Deepseek V4 Flash Vision Video Rag

Community

liangdabiao

deepseek-v4-flash-vision-video-rag

基于DeepSeek视觉大模型（deepseek-v4-flash-vision-exp）的视频理解与问答（video RAG）。当用户提到视频、录像、mp4、监控、游戏录像、影片、动画，并想提问、查找、定位、总结或理解其内容，或问"什么时候发生的""某一秒在干什么"、要时间戳定位、片段回放时，使用本skill。支持剧情理解、事件时间线、OCR字幕/HUD识别；回答带 [MM:SS] 时间戳引用，并生成可播放片段与关键帧展示给用户。无音频分析能力。

Overview

Publisher	liangdabiao
Repository	`deepseek-v4-flash-vision-video-rag`
Skill name	`deepseek-v4-flash-vision-video-rag`
Stars	73
Forks	5
Bundled files	6
Links	Skill source Repository Homepage

Markdown instructions
A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.
Works with any LLM
AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.
6 bundled files
Scripts, templates, and references the model can read while it works. Files are read-only and never executed.
Open source
Published by liangdabiao on GitHub. Read the source before you install it.

Installation

Install the Deepseek V4 Flash Vision Video Rag AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

Open the app and go to Plugins → Skills.
Choose "Install from GitHub".
Paste the skill folder URL below and confirm.
https://github.com/liangdabiao/deepseek-v4-flash-vision-video-rag/tree/main
Enable the skill in any chat where you want it available.

Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills

git clone --depth 1 https://github.com/liangdabiao/deepseek-v4-flash-vision-video-rag.git \
  .claude/skills/deepseek-v4-flash-vision-video-rag

Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Deepseek V4 Flash Vision Video Rag in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Deepseek V4 Flash Vision Video Rag on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Deepseek V4 Flash Vision Video Rag is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

Loaded only when it is needed
The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.
Switch models mid-chat
Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

DeepSeek V4-Flash Vision Video RAG

让视觉大模型"看懂"一段视频：先把视频按时间轴抽帧阅读、建立索引（一次性、缓存复用），再对用户问题做本地粗筛 → 视觉精排 → 深读回答，回答带 [MM:SS] 时间戳引用，并自动切出可播放片段、关键帧和自包含 HTML 预览。

模型没有音频模态——所有理解基于画面；音频轨会被忽略（如实告知用户）。

环境要求

Python 依赖：openai（本机已装）
ffmpeg / ffprobe 必须在 PATH（本机已装；缺了先安装）
API key：从环境变量 DEEPSEEK_API_KEY 或本机 ~/.deepseek_api_key 文件（首行）读取，代码中不内置任何密钥；缺失时报错并给出设置指引
脚本在 scripts/ 下，Windows 直接 python scripts/xxx.py

工作流

第 1 步：建立索引（每份视频一次，之后走缓存）

bash
python scripts/ingest.py "<视频路径>"

ffprobe 元数据 → 宏观抽帧（≤60s→1fps，≤10min→0.5fps，≤30min→0.25fps，更长→0.1fps， --fps 可覆盖）→ Files API 上传 → 25 帧/批视觉卡片 → 段聚合（30s/段）→ 本地场景切换检测
30s 视频约 30 秒完成；重复运行直接命中缓存
常用参数：--force 重建｜--limit-frames N 先试前 N 帧｜--clean 清缓存
输出打印帧卡片数/段数/场景切换/运动统计，可据此向用户简报视频概况

第 2 步：回答用户问题

bash
python scripts/ask.py "<视频路径>" "用户的问题"

本地粗筛（TF-IDF，零费用）→ 视觉精排选 top3 段（1 次调用，关推理）→ 深读选中段全部帧（1 次调用，开推理，max_tokens=32768）→ 带 [MM:SS] 引用的回答
微观层自动触发：问题含动作类词（瞄准/击杀/瞬间/轨迹…）且深读有引用时，对运动最强的引用时刻抽 12fps 中心放大帧追加一次分析（+1 次调用）； --micro 00:04-00:07 强制指定窗口，--no-micro 禁用

stdout 分段：=== 答案 ===、=== 引用时刻 ===、=== 关键帧 ===、 === 可播放片段 ===（均为 ./video-vision-out/ 下文件）、=== 预览 ===。

第 3 步：向用户展示（必须做）

终端里用户未必能看到 inline 图片/视频，必须给出可自行打开的路径，按优先级：

HTML 预览（首选）：./video-vision-out/<视频名>_<问题>.html —— 自包含单文件，答案 + 引用时刻的可播放片段（base64 内嵌 <video>）+ 关键帧，双击浏览器打开。 Windows 下可代用户执行 start "<路径>"
单独的片段 mp4 / 关键帧 jpg 路径：用户只要某个时刻时逐个给出
支持 inline 的环境可再 Read 图片作补充，但不得只依赖它

用户想直接看某个时刻/片段：

bash
python scripts/show.py "<视频路径>" --at 00:07          # 该时刻帧
python scripts/show.py "<视频路径>" --clip 00:05-00:12  # 可播放片段
python scripts/show.py "<视频路径>" --micro 00:04       # 12fps 中心放大帧序列

回答规范

以脚本产出的答案为事实基础，保留 [MM:SS] 时间戳引用标记
时刻引用是 1fps 采样的估计值（误差 ±2s），涉及精确时机时提醒用户以片段回放为准
答案说"帧内容不足"时如实转告，不要编造；音频相关请求明确说明无音频能力
展示时必须给 HTML 预览或媒体文件路径，不能只靠 inline 展示

注意事项

首次 ingest 慢是正常的；之后每次提问 2~3 次调用
深读/微观用 max_tokens=32768（开推理时 reasoning 计入预算，小了会得到空回复）
视频抽帧与采样依据实测最佳实践（GIF 不可用、两级采样、detail=low 禁用等），详见 references/video-sampling.md；索引结构见 references/index-schema.md
改 scripts/ds_client.py 前先读 references/video-sampling.md 的踩坑清单

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Deepseek V4 Flash Vision Video Rag AI skill do?

Why use Deepseek V4 Flash Vision Video Rag on TypingMind?

Because you install it once and use it with any model. Deepseek V4 Flash Vision Video Rag is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Deepseek V4 Flash Vision Video Rag in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/liangdabiao/deepseek-v4-flash-vision-video-rag/tree/main. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Deepseek V4 Flash Vision Video Rag?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Deepseek V4 Flash Vision Video Rag?

As many as you like. As long as a model supports skills, you can use Deepseek V4 Flash Vision Video Rag with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Deepseek V4 Flash Vision Video Rag AI skill free?

It is published on GitHub by liangdabiao. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Nemo Curator

NousResearch

OrganizationPopular

Curate LLM training data: dedupe, filter, PII redaction.

Obliteratus

OBLITERATUS: abliterate LLM refusals (diff-in-means).

Prompt Lookup

Activates when the user asks about AI prompts, needs prompt templates, wants to search for prompts, or mentions prompts.chat. Use for discovering, retrieving, and improving prompts.

Banner Design

Design banners for social media, ads, website heroes, creative assets, and print. Multiple art direction options with optional generated or supplied visuals. Actions: design, create, generate banner.…

Overview

Installation

Install in TypingMind

Install in another agent

Use it in TypingMind

Works with any AI model

Skill instructions

DeepSeek V4-Flash Vision Video RAG

环境要求

工作流

第 1 步：建立索引（每份视频一次，之后走缓存）

第 2 步：回答用户问题

第 3 步：向用户展示（必须做）

回答规范

注意事项

Bundled files

Frequently asked questions

Related AI Skills

Set up your own AI workspace now