本、图片、视频、音频等多模态输入创作视频提示词时触发，覆盖 @ 引用、运镜复刻、特效模仿、延长与编辑、音乐卡点及电商、短剧、科普等场景。本文说明定位、能力、基于 skillhub CLI 的安装路径（含 Cursor 等）及典型对话用法，并注明 requires_api_key=false 与社区来源限制。
Read Morekling v3 omni：在 Agent 中接入可灵 v3 Omni 多模态视频生成
kling v3 omni（@user_f95f21b8/dlazy-kling-v3-omni，v1.0.9）是 SkillHub 上的 design-media 类社区 Skill，通过 dLazy CLI 调用 Kling v3 Omni，支持提示词与图片等多模态输入生成动态视频。文章说明 Skill 定位、核心 CLI 参数、SkillHub 与 @dlazy/cli@1.0.9 安装路径、API Key 配置，以及官方命令示例与异步、计费相关注意点。
Read MoreSeedance 2.0 视频提示词撰写指南：为多模态视频生成整理提示词结构
SkillHub 社区 Skill「Seedance 2.0 视频提示词撰写指南」（@user_d5c3472c/seedance-prompt，v1.0.0）面向即梦 Seedance 2.0，在用户以文本、图、视频、音频等多模态输入写视频提示词时触发。涵盖 @ 引用、运镜复刻、特效模仿、延长与编辑、音乐卡点及电商广告、短剧、科普等场景；标注 requires_api_key=false。文章说明其定位与能力边界，给出 skillhub CLI 安装命令及 Cursor、Claude Code、Codex 等 skills 目录，并示例如何通过对话触发结构化提示词输出。
Read Moreseedance 2.0：在 Agent 里调用 ByteDance Seedance 2.0 视频生成
seedance 2.0 是 SkillHub 社区 Skill（@user_f95f21b8/dlazy-seedance-2-0，v1.0.9），通过 dLazy CLI 调用托管 API，对接 modelId doubao-seedance-2_0。支持多模态参考（图/视频/音频）、首尾帧与文生视频，可配置分辨率、画幅与时长。安装需 skillhub install 到 Agent skills 目录，并安装 @dlazy/cli@1.0.9、配置 dLazy API Key。适合在 Cursor 等环境中用自然语言驱动成片，需注意云端上传与账户余额。
Read MoreQwen-VL 图片识别：为纯文本 Agent 补上视觉理解
介绍 SkillHub 社区 Skill「Qwen-VL 图片识别」（@user_213bfd3b/qwen-vl-image-reader，v1.0.2）。它通过阿里通义千问 Qwen-VL 把截图、文档、图表等解析为结构化 Markdown，供 DeepSeek 等纯文本 Agent 继续推理。文章说明定位、核心能力、requires_api_key 要求，并给出基于 skillhub.cn 官方文档的 CLI 安装步骤与 Cursor 等客户端 skills 目录示例；典型用法与配置细节以安装后的 SKILL.md 为准。
Read MorePrompt Generation Master: A Unified Multimodal Prompt Skill for Images and Videos
Prompt Generation Master (@user_5d00e27d/image-prompt-skill, v1.0.15) is a SkillHub community skill designed as a unified prompt generator supporting both image and video modality: text-to-image and image-to-image, text-to-video and image-to-video, image/video decompilation, and context fine-tuning. It claims automatic adaptation to art styles and platform syntax, with built-in RAG-based style presets and safety validation; requires_api_key=false. Based on the official SkillHub introduction and installation documentation, this article outlines its capability boundaries, the skillhub CLI installation command, and the skills directories for different Agents, providing a reproducible usage workflow and precautions.
Read Moregemini-api-dev: An Official Development Skill for the Gemini API
gemini-api-dev is an official Agent Skill in the Google gemini-skills repository, using Critical Rules to enforce the current model name, the new google-genai SDK, and the interactions API. It covers text, multimodal, function calling, structured output, and migration from generateContent. The article explains the installation command npx skills add, Python/TypeScript Quick Start, multi-turn previous_interaction_id, and Deep Research background task mode, and reminds users of the store default and the policy for disabling legacy models.
