AnyCoder - AI Code Generator with React Frontend

🎨 Features

• Modern React UI: Apple-inspired design with VS Code layout
• Real-time Streaming: Server-Sent Events for live code generation
• Multi-Model Support: MiniMax M2, DeepSeek V3, and more via HuggingFace InferenceClient
• Multiple Languages: HTML, Gradio, Streamlit, React, Transformers.js, ComfyUI
• Authentication: HuggingFace OAuth + Dev mode for local testing
• One-Click Deployment: Deploy generated apps directly to HF Spaces

🏗️ Architecture

anycoder/
├── backend_api.py           # FastAPI backend with streaming
├── frontend/                # Next.js React frontend
│   ├── src/
│   │   ├── app/            # Pages (page.tsx, layout.tsx, globals.css)
│   │   ├── components/     # React components
│   │   ├── lib/            # API client, auth utilities
│   │   └── types/          # TypeScript types
│   └── package.json
├── requirements.txt        # Python dependencies
├── Dockerfile              # Docker Space configuration
└── start_fullstack.sh      # Local development script

🚀 Quick Start

Local Development

1. Backend:

1. Frontend (new terminal):

1. Open http://localhost:3000

Using start script:

🐳 Docker Space Deployment

• Builds the Next.js frontend
• Runs FastAPI backend on port 7860
• Uses proper user permissions (UID 1000)
• Handles environment variables securely

🔑 Authentication

• Dev Mode (localhost): Mock login for testing
• Production: HuggingFace OAuth with manage-repos scope

📝 Supported Languages

• html - Static HTML pages
• gradio - Python Gradio apps
• streamlit - Python Streamlit apps
• react - React/Next.js apps
• transformers.js - Browser ML apps
• comfyui - ComfyUI workflows

🤖 Available Models

• Gemini 3 Pro Preview (Default) - Google's latest with deep thinking & Google Search
• MiniMax M2 (via HF router with Novita)
• DeepSeek V3/V3.1
• DeepSeek R1
• And more via HuggingFace InferenceClient

🎯 Usage

1. Sign in with HuggingFace (or use Dev Login locally)
2. Select a language and AI model
3. Describe your app in the chat
4. Watch code generate in real-time
5. Click 🚀 Deploy to publish to HF Spaces

🛠️ Environment Variables

• HF_TOKEN - HuggingFace API token (required)
• GEMINI_API_KEY - Google Gemini API key (required for Gemini 3 Pro Preview)
• POE_API_KEY - Poe API key (optional, for GPT-5 and Claude models)
• DASHSCOPE_API_KEY - DashScope API key (optional, for Qwen models)
• OPENROUTER_API_KEY - OpenRouter API key (optional, for Sherlock models)
• MISTRAL_API_KEY - Mistral API key (optional, for Mistral models)

📦 Tech Stack

• Next.js 14
• TypeScript
• Tailwind CSS
• Monaco Editor

• FastAPI
• HuggingFace Hub
• Server-Sent Events (SSE)

📄 License