Detailed Introduction
DeepDiagram is an open-source agentic visualization platform that turns natural language and multimodal inputs into structured, interactive visuals. Specialized agents collaborate (mind map, flowchart, Mermaid, chart agents) to provide real-time previews, in-canvas editing, and high-quality exports while preserving execution traces and versioned history for reproducibility and iteration. See the Demo.
Main Features
DeepDiagram offers intent-aware agent routing and ReAct orchestration to select the best visualization agent for a request. It supports multimodal and document parsing (PDF, DOCX, images), interactive canvas controls (zoom, drag), and high-resolution exports (SVG/PNG). Session history, branching, and versioning make it easy to compare and restore generated results.
Use Cases
Ideal for knowledge mapping and brainstorming, business process design, technical documentation visualization, and data storytelling. Teams can quickly convert unstructured requirements or documents into mind maps, flowcharts, or interactive charts to speed communication and reviews; it also digitizes whiteboards and sketches into structured diagrams.
Technical Features
The frontend uses React + Vite with visualization libraries (React Flow, Mind-elixir, Mermaid, ECharts). The backend runs FastAPI + LangGraph for agent orchestration and SSE streaming; document parsing and persistence rely on PostgreSQL. The project is licensed under AGPL-3.0; see the repository for deployment instructions.