About A visual interface for creating and customizing LLM agents using drag-and-drop components. Designed for both developers and non-developers.

🌟 LangStar

English | 한국어

<a name="english"></a>

English Version

🚧 LangStar is currently under development. This project is currently in active development. Features, structure, and documentation may change frequently.

LangStar is a visual no-code/low-code platform that enables both developers and non-developers to easily build LLM-based agents through a drag-and-drop interface. Built workflows are automatically converted to Python code and can be deployed.

🎯 Project Goals

• Lower the barrier to entry for LLM agent development
• Provide intuitive drag-and-drop based visual modeling
• Automatically generate clean, modularized Python code
• Provide one-click deployment and testing environment
• Save development time and improve accessibility

📤 Key Features

🎨 Visual Workflow Builder

• Compose LLM agent workflows with drag-and-drop interface
• Real-time node connection and data flow visualization
• Support for individual node execution and debugging

🔧 Various Node Types

• Start Node: Workflow starting point and initial variable setup
• Prompt Node: Prompt template creation and variable substitution
• Agent Node: LLM model execution (AWS Bedrock support)
• Function Node: User-defined Python function execution
• User Node: Custom Python code blocks
• Condition Node: Conditional branching
• Merge Node: Multi-input data merging
• End Node: Workflow termination point

🤖 AI Model Connection Management

• AWS Bedrock fully supported (Gemini, Nova, etc.)
• OpenAI, Google, Anthropic UI support (backend integration in progress)
• API key and configuration management
• Model parameter adjustment (temperature, max tokens, etc.)

🚀 Local Execution and Testing

• Automatic LangGraph code generation
• Local execution environment and testing
• **Execution versio