One JSON object: QuireKit's embed object carries mind maps, flowcharts, and whiteboards into any editor

Because the embed object is document-agnostic, teams can stop building diagram support per editor and instead wire one adapter per document type. The structure also means AI tools can generate and modify diagrams without round-tripping through screenshots.

QuireKit ships five diagram components as MIT-licensed standalone packages: mind map, flowchart, whiteboard, text diagram (Mermaid/PlantUML/Graphviz), and a multi-dimensional table. Each one renders inside a document as a normal image you can resize and align, while the data behind it stays structured and editable. The same JSON object works as a block in rich text, a code fence in Markdown, and an image object in a paged Word-style document.

The design splits editing from storage: the editor has no idea what document it lives in, and each document type needs only a thin adapter that reads and writes the embed object. Two conventions connect them: a universal embed object carries the diagram data, a stable ID, and a preview image; an edit request passes the current data plus a commit callback that returns 'saved', 'conflict', 'readonly', or 'missing'. Kernels are swappable: the flow chart uses AntV X6, mind map uses simple-mind-map, whiteboard uses Excalidraw, and text diagrams use the official Mermaid/PlantUML/Graphviz libraries, all behind adapters.

The structured data is what lets AI write diagrams directly. Each component exposes command descriptions for inserting and updating diagrams; an AI assistant can decide where a flowchart or mind map fits, insert a real editable diagram, and recover from syntax errors. The integration paths cover ProseMirror/Tiptap block nodes, Markdown code fences, paged documents, and direct API calls.