从官方源安装

openclaw openclaw解答 2026-04-09 3

我将为你介绍如何集成和使用AI小龙虾OpenClaw的Python SDK：

从官方源安装-第1张图片-OpenClaw下载官网 - OpenClaw电脑版 | ai小龙虾

安装方法

安装SDK

# 或从GitHub安装最新版本
pip install git+https://github.com/openclaw/openclaw-python.git

可选依赖安装

# 如果需要语音功能
pip install "openclaw[audio]"
# 如果需要图像处理功能
pip install "openclaw[vision]"
# 安装完整功能
pip install "openclaw[all]"

基本使用方法

初始化客户端

from openclaw import OpenClaw