[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-85151":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":12,"contributorsCount":12,"subscribersCount":12,"size":12,"stars1d":12,"stars7d":12,"stars30d":12,"stars90d":12,"forks30d":12,"starsTrendScore":12,"compositeScore":14,"rankGlobal":9,"rankLanguage":9,"license":9,"archived":15,"fork":15,"defaultBranch":16,"hasWiki":17,"hasPages":15,"topics":18,"createdAt":9,"pushedAt":9,"updatedAt":19,"readmeContent":20,"aiSummary":9,"trendingCount":12,"starSnapshotCount":12,"syncStatus":21,"lastSyncTime":22,"discoverSource":23},85151,"Auto-Visio-Helper","0Antique\u002FAuto-Visio-Helper","0Antique","用于帮助Codex操作本地Visio生成可编辑科研绘图的skill，支持自然语言转图、图片复现、JVSDX 生成和预览导出。欢迎大家使用~~",null,"Python",56,0,51,33,false,"master",true,[],"2026-06-15 10:04:54","# Auto-Visio-Helper\n\n> 把“论文图想法 \u002F 截图 \u002F 草图”变成真正可编辑的 Microsoft Visio `.vsdx`。\n\n![License: MIT](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-yellow.svg)\n![Codex Skill](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FCodex-Skill-blue)\n![Visio](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMicrosoft-Visio-3955A3)\n![Python](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPython-3.10%2B-3776AB)\n![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002F0Antique\u002FAuto-Visio-Helper?style=social)\n\n![Auto-Visio-Helper poster](assets\u002Fauto-visio-helper-poster-image2.png)\n\n**Auto-Visio-Helper** 是一个面向 Codex 的 Visio 科研绘图 skill。它不让 AI 直接“凭感觉拖拽图形”，而是先把你的自然语言需求、论文截图、PPT 截图或手绘草图整理成可检查的 JSON 绘图规范，再调用本地 Microsoft Visio 自动生成可编辑的 `.vsdx` 文件，并导出 `.png` 预览供你确认。\n\n如果你也经常遇到这些问题，这个仓库就是为你准备的：\n\n- AI 生成的论文图很好看，但只能得到一张不可编辑的图片。\n- 论文截图想复现，手动在 Visio 里对齐、连线、调字号太耗时间。\n- 模型结构、流程图、系统架构图需要反复改版，必须保留可编辑源文件。\n- 想让 Codex 更稳定地完成科研绘图，而不是每次从零提示。\n\nEnglish documentation: [README-EN.md](README-EN.md)\n\n## 亮点\n\n- **可编辑优先**：最终产物是 Visio 形状、连接线、文本框、图层和命名对象，不是一张扁平截图。\n- **先规划再绘图**：先生成 JSON drawing spec，确认布局、风格、元素和导出格式后再渲染。\n- **适合科研图**：内置模型结构图、方法流程图、系统架构图、实验流程图、概念图、时序图等绘图模式。\n- **支持参考图复现**：可根据论文截图、PPT 截图、手绘草图重绘为可编辑 Visio 图。\n- **本地 Visio 自动化**：通过 `scripts\u002Frender_visio.py` 调用 Microsoft Visio COM 自动化生成 `.vsdx`。\n- **可预览可迭代**：支持 dry-run 校验和 `.png` 预览导出，方便逐轮修改到论文可用。\n\n## 效果预览\n\n### 最大池化示意图\n\n| 用户需求 \u002F 参考输入 | Auto-Visio-Helper 输出 |\n| --- | --- |\n| ![max pooling prompt](assets\u002F1780674451366.png) | ![max pooling demo](demo\u002Fmax_pooling_demo.png) |\n\n### YOLO 架构复现\n\n| 参考图 | 可编辑 Visio 复现效果 |\n| --- | --- |\n| ![YOLO reference](assets\u002F1780678594787.png) | ![YOLO architecture](assets\u002Fyolo11_architecture.png) |\n\n### 技术架构图复现\n\n| 参考图 | 可编辑 Visio 复现效果 |\n| --- | --- |\n| ![technical architecture reference](assets\u002F1780678953352.png) | ![technical architecture](assets\u002Ftech_architecture.png) |\n\n## 适合场景\n\n- 论文方法图、总体框架图、技术路线图。