Anionex/agent-vision-toolkit

为纯文本模型"看图“设计更好的视觉工具箱和技能，支持多图理解，图片问答，前端UI还原、GUI 自动化等，并可选无缝接入多个主流agent，直接识别粘贴图片｜ A vision toolkit and skill designed for text-only llms — image Q&A, long-screenshot OCR, frontend UI restoration, and GUI automation, with optional seamless integration for Codex, Claude Code, Pi, Oh My Pi, and OpenCode

Visit GitHub

Python

1.3k

1 contributors

MIT License

website

FeaturedAdvertise to 180K+ monthly visitors

Busbar

Control what AI can do before it acts

Bifrost

Fastest Enterprise AI Gateway for Routing, Governing, and Securing all your AI Traffic

Omnigraph

Versioned graph database on object storage. Git-style workflows and multimodal retrieval

Websitebench

Can AI agents reverse-engineer real websites?

Momentum

How fast Anionex/agent-vision-toolkit is gaining stars, and where it lands on Trendshift and GitHub Trending

Activity and trending

Stars, forks and merged pull requests each day or month, lined up with the repository's rank on Trendshift Trending

Sep 11 – Nov 10, 2026

Stars +77 in 61 days

Forks +6 in 61 days

Merged PRs 2 in 61 days

Trendshift Trendingwhere the repository ranked each day it made the top 25

Not in the Trendshift top 25 in this period

Sep 11Sep 19Sep 27Oct 5Oct 13Oct 21Oct 29Nov 6

Stars, forks, merged pull requests and Trendshift Trending rank, by day
Day	Stars gained	Forks gained	Merged PRs
Fri, Sep 11, 2026	+3	+1	0
Sat, Sep 12, 2026	+1	+0	0
Sun, Sep 13, 2026	+1	+0	0
Mon, Sep 14, 2026	+4	+0	0
Tue, Sep 15, 2026	+5	+0	0
Wed, Sep 16, 2026	+1	+0	0
Thu, Sep 17, 2026	+0	+0	0
Fri, Sep 18, 2026	+0	+1	0
Sat, Sep 19, 2026	+1	+0	0
Sun, Sep 20, 2026	+0	+0	0
Mon, Sep 21, 2026	+3	+0	0
Tue, Sep 22, 2026	+0	+0	0
Wed, Sep 23, 2026	+2	+0	0
Thu, Sep 24, 2026	+1	+0	0
Fri, Sep 25, 2026	+2	+0	0
Sat, Sep 26, 2026	+2	+1	0
Sun, Sep 27, 2026	+0	+0	0
Mon, Sep 28, 2026	+3	+0	0
Tue, Sep 29, 2026	+0	+0	0
Wed, Sep 30, 2026	+1	+0	0
Thu, Oct 1, 2026	+0	+0	0
Fri, Oct 2, 2026	+0	+0	0
Sat, Oct 3, 2026	+0	+0	0
Sun, Oct 4, 2026	+0	+0	0
Mon, Oct 5, 2026	+0	+1	0
Tue, Oct 6, 2026	+3	+0	0
Wed, Oct 7, 2026	+1	+0	0
Thu, Oct 8, 2026	+1	+0	1
Fri, Oct 9, 2026	+2	+0	0
Sat, Oct 10, 2026	+3	+0	0
Sun, Oct 11, 2026	+3	+0	0
Mon, Oct 12, 2026	+1	+0	0
Tue, Oct 13, 2026	+4	+1	1
Wed, Oct 14, 2026	+1	+0	0
Thu, Oct 15, 2026	+1	+0	0
Fri, Oct 16, 2026	+1	+0	0
Sat, Oct 17, 2026	+0	+0	0
Sun, Oct 18, 2026	+0	+0	0
Mon, Oct 19, 2026	+3	+0	0
Tue, Oct 20, 2026	+1	+0	0
Wed, Oct 21, 2026	+3	+0	0
Thu, Oct 22, 2026	+4	+0	0
Fri, Oct 23, 2026	+3	+0	0
Sat, Oct 24, 2026	+1	+0	0
Sun, Oct 25, 2026	+0	+0	0
Mon, Oct 26, 2026	+1	+0	0
Tue, Oct 27, 2026	+0	+0	0
Wed, Oct 28, 2026	+1	+0	0
Thu, Oct 29, 2026	+0	+0	0
Fri, Oct 30, 2026	+1	+0	0
Sat, Oct 31, 2026	+1	+0	0
Sun, Nov 1, 2026	+1	+0	0
Mon, Nov 2, 2026	+0	+0	0
Tue, Nov 3, 2026	+0	+0	0
Wed, Nov 4, 2026	+0	+0	0
Thu, Nov 5, 2026	+3	+0	0
Fri, Nov 6, 2026	+3	+0	0
Sat, Nov 7, 2026	+0	+1	0
Sun, Nov 8, 2026	+0	+0	0
Mon, Nov 9, 2026	+0	+0	0
Tue, Nov 10, 2026	+0	+0	0
Total	+77	+6	2

