عن المشروع

Next AI Draw.io is a Next.js web application that integrates AI capabilities with draw.io diagrams. Users create, modify, and enhance diagrams through natural language commands and AI-assisted visualization. Key capabilities described in the README: - LLM-powered diagram creation and manipulation via natural language - Image-based diagram replication: upload existing diagrams or images for the AI to replicate and enhance - PDF and text file upload to extract content and generate diagrams - AI reasoning display for supported models (OpenAI o1/o3, Gemini, Claude, etc.) - Diagram history with version tracking and restore of previous versions - Interactive chat interface for real-time refinement - Cloud architecture diagram support (AWS, GCP, Azure) - Animated connectors between diagram elements Integration and deployment: - MCP server usable with Claude Desktop, Cursor, and VS Code; also installable via Claude Code CLI - Desktop applications for Windows, macOS, and Linux from the Releases page - Docker guide, plus deployment options for EdgeOne Pages, Vercel, and Cloudflare Workers - Local setup via git clone, npm install, env.example to .env.local, and npm run dev on port 6002 Provider support includes AWS Bedrock (default), OpenAI, Anthropic, Google AI, Google Vertex AI, Azure OpenAI, Ollama, OpenRouter, AIHubMix, DeepSeek, SiliconFlow, ModelScope, SGLang, Vercel AI Gateway, ByteDance Doubao, and Atlas Cloud. All providers except AWS Bedrock and OpenRouter support custom endpoints. Administrators can configure server-side models via AI_MODELS_CONFIG or ai-models.json, and an admin panel at /admin (protected by ADMIN_PASSWORD) manages models, access codes, features, observability, and quota. Technically, the app uses Next.js for the frontend and routing, the Vercel AI SDK (ai and @ai-sdk/*) for streaming responses and multi-provider support, and react-drawio for diagram representation. Diagrams are represented as draw.io XML that the AI generates or modifies. The README notes the task requires strong models for long-form output with strict formatting constraints, recommending Claude Sonnet 4.5, GPT-5.1, Gemini 3 Pro, and DeepSeek V3.2/R1, and mentions the claude series was trained on draw.io diagrams with cloud architecture logos. A live demo is available, and users can supply their own API key (stored locally in the browser) to bypass demo usage limits. The project is licensed under Apache 2.0.