\n- YOLO、Transformer、CNN、Encoder-Decoder、检测头、注意力模块等模型结构图。\n- 数据处理、训练、推理、评估、部署流程图。\n- 软件系统架构图、模块交互图、接口调用图。\n- 实验设计、消融实验、Benchmark 对比流程。\n- 将已有截图重绘为可编辑 Visio 图，方便后续投稿、答辩和汇报修改。\n\n## 工作流\n\n```mermaid\nflowchart LR\n    A[\"自然语言需求 \u002F 参考图片\"] --> B[\"Codex 分析图类型与布局\"]\n    B --> C[\"生成 JSON drawing spec\"]\n    C --> D[\"用户确认结构、风格与导出格式\"]\n    D --> E[\"本地 Visio 渲染 .vsdx\"]\n    E --> F[\"导出 PNG 预览\"]\n    F --> G[\"检查并迭代到可投稿版本\"]\n```\n\n## 快速安装\n\n将仓库克隆到 Codex skills 目录：\n\n```powershell\ngit clone https:\u002F\u002Fgithub.com\u002F0Antique\u002FAuto-Visio-Helper.git $env:USERPROFILE\\.codex\\skills\\auto-visio-helper\n```\n\n然后重启 Codex，或按你的 Codex 环境重新加载 skills。\n\n## 使用示例\n\n在 Codex 中直接调用 skill：\n\n```text\nUse $auto-visio-helper to create an editable Visio diagram for my YOLO11 pest detection method.\nInclude dataset preprocessing, backbone, feature fusion, detection head, loss, and output prediction.\nUse a clean paper-style layout and export a PNG preview.\n```\n\n中文也可以：\n\n```text\n使用 $auto-visio-helper 帮我画一张可编辑的 Visio 论文方法图。\n内容包括：数据预处理、主干网络、特征融合、检测头、损失函数和预测输出。\n请先给出 JSON 绘图规范，确认后再生成 .vsdx 和 .png 预览。\n```\n## 运行依赖\n\n只做 drawing spec 生成和 dry-run 校验：\n\n- Python 3.10+\n\n真正调用本地 Visio 生成 `.vsdx`：\n\n- Windows\n- Microsoft Visio 桌面版\n- Python 包 `pywin32`\n\n安装 `pywin32`：\n\n```powershell\npython -m pip install pywin32\n```\n\n## 目录结构\n\n```text\nAuto-Visio-Helper\u002F\n├── SKILL.md\n├── README.md\n├── README-EN.md\n├── agents\u002F\n│   └── openai.yaml\n├── assets\u002F\n│   ├── auto-visio-helper-poster-image2.png\n│   └── ...\n├── demo\u002F\n│   ├── max_pooling_demo.vsdx\n│   ├── yolo11_architecture.vsdx\n│   └── tech_architecture.vsdx\n├── references\u002F\n│   ├── diagram_types.md\n│   ├── drawing_spec.md\n│   ├── style_guide.md\n│   ├── visio_automation.md\n│   └── 绘图模版.vsdx\n└── scripts\u002F\n    └── render_visio.py\n```\n\n## Star History\n\n如果这个 skill 帮你把科研绘图从“截图糊上去”推进到“可编辑、可复用、可投稿”，欢迎给项目点一个 Star。\n\n[![Star History Chart](https:\u002F\u002Fapi.star-history.com\u002Fsvg?repos=0Antique\u002FAuto-Visio-Helper&type=Date)](https:\u002F\u002Fstar-history.com\u002F#0Antique\u002FAuto-Visio-Helper&Date)\n\n> Hi，这里是Antique的AI小站。作为一名 AI Explorer，我坚信 AI 必将重塑当下生活方式，进而推动生产关系的全新变革，且终将以你我皆无法预料的形式改造人类文明。诸位，拥抱AI吧……\n\n",2,"2026-06-15 02:30:10","CREATED_QUERY"]