HN 표시: AntroCode-종속성 없음, 단일 파일 로컬 AI 클라이언트, 4D에서 159 클론

「We got tired of downloading hundreds of MB of dependencies just to run a simple UI. AntroCode is an ultra-lightweight, plug-and-play, pure frontend LLM client.」 🚫 Zero Dependencies: Say goodbye to bloated node_modules and cumbersome environments. ⚙️ No Backend Server: No backend deployment needed; just one Python script to launch. 🧠 Native DeepSeek Support: Perfectly compatible with Think / Fast modes. 💻 Developer-Friendly: Designed for those who appreciate a clean, fast workflow. I stripped away all the complex infrastructure, leaving only pure productivity and absolute data security: Tired of npm install ing hundreds of MB of dependencies every time? AntroCode uses a single Python script to dynamically generate a single-file HTML. Double-click to run, your browser opens in seconds, and you're ready to work. That's the elegant efficiency a true geek deserves. Your data, your computer. All chat history, workspace state, and API Keys are stored exclusively in your browser's LocalStorage . No intermediary servers can intercept your proprietary code, perfectly meeting the highest compliance needs for enterprises and freelance developers. This isn't a toy chat window; it's a hardcore coding assistant: - Syntax Highlighting & One-Click Copy: Built-in highlight.js support renders code blocks of any complexity perfectly, complete with a one-click Copy function. - Local Token Estimation: Features a self-developed token estimation algorithm. Monitor consumption per workspace in real-time to precisely control API costs, ensuring every cent counts. - Customizable Context Depth: Freely adjust the number of historical messages carried in the context window via settings, avoiding wasteful token usage. A dark mode, geek-themed interface with a signature tech-green accent. Intuitive sidebar management lets you easily switch between different Agent and Project workspaces. Provides an IDE-like immersive conversation experience. Native rendering support for Chain of Thought (CoT) from models like DeepSeek-Reasoner (Think mode). The AI's thinking process is elegantly folded and visually separated, allowing you to see every logical step the AI takes to solve a problem. Get AntroCode up and running in just 10 seconds: - Download the Script: Download AntroCode_1.py from this project. - Run the Script: python AntroCode_1.py Start Using: The script will automatically generate antrocode_ui_pro_max.html in your current directory and open it in your browser. Set Your API Key: Click ⚙️ Settings in the bottom-left corner, enter your API Key, and start chatting! 🗺️ Roadmap AntroCode is just the first step. We're building the future Edge AI ecosystem: [x] Phase 1: Launch the ultra-lightweight, painless deployment AntroCode UI. [ ] Phase 2 (Coming Soon): Integrate support for local open-source models (Ollama, etc.) enabling true 'zero API cost' local inference. [ ] Phase 3: Release Aston 1, a top-tier model fine-tuned specifically for Python frameworks, unlocking unprecedented local code generation capability. [ ] Phase 4: Launch a free educational program for students and a proprietary Enterprise Edition. 🤝 Contributing & Community Issues and Pull Requests are welcome! If you like AntroCode, please give us a ⭐️ Star – it helps our open-source development immensely! By AntroMind.