Quick Start

Installation

Install with pip (recommended)

Requires Python 3.12+. One command is enough.

pip install cellium-agent

# Optional: Feishu channel dependency
pip install "cellium-agent[lark]"

Portable package (no Python setup)

Download the archive for your platform from Releases and run it directly.

# Linux x64
curl -LO https://github.com/Cellium-Project/Cellium-Agent/releases/latest/download/Cellium-Agent-Linux.tar.gz
tar -xzf Cellium-Agent-Linux.tar.gz
cd Cellium-Agent-Linux && ./start-cellium.sh

On Windows, unzip and run CelliumAgent.exe; on Linux / macOS run ./start-cellium.sh. ARM64 and macOS builds are on the Releases page.

Install from source

git clone https://github.com/Cellium-Project/Cellium-Agent.git
cd Cellium-Agent
pip install -e .

Python

Python 3.12+ is required.

Dependencies

Configure Model

Edit config file to set API key

# config/agent/llm.yaml
llm:
  models:
  - name: model-1
    api_key: "${LLM_MODELS_API_KEY}"
    base_url: "https://api.openai.com/v1"
    model: "gpt-4o"
    temperature: 0.7
    timeout: 120
  current_model: model-1

其他配置文件（agent.yaml、server.yaml、memory.yaml 等）使用默认值即可，无需修改。

Start Service

After installing with pip, run cellium-web for the WebUI or cellium for the terminal TUI.

cellium-web     # WebUI (default port 18000)
cellium         # terminal TUI

The portable build needs no cellium-web command — run ./start-cellium.sh (or double-click CelliumAgent.exe on Windows).

After startup, visit http://localhost:18000 to open the chat interface

• Chat UI: http://localhost:18000
• API Docs: http://localhost:18000/docs