Rss3208/Visiomaster

What it solves

Visiomaster solves the problem of converting static images of flowcharts, architecture diagrams, and academic module diagrams (including those generated by AI) into fully editable Microsoft Visio (.vsdx) files. Instead of simply embedding an image into a slide or document, it reconstructs the diagram's structure—shapes, text, containers, and connectors—so they can be modified manually.

How it works

The tool uses a multi-stage pipeline that separates image understanding from rendering:

1. Analysis: It starts with an image and generates a scene.json file, which acts as an intermediate description of the diagram's nodes, edges, fonts, and layout.
2. Validation & Audit: A series of Python scripts validate the scene's complexity and structure, ensuring that connectors and arrows are correctly mapped to the original image's facts.
3. Rendering: Using the Visio COM interface on Windows, the tool renders the scene.json description into a native .vsdx file, while also exporting .svg and .png for review.
4. Iterative Review: A "visual review loop" allows users or vision-capable LLMs to compare the original image with the reconstructed PNG and generate a repair plan to refine the scene.json for subsequent rebuilds.

Who it’s for

• Researchers and academics needing to recreate professional diagrams from papers.
• Professionals who need to convert AI-generated flowcharts into editable corporate formats.
• Users of AI coding assistants (like Codex or Claude Code) who want to automate the creation of Visio diagrams via a defined skill/context.

Highlights

• Editable Output: Produces genuine .vsdx files rather than static image overlays.
• Strict Reconstruction Workflow: Implements a rigorous process including "Arrow Plan Gates" and review bundles to ensure high fidelity to the original source.
• Semantic Component Library: Uses a controlled vocabulary of components (e.g., process_box, residual_loop, merge_bus) to maintain stability across different Visio versions and languages.
• AI-Agent Ready: Can be installed as a Codex Skill, allowing AI agents to drive the reconstruction process using the provided scripts and schema.