Clean Graph

A minimal conversational AI agent built with LangGraph, designed for agent scaffolding and playground, providing a practical starting point for learning, building, and experimenting with advanced agent technologies.

中文版

Prerequisites

Required Tools

• Python: Programming language environment
• pip: Python package manager
• Miniconda: Lightweight conda environment manager
• VSCode: Recommended development environment

Installation Guides

• Python for MacOS
• Miniconda installation
• pip Installation
• Visual Studio Code on macOS

Quick Start

1. Environment Setup

Create a conda environment:

conda create -n clean_graph python=3.12
conda activate clean_graph

2. Install Dependencies

pip install -r requirements.txt

3. Configure Environment

Edit .env file with your configuration:

# Required: LLM API Configuration (supports any OpenAI-compatible API)
LLM_API_BASE=http://localhost:1234/v1  # LM Studio, Ollama, or other OpenAI-compatible API endpoint
LLM_MODEL=qwen/qwen3-next-80b          # Your model name
LLM_API_KEY=your-api-key-here          # API key for authentication

# Optional: LangSmith Tracing
LANGSMITH_TRACING=false                # Set to 'true' to enable
LANGSMITH_API_KEY=your-langsmith-key   # Your LangSmith API key

4. Launch LangGraph Studio

langgraph dev --no-reload

Access the LangGraph Studio interface at http://localhost:2024.

Configuration

LLM Setup

This project supports multiple LLM providers, with LM Studio recommended for local development:

LM Studio Setup

1. Download and install LM Studio
2. Select and download suitable models in LM Studio
3. Start the local server (typically at http://localhost:1234)

LangSmith Registration

For the best development experience, register for a LangSmith account:

1. Visit LangSmith and create an account
2. Get your API key
3. Configure LANGSMITH_API_KEY in your .env file

Note: Even when using local graphs with LangGraph Studio, you still need to register for a LangSmith Key (free) and have logged into LangSmith (as of November 2025).

Environment Variables

Variable	Required	Description	Default
LLM_API_BASE	Yes	Base URL for LLM API endpoint	-
LLM_MODEL	Yes	Model name to use	-
LLM_API_KEY	Yes	API key for authentication	-
LANGSMITH_TRACING	No	Enable LangSmith tracing	false
LANGSMITH_API_KEY	Yes	LangSmith API key	-

Supported LLM Providers

This project is compatible with any OpenAI-compatible API:

• Local LLMs: LM Studio, Ollama, LocalAI
• Cloud Providers: OpenAI, Together AI, Groq, etc.

Project Structure

clean_graph/
├── src/
│   ├── __init__.py      # Package initialization
│   ├── graph.py         # Main LangGraph application logic
│   └── llms.py          # LLM configuration and setup
├── .env                 # Environment configuration
├── langgraph.json       # LangGraph application definition
├── requirements.txt     # Python dependencies
├── README.md           # English documentation
└── README_CN.md        # Chinese documentation

Dependencies

• langchain~=1.0 - Core LangChain framework
• langgraph~=1.0 - Graph-based AI application framework
• langchain-core~=1.0 - Core LangChain components
• langchain_openai - OpenAI API client
• python-dotenv~=1.0 - Environment variable management
• langgraph-checkpoint>=2.1.0 - State checkpointing
• langgraph-cli[inmem] - Development tools (in-memory storage)
• pydantic~=2.0 - Data validation and settings management

References

• LangGraph Studio Doc
• LangGraph Doc