Read Moreopencode-eyes: Giving Text Models a Pair of Eyes (StepFun step-3.7-flash)
The DeepSeek Harness plugin, opencode-eyes, provides visual capabilities for pure text large language models by mounting StepFun’s step-3.7-flash model to enable multimodal completion. The plugin packages an MCP server for a ready-to-use setup. To use it, users need to configure the `STEP_API_KEY`, with support for local image paths and custom prompts. The plugin allows adjusting parameters such as timeout, image size, and quality, while users should be mindful of runtime directory and permission limitations.
Read Moredsh-wsl-shot: Save Windows Clipboard Images as WSL Files
dsh-wsl-shot is a DeepSeek Harness plugin designed to address the issue that agents in a WSL environment cannot directly read images from the Windows clipboard. It uses tool calls to save Windows clipboard images as files accessible in WSL, enabling agents to perform multimodal analysis. Installation requires DSH version >= 0.1.2, and the service must be restarted before use. This plugin is part of the dsh-wsl-kit suite and must be used with a multimodal model.
Read Moredsh-comfyui-image: Run ComfyUI Locally to Generate Multimodal Assets
The DeepSeek Harness plugin `dsh-comfyui-image` enables agents to directly control local ComfyUI, leveraging GPU resources to generate images, videos, audio, and 3D assets without cloud API costs. The plugin supports 116+ workflow templates across 67 model families, providing core features such as text generation, template browsing, status reporting, and skill decision-making. Installation requires Harness 0.2.0-rc.2 or later, and supports automatic detection of ComfyUI status or launching headless mode. Environment variables allow flexible configuration of the port and model paths. Models are downloaded from mirror sites by default, and local generation tasks can be invoked directly using natural language instructions.
Read MoreDeepSeek Harness Plugin: dsh-vision-autoswitch — Automatically Switches to a Vision Model on Image Input, Then Reverts After the Turn Ends
`dsh-vision-autoswitch` is a DeepSeek Harness plugin maintained by CultOfLuna, designed to eliminate the cumbersome process of manually switching models in multimodal conversations. When it detects that an input contains images, the plugin automatically routes the current turn to `deepseek-v4-flash-vision-exp` and switches back to the user’s originally selected model after processing. It supports real-time display of the current model, leaves text-only turns unchanged, and requires manual installation by editing the configuration file. Its automated routing effectively lowers the barrier to using multimodal conversations.
Read MoreDeepSeek Harness Iris Plugin Introduction
@mokuyoaxis/dsh-iris is a plugin that provides multimodal capabilities for DeepSeek Harness (DSH), integrating features such as image generation, video processing, speech synthesis, and visual understanding. Through a plugin-based architecture, it centralizes complex media production tasks in the Iris workbench, supporting the full pipeline from content generation to visual analysis. Installation requires Node.js >= 22.0.0, a compatible host version, and dependencies on ffmpeg and sharp. Running with host process privileges, the plugin provides features such as asynchronous task tracking, multi-provider model pools, and prompt optimization, aiming to lower the barrier to implementing multimodal tasks in agent development.
Read More@blue/dsh-image-model-router
`@blue/dsh-image-model-router` is a DeepSeek Harness plugin that automatically routes requests to the appropriate model based on input content (images or plain text). It intercepts requests to detect images, sends image-containing requests to a vision model (default: `deepseek-v4-flash-vision-exp`), and keeps plain-text requests on the original model, thereby avoiding resource waste and invocation failures. After installation, it must be manually configured and enabled in `cordis.patch.yml`, and it supports image inputs returned by tools.
Read Moredsh-omni-workstation: DeepSeek Harness Omni-Modal Workstation Plugin
`dsh-omni-workstation` is an omni-modal workstation plugin for DeepSeek Harness, designed to simplify the integration of visual, auditory, and video creation capabilities. It provides image analysis, generation, video synthesis, and speech processing through a unified configuration entry point and supports the ComfyUI protocol. The plugin features a flexible modular design that allows individual features to be enabled or disabled to reduce costs. It also supports asynchronous multi-GPU processing and multiple protocols, significantly reducing developers' effort in maintaining Skill prompts.
