xlight/deepseek-visionary 预览 preview

xlight/deepseek-visionary

[插件名称]让您的Agent拥有视觉能力,基于DeepSeek官方多模态模型(兼容DSH、Zed、OpenCode、Codex、Claude Code、Cursor、Claude Desktop)

Project Overview项目介绍

This repository hosts an MCP service that exposes DeepSeek’s web-based vision model to a wide range of AI agents, including DeepSeek Harness (DSH), Zed, OpenCode, Codex, Claude Code, Cursor, and Claude Desktop. It is a full rewrite in Rust of the original Python version deepseek-vision-mcp, packaged as a single native binary that works across multiple platforms, so you only need to install it once to use it across all your AI agents. For DSH users, the service can be installed with a single dsh plugin add @xlight-oss/visionary-dsh command, which includes five pre-built tools and image bridging capability for text models.

The standard workflow starts with installing the binary via one of several methods: you can use the one-line installer script for macOS/Linux, the PowerShell script for Windows, install it via Homebrew, pull the npm global package, or manually download the binary from GitHub Releases and add it to your system PATH. After installation, you run visionary-server login to trigger an automatic browser login that fetches your DeepSeek credentials, so you don’t need to manually copy API keys or tokens to start using the service. It is built for developers and general users who want to add image recognition and OCR capabilities to their existing AI agents, and it supports multi-turn conversations with persistent session state for follow-up analysis and multi-image comparison.

The tool officially supports macOS (both Apple Silicon and Intel architectures), Linux (x86_64 and aarch64), and Windows (x86_64), and requires that you have at least one Chrome-based browser (Chrome, Chromium, or Edge) installed for the automatic login step. OCR results work best on clear screenshots and documents; blurry images, handwritten text, or complex layouts may return incomplete results. The project is released under the open-source MIT license, and first-time Windows users may see a SmartScreen security prompt that requires manual approval before the first run can complete.

本项目是一个提供 DeepSeek 网页版视觉模型能力的工具,可集成到任意支持 MCP 的 AI 助手,包括 Zed、OpenCode、Codex、Claude Code、Cursor、Claude Desktop 以及 DeepSeek Harness(DSH)。它是Python版 deepseek-vision-mcp 的Rust全量重写,单原生二进制多平台分发,一次安装就能在多个代理中使用。DSH 用户可通过 dsh plugin 一键安装原生插件包,包含5个原生工具和文本模型图片桥接功能。

典型使用流程为,先通过一键脚本、Homebrew、npm 或手动下载二进制完成安装,首次运行调用 visionary-server login 触发浏览器自动登录获取凭据,无需手动复制 API key 或 token,操作十分便捷。它面向需要让现有 AI 助手获得图像识别、OCR 文字提取能力的开发者和普通用户,支持多轮对话,可保存会话状态方便后续续聊和多图对比分析。

本项目支持 macOS(Apple Silicon/Intel)、Linux(x86_64/aarch64)和 Windows(x86_64)平台,运行依赖系统中已安装 Chrome/Chromium/Edge 其中一款浏览器用于自动登录。OCR 对清晰截图和文档识别效果较好,模糊、手写或复杂版式可能结果不完整。项目采用 MIT 许可证开放源代码,首次运行 Windows 可能会遇到 SmartScreen 提示,手动放行即可。

Pre-install check安装前体检Compatibility · Security兼容性 · 安全性 3 warnings3 项注意
  • No license declared - all rights reserved by default; ask the author before commercial use or redistribution未声明开源许可证 —— 默认「保留所有权利」,商用或再分发前先问作者
  • 18 stars - an early-stage project星标 18,属于早期项目
  • No DSH plugin manifest detected - it may only carry the dsh-plugin topic, so the install method must be confirmed on the spot未检测到 DSH 插件清单:可能只是打了 dsh-plugin 话题,安装方式要现场确认
  • Not DSH-native: a multi-platform tool that may require Node / Electron or another runtime first非 DSH 原生,是多平台兼容工具:可能要先装 Node / Electron 等运行时
  • Desktop client: installation downloads an executable - verify the publisher and checksums桌面客户端:安装会下载可执行文件,请核对发布者与校验和
DSH walks through these 9 checksDSH 会逐条核对这 9 项

Compatibility兼容性

  • DSH, Node, OS and profile requirementsDSH 版本 / Node 版本 / 操作系统 / profile 是否满足要求
  • External dependencies and runtimes (Electron / Python / Docker, ...)外部依赖与运行时(Electron / Python / Docker 等)是否齐备
  • Conflicts with installed plugins: command names, skill / tool names, ports, duplicate MCP registration与已装插件是否冲突:命令名、skill / tool 重名、端口占用、重复 MCP 注册