#1 on the list
Rank that day
Not trending

Mentions and achievements

Where people are talking about Anionex/agent-vision-toolkit, and every badge it has earned

Aug 191 post
- 开源DSH Desktop 目前只有window客户端，有以下几个特点： 1，原生 Electron 窗口，非命令行 2，增加手动目录选择 3，增加右侧边栏工具，接入的是Better Sidebar（ 4，增加了视觉识别，接入的是Vision Toolkit （ 5，增加了底部命令行。但是失灵时不灵。还在排查稳定性。开源地址：
  Read full on X, posted by @renqw5271
  @renqw5271 · x.com
Aug 171 post
- 给跑在纯文本模型（比如 DeepSeek）上的 coding agent 配一双眼睛：一套 CLI 工具加 skill，再加本地透明代理，让只能读文字的模型也能看图、问答、OCR、还原 UI、操作 GUI。
  Read full on X, posted by @QingQ77
  @QingQ77 · x.com
Aug 161 post
- dsh-vision-toolkit star刚破了500 🌟 ～感谢大家的支持这应该是dsh生态第一个视觉工具集插件（内测时开始开发），专为纯文本模型进行进阶视觉任务而设计，包含工具集和对应的skill 不仅仅是粘贴图片-得到图片描述，该项目还能够： 1. 一行命令安装，无需配置api key，先免费用起来（内置免费额度） 2. 根据截图重建UI为html（如图） 3.…
  Read full on X, posted by @anion_ex
  @anion_ex · x.com
Aug 141 post
- 梁圣的 DeepSeek 重磅发布 Harness：最大特色：一切皆插件！开发者不需要改源码，只通过配置就能组合AI能力，现在社区插件排名： 1. colleague-skill 同事 Skill 将冰冷的离别化为温暖的 Skill。提供飞书/钉钉/Slack 消息、文档、邮件、截图 + 主观描述，生成真正能替该同事工作的 AI Skill，支持数字生命传承。 2. iPolloWork…
  Read full on X, posted by @web3annie
  @web3annie · x.com
Aug 131 post
- DSH 里的 Agent，不是出生时能力就定死了。它的架构允许自己在任务跑到一半的时候，发现“现在的我不够用了”，然后给自己挂新的工具、Skill、Service，甚至换一套“身份”继续干活。比如前半段还是一个只读的调研 Agent，后面需要落地了，就动态切换自己的skill集、工具集，变成一个开发 Agent；也就是说，Agent 不只是“在 Harness…
  Read full on X, posted by @anion_ex
  @anion_ex · x.com
Aug 82 posts
- 🔥 Top 10 GitHub Projects This Week (Aug 9) — for AI devs & founders From a CPU-only 2.7T-param LLM to an agent-first CRM and an AI-native office suite. These 10 open-source repos exploded this week ↓
  Read full on X, posted by @FrankLeeLife
  @FrankLeeLife · x.com
- 别急着给纯文本模型换多模态底座，视觉能力也可以放进 agent 的工具层它把看图拆成描述、定位、检测、裁剪、轨迹恢复等工具，再用任务意图作为视觉模型的关注提示，减少泛泛描述造成的信息损失。还提供长截图 OCR、UI 还原、GUI 操作流程，并能接入 Codex、Claude Code 等。代价是仍需一个兼容接口的视觉模型，最终效果也取决于两边模型。
  Read full on X, posted by @vintcessun
  @vintcessun · x.com
Aug 53 posts
- Give text-only LLM agents vision: img Q&A, OCR, screenshot analysis. See link below 👇
  Read full on X, posted by @RepoGems
  @RepoGems · x.com
- 限制说清楚： • 图片转文字不是视觉token，信息有损 • 描述质量取决于视觉模型，垃圾模型进垃圾描述出 • 缓存进程内，重启清空 • 需要额外的视觉API，每次看图消耗token • 已经用GPT-4o/Claude 3.5的不需要这层但对用DeepSeek V4做Agent开发的人，这东西填补了一个空白。
  Read full on X, posted by @BencyMarrit
  @BencyMarrit · x.com
- 大表哥报。DeepSeek V4 在 Codex 里看不了图？有人4天前刚把这坑填了。本地代理截获图片请求→视觉模型转文字→纯文本模型收到描述。Agent不用改，MCP不用加。
  Read full on X, posted by @BencyMarrit
  @BencyMarrit · x.com