Read MoreDeepSeek Visual Plugin
The DeepSeek Harness plugin intercepts messages containing images and converts images that pure text models cannot process into textual descriptions generated by a vision model, enabling it to interpret images. It allows multimodal models to pass through directly, while pure text models automatically invoke a vision model to translate them. Users must deploy the code to `$DSH_HOME/profiles/node_modules` and configure the model and API key in `cordis.patch.yml`. The plugin covers session logs and tool execution, supports deployment on Windows, Linux, and macOS, and includes a self-check script.
Read Moredsh-official-vision: DSH Plugin for Direct Access to DeepSeek Official Vision API
`dsh-official-vision` is a DeepSeek Harness (DSH) plugin designed to connect directly to the official DeepSeek vision API without relying on third-party models. The plugin registers the `deepseek-v4-flash-vision-exp` route, supports formats such as JPEG and PNG, provides two transmission methods—Base64 inline and the Files API—and automatically passes through plain-text requests. It addresses the intermediate model dependency issue in existing vision plugins, allowing users to call DeepSeek’s multimodal capabilities directly and configure the API key, image detail level, and large-image handling options.
Read Morellm-deepseek-image-admit: DeepSeek Image Admission
DeepSeek's official model metadata restricts image admission. This plugin takes over the official routing and modifies the metadata—completing `inputModalities` to `['text', 'image']`—to bypass the harness's image blocking and allow image messages to pass admission pre-checks. It relies on the official artifact layout, transmits no images via the API, and must be used with plugins such as `image-bridge` to transcribe image content during the `pre-step` stage, without changing user behavior. The installation command is `dsh plugin --profile web add github:Seryta/llm-deepseek-image-admit`.
Read Moredsh-gemini-multimodal: Multimodal Perception for DeepSeek Harness
**dsh-gemini-multimodal** is a plugin developed for DeepSeek Harness. By calling the Google Gemini API or a local CLI, it addresses the limitation that DeepSeek models cannot directly process unstructured data such as screenshots, audio, and video. The plugin provides capabilities including multimodal analysis, transcription, image generation, and document understanding, and supports two provider modes: `gemini_api` and `antigravity_cli`. After installation, users can invoke the relevant tools in DeepSeek Harness, allowing the main model to focus on logical reasoning while enabling “seeing” and “hearing” capabilities.
Read MoreDeepSeek Harness (DSH) Multimodal Vision Bridge Plugin
The DSH (DeepSeek Harness) plugin `dsh-vision-bridge` is designed to provide visual capabilities to text-only models that do not natively support images, such as DeepSeek. The plugin proxies `llm/stream` requests, converts images into structured Chinese descriptions that are injected into the context, and provides two tools: `view_image` and `ocr_image`. It supports native multimodal routing detection to optimize performance and allows hot-updating configuration from the settings page. After installation, users can upload images directly in a conversation; the plugin handles them automatically or via tool calls, without requiring any modifications to the DSH core source code.
Read MoreDeepSeek Harness Image Multimodal Description Plugin (`sd1g1/dsh-image-describe`)
`sd1g1/dsh-image-describe` is a DeepSeek Harness host plugin designed to address the limitation that text-only models cannot process images directly. By injecting a `describe_image` tool, it enables the primary model to "see" and describe images. Key features include support for local file paths and conversation attachments, automatic scaling, and automatic configuration. The plugin requires `vision.provider` and `vision.model` to be configured to become active and supports formats such as PNG and JPEG. Once installed, the primary model can perform multimodal image description through tool calls, extending the application boundaries of text-only models.
Read MoreDeepSeek Harness Plugin: dsh-codex-media
dsh-codex-media is a DeepSeek Harness plugin designed to use a locally signed-in Codex CLI with a ChatGPT Plus subscription to handle image generation and editing tasks, thereby saving DeepSeek tokens. It registers four core tools via the MCP protocol: image generation, image editing (with multiple reference image inputs), clipboard-to-file saving (requires PowerShell), and the Codex subagent. The plugin supports installation via the command line; you need to add the MCP server settings to the configuration file and restart the service. Generated images are saved locally by default, and the feature consumes your ChatGPT subscription quota.
