本指南帮助你在几分钟内安装 Agentica 并运行第一个 AI 智能体。
pip install -U agentica
从源码安装(开发模式):
git clone https://github.com/shibing624/agentica.git
cd agentica
pip install -e .
在 ~/.agentica/.env 中配置,或直接设置环境变量:
# 推荐:智谱AI(glm-4.7-flash 免费,支持工具调用,128k 上下文)
export ZHIPUAI_API_KEY="your-api-key"
# OpenAI
export OPENAI_API_KEY="sk-xxx"
# DeepSeek
export DEEPSEEK_API_KEY="your-api-key"
# Anthropic (Claude)
export ANTHROPIC_API_KEY="your-api-key"
import asyncio
from agentica import Agent, ZhipuAI
async def main():
agent = Agent(model=ZhipuAI())
result = await agent.run("一句话介绍北京")
print(result.content)
asyncio.run(main())
输出:
北京是中国的首都,是一座拥有三千多年历史的文化名城,也是全国的政治、文化和国际交流中心。
如果你不需要 async,可以使用同步适配器:
from agentica import Agent, OpenAIChat
agent = Agent(model=OpenAIChat(id="gpt-4o-mini"))
result = agent.run_sync("什么是量子计算?")
print(result.content)
import asyncio
from agentica import Agent, OpenAIChat
async def main():
agent = Agent(model=OpenAIChat(id="gpt-4o-mini"))
async for chunk in agent.run_stream("写一首关于春天的诗"):
if chunk.content:
print(chunk.content, end="", flush=True)
print()
asyncio.run(main())
from agentica import Agent, OpenAIChat
agent = Agent(model=OpenAIChat(id="gpt-4o-mini"))
for chunk in agent.run_stream_sync("写一首关于春天的诗"):
if chunk.content:
print(chunk.content, end="", flush=True)
print()
给 Agent 配备工具,让它能执行实际操作:
import asyncio
from agentica import Agent, OpenAIChat, BaiduSearchTool
async def main():
agent = Agent(
model=OpenAIChat(id="gpt-4o"),
tools=[BaiduSearchTool()],
instructions=["使用搜索工具获取最新信息,用中文回答"],
)
result = await agent.run("今天的科技新闻有哪些?")
print(result.content)
asyncio.run(main())
使用 Pydantic 模型获取结构化数据:
import asyncio
from pydantic import BaseModel, Field
from agentica import Agent, OpenAIChat
class CityInfo(BaseModel):
name: str = Field(description="城市名称")
country: str = Field(description="所属国家")
population: str = Field(description="人口")
highlights: list[str] = Field(description="主要特色")
async def main():
agent = Agent(
model=OpenAIChat(id="gpt-4o-mini"),
response_model=CityInfo,
)
result = await agent.run("介绍北京")
city: CityInfo = result.content
print(f"{city.name} ({city.country})")
print(f"人口: {city.population}")
for h in city.highlights:
print(f" - {h}")
asyncio.run(main())
安装后可直接在终端使用:
# 交互模式
agentica
# 单次查询
agentica --query "解释什么是 RAG"
# 指定模型
agentica --model_provider zhipuai --model_name glm-4.7-flash
# 启用工具
agentica --tools baidu_search shell

Agentica 支持 20+ 模型提供商,详见 模型提供商指南:
from agentica import (
OpenAIChat, # GPT-4o, GPT-4o-mini
DeepSeek, # deepseek-chat, deepseek-reasoner
ZhipuAI, # glm-4.7-flash (免费)
Claude, # claude-sonnet-4-20250514
Qwen, # qwen-plus, qwen-turbo
Moonshot, # moonshot-v1-128k
Doubao, # doubao-pro-32k
Ollama, # 本地模型
# ...更多
)
# 使用不同模型
agent = Agent(model=DeepSeek(id="deepseek-chat"))
agent = Agent(model=ZhipuAI(id="glm-4.7-flash"))
agent = Agent(model=Ollama(id="llama3.1"))