Security安全性

  • Repo matches the facts registered here; archived or abandoned?仓库是否与页面登记一致,是否归档或长期停更
  • Safety of preinstall / install / postinstall and install.sh / setup.ps1preinstall / install / postinstall 与 install.sh、setup.ps1 是否安全
  • curl|bash, download-then-execute, obfuscation, unrelated domains → stop immediatelycurl|bash、下载即执行、混淆代码、无关域名 → 立刻停止
  • Typosquatting or unmaintained packages among the new dependencies新增依赖里有没有 typosquatting 或无人维护的包
  • Requested permissions vs. what the feature actually needs申请了哪些权限、是否超出功能所需(filesystem / network / shell / clipboard)
  • Any sudo / admin requirement, plus uninstall and rollback是否要求 sudo / 管理员权限,以及卸载与回滚方式

Anything uncertain must be marked unknown with a note on how to confirm it. This site's signal screen is a static snapshot, not a security audit.拿不准的必须标「未知」并说明要我怎么确认。本站的信号筛查是静态快照,不能替代安全审计。

Or use CLI install (for developers)或使用命令行安装(适合开发者)

CLI Install命令行安装

dsh plugin --profile web add @xlight-oss/visionary-dsh

把 xlight/deepseek-visionary 加入你的 DSH 配置(web profile)即可启用。

READMEREADME

img

DeepSeek Visionary

让 DeepSeek 网页版视觉模型,成为你所有 AI 助手的"眼睛"。 在 Zed、OpenCode、Codex、Claude Code、Cursor、Claude Desktop 等任意支持 MCP 的 agent,以及 DeepSeek Harness(DSH,原生插件或 skill + CLI)中直接识图——浏览器自动登录,无需 API key、无需手动复制 token。

Python 版 deepseek-vision-mcp 的 Rust 全量重写:单原生二进制、多平台分发,一处安装处处可用。DSH 用户更可 dsh plugin 一键安装原生插件包 @xlight-oss/visionary-dsh——5 个原生工具 + 文本模型图片桥接,一包全齐。

架构

graph TD
    subgraph 宿主[任意 MCP 宿主]
        AG["Zed / OpenCode / Codex / Claude Code / Cursor / Claude Desktop"]
        AG -->|spawn 独立进程| SRV
    end
    subgraph DSH[DeepSeek Harness]
        DP["@xlight-oss/visionary-dsh 插件<br/>deepseek_vision 等 5 个原生工具<br/>+ 文本模型图片桥接"]
        DP -->|宿主进程 spawn| SRV
    end
    subgraph visionary-server 原生二进制
        SRV["CLI + MCP stdio 服务<br/>vision / ocr / status / login / logout / skill / init / doctor<br/>mcp-stdio CLI"]
        CFG["~/.deepseek-visionary/config.json<br/>token + smidV2 + cf_clearance + 会话"]
        SRV --> CFG
    end
    SRV -->|HTTPS| DS["DeepSeek 网页后端"]
    SRV -->|CDP 启动 + 监听| BRO["Chrome 系浏览器<br/>仅登录时出现"]
  • visionary-server:单二进制,默认 CLI 模式(vision / ocr / status / login / logout / skill / init / doctor),mcp-stdio 子命令显式启动 MCP stdio 服务;实现完整 vision / OCR 流水线(PoW → 上传 → fork → HIF 签名 → SSE 流式 completion)与 CDP 自动登录
  • @xlight-oss/visionary-dsh:DSH 原生插件包(npm,纯 ESM 无构建,单包双插件行),经 ctx.tools 注册 deepseek_vision / deepseek_ocr 等 5 个原生工具,宿主进程内 spawn visionary-server 复用 Rust 管道(续聊/登录不受 bash 沙箱限制);内置文本模型图片桥接(纯文本模型会话粘贴图片自动放行 + 改写为文本引导)
  • visionary-zed-ext:Zed 扩展壳(仅 Zed 需要),按平台从 GitHub Releases 下载/缓存 visionary-server 并启动

安装

1. 安装二进制

# macOS / Linux 一键脚本
curl -LsSf https://github.com/xlight/deepseek-visionary/releases/latest/download/visionary-server-installer.sh | sh

# Windows(PowerShell 一键,自动绕过执行策略)
powershell -NoProfile -ExecutionPolicy Bypass -Command "irm https://github.com/xlight/deepseek-visionary/releases/latest/download/visionary-server-installer.ps1 | iex"

# 或 Homebrew
brew install xlight/tap/visionary-server

# 或 npm(全平台)
npm install -g @xlight-oss/visionary-server

Showing the opening section of the README — the full document lives in the repository以上为 README 开头摘要,完整文档在仓库内 · View the full README on GitHub →在 GitHub 查看完整 README →

← 上一个 Prev dsh-approve-for-me 下一个 Next prompt-skill-armory →