Read MoreDeepSeek Harness Plugin dsh-subagent-vision: Enabling Pure-Text Main Agents to Read Images Directly in Conversations
The DeepSeek Harness plugin `dsh-subagent-vision` addresses the inability of pure-text models to handle images. It delegates image processing tasks to a configured vision subagent, enabling the main agent to remain text-only while receiving visual results. The plugin supports pasting, dragging and dropping, and importing images via file path; provides a native Web GUI experience; and enforces a 25 MB image size limit. After installation, users only need to describe the task to the agent, and the plugin automatically invokes the subagent's `read_image` tool to complete it.
Read Moredsh-autovision: Adding Visual Capabilities to DSH Text Models
The `dsh-autovision` plugin provides native visual capabilities for DeepSeek Harness (DSH) text-only models. By registering a transparent twin provider, it automatically routes image requests to a configured multimodal model for transcription and then injects the resulting text into the context, enabling “paste-to-text.” Users do not need to switch models or expose API keys, and can paste images directly or read files through Agent tools. The plugin is easy to install and offers persistent configuration and an LRU cache to mitigate latency, but users should be aware of limitations such as the first image being skipped, a 5 MB file size limit, and minor UI display issues.
Read Moredsh-image-bridge: Enabling Image Input Support for Text Models in DeepSeek Harness
`dsh-image-bridge` is a DeepSeek Harness (DSH) plugin designed to address the limitation that text-only models in DSH cannot process images. It wraps interfaces and leverages DSH’s surface replace mechanism to allow text models to receive images, save them to disk, and convert them into a text format for the model to invoke vision or MCP tools. The plugin supports image thumbnail display and is easy to install, but users should note limitations when mixing commands with images and when handling previous images.
Read Moreauto-vision: DSH All-in-One Image Plugin
`auto-vision` is a DSH plugin designed to resolve errors or rendering issues caused by directly pasted images. It automatically cleans image data and registers the `see_image` tool, allowing text models to invoke vision models to read content automatically without modifying the original logic. The plugin supports free vision models from Zhipu and ModelScope, and is also compatible with any OpenAI-compatible platform. During installation, the plugin directory must be copied into `node_modules`, and its configuration must be added to `cordis.patch.yml`. Users can configure the API token via environment variables or a credentials file. By default, it uses ModelScope, but it can be switched to Zhipu or a local model. The DSH process must be restarted before use.
Read MoreDeepSeek Harness Vision Bridge Plugin: analyze-image-tool
`analyze-image-tool` is a general-purpose visual interface plugin that provides a “visual bridging” layer for DeepSeek Harness text models. It supports any OpenAI-compatible endpoint, allowing multimodal models to analyze images through the `analyze_image` tool call. Core features include universal endpoint configuration, sandboxed secure reads, schema-safe compilation, automatic sanitization of API keys and reasoning traces, structured return payloads, and a WebUI settings panel. Installation requires Node.js ≥ 18.17 and DSH 0.1.x. Note that the image-pasting feature requires installing the `main` branch, and historical sessions cannot read previous attachments after a restart.
Read Moredsh-modlens: A DSH Plugin for Adding Visual Capabilities to Text-Only LLMs
The dsh-modlens plugin for DeepSeek Harness (DSH) is designed to add vision capabilities to text-only LLMs by converting images into structured JSON evidence. It supports multiple engines (such as Gemini, Doubao/Ark, and Claude). Users can configure the API key and model in a configuration file or specify the engine via parameters at invocation. Once installed and enabled, the model gains the `(modlens vision)` capability and the `modlens_read_image` tool, with support for local diagnostic verification.