Anionex/dsh-vision-toolkit

TypeScript

[dsh]为纯文本模型设计更强大的视觉工具箱：一行安装使用、粘贴图片直接识别、多张图片问答、截图到前端UI 还原等｜DeepSeek Harness-native integration for agent-vision-toolkit: image Q&A, long-screenshot OCR, UI restoration, grounding, pixel diff, Artifacts, and Web UI.

93757

RunhuaHuang/VisionPower

JavaScript

VisionPower: portable image understanding MCP server for Codex, Claude Desktop, Cursor, and other agents.

688

inclusionAI/UI-Venus

HTML

UI-Venus is a general-purpose foundation GUI agent for mobile apps, web platforms, and desktop operating systems using only screenshots as input.

1.1k92

JochenYang/luma-mcp

TypeScript

多模型视觉理解 MCP 服务器，为不支持图片理解的 AI 编码模型提供视觉能力：分析截图、报错、UI 与文档，可接入多家主流视觉大模型。Multi-model vision MCP server that adds image understanding to AI coding models without native vision — analyze screenshots, errors, UI and documents via major vision LLM providers.

11513

ysr666/dsh-vision-router

JavaScript

Eyes for text-only DeepSeek Harness agents: built-in free vision chain (no key) + pixel-level vision tools (Q&A, grounding, crop, pixel diff, colors, OCR, SVG trace, cutout, screenshots). One-command install, no Python, image turns work like ordinary tool-calling turns.

1.3k71

xiincs/claude-code-vision-skill

Python

为 Claude Code 赋能多模态视觉能力，适配纯文本 LLM 底座，用于截图 / UI / 图表分析；搭配 browser-harness 可做前端布局自动化检查。

1689

zai-org/CogAgent

Python

An open-sourced end-to-end VLM-based GUI Agent

1.2k109

liustack/modlens

TypeScript

The first vision plugin for DeepSeek Harness, and the vision bridge for every text-only coding agent. Paste an image, get structured JSON evidence (OCR, layout, semantics). | 全网最强 DeepSeek Harness 外挂视觉插件，为 DeepSeek、GLM 等纯文本模型外挂视觉能力，粘贴图片即得结构化 JSON 证据（OCR、版面、语义）。

4.9k163

Sorwcyra/ds-vision-skill

PowerShell

ds-vision-skill: Vision skill suite to add multimodal image capabilities for DeepSeek, and other text-only models as well.

15811

Anionex/agent-vision-toolkit

Momentum

Activity and trending

Mentions and achievements

You might also like

You might also like

Anionex/dsh-vision-toolkit

RunhuaHuang/VisionPower

inclusionAI/UI-Venus

JochenYang/luma-mcp

ysr666/dsh-vision-router

xiincs/claude-code-vision-skill

zai-org/CogAgent

liustack/modlens

Sorwcyra/ds-vision-skill