Read Moredsh-bridge-preview: In-Chat Inline Preview for dsh-tool-vision Image Bridging
`dsh-bridge-preview` is a client-side plugin designed to address the issue that DeepSeek Harness text-only models cannot display images. It leverages MutationObserver to monitor the DOM and automatically converts the image path text generated by `dsh-tool-vision` into a visual inline preview within chat bubbles. The plugin uses a pure additive architecture and does not modify underlying data. It also supports clicking a thumbnail to view the full-size image and includes automatic fallback handling. Users must first install and enable `dsh-tool-vision`, then install this plugin via a terminal command and restart for it to be ready to use.
Read Moremimo-vision: DSH Native Vision Plugin
mimo-vision is a native plugin designed for DeepSeek Harness (DSH), providing text description capabilities for host models that lack native vision support. As a first-class citizen of DSH, it registers the `describe_image` tool and uses DSH infrastructure to manage dependencies and credentials, sending images to a multimodal model and parsing the resulting descriptions. The plugin supports routing to free or paid APIs, preferring free APIs by default, and can be installed with the command `dsh plugin --profile web add mimo-vision`. It requires Node.js `^22.19` or `>=24`, and an OpenCode API Key must be configured. When used, the host model can invoke the tool to generate text descriptions for images at specified paths.
Read MoreDeepSeek VisionPlus: Text goes to DeepSeek, images automatically routed to free vision model pools
When using DeepSeek Harness (DSH), there is an unavoidable limitation: DeepSeek's official models are text-only and cannot process images. When screenshots, charts, or interface photos appear in a message, one must either manually transcribe the content or switch the entire conversation to another multimodal model, resulting in a break in context. DeepSeek VisionPlus takes a different approach: text requests continue to be handled by the DeepSeek official API, while images are automatically routed to a pool of free visual models, with the understanding results returned to the same conversation. DSH's philosophy is that "everything is a plugin," and this capability is implemented as a plugin. Below is an introduction to its features, installation, and daily usage.
Read Moredsh-vision-plugin: Add an "Image Input" declaration checkbox to the DSH model directory
# Introduction When integrating custom models into DSH, you may encounter a scenario where the endpoint itself supports multimodal capabilities, but sending an image directly in the chat dialog results in an error stating "Image input not supported." The reason is that the DSH host reads the `input: ["text","image"]` declaration on the model entry to decide whether to pass the original image directly to the model, while the interface in "Settings → Models → Model Catalog" has never exposed an entry point for this declaration. dsh-vision-plugin aims to address this step: by adding an "Image Input" checkbox on the model page in the Model Catalog. Once checked, the declaration is written into DSH settings, allowing the custom model to directly receive images in the chat dialog. # What is it dsh-vision-p
Read MoreDSH-Vision: A Plugin Adding Image Recognition Capabilities to DeepSeek Harness
dsh-vision is a DeepSeek Harness (DSH) bundle plugin maintained by Terry12138qy, licensed under the MIT License. This plugin is designed to address the issue that the DSH main model lacks native image recognition capabilities. By encapsulating the image recognition workflow, it sends local image paths or network URLs to Alibaba Cloud Bailian or Zhipu multimodal models, ultimately returning Chinese text descriptions for use by agents or Code Mode. Core features include automatically reading and encoding images, prioritizing the main model `qwen3.5-omni-plus`, and automatically switching to the fallback model `glm-4.6v-flash` for retries upon failures (e.g., quota exhaustion, rate limiting). It supports invocation via `await tools.vision(...)`.
Read Moredsh-video-lens: Equipping Pure Text DSH Agents with Video Understanding Capabilities
dsh-video-lens is a DeepSeek Harness plugin maintained by dundunhan (v0.3.1, MIT). It utilizes three tools—video_probe, video_analyze, and video_ask—to read local video metadata via ffprobe/ffmpeg, perform scene-aware frame extraction with optional ASR, and interface with OpenAI-compatible vision models, outputting structured evidence JSON with timelines. Installation requires running `pnpm add dsh-video-lens` in the DSH Profile directory and registering the bundle, along with configuring environment variables such as `VIDEO_LENS_API_KEY`. It is suitable for video review, meeting recording summaries, and time-anchored Q&A; users are advised to review the source code and permission notes before use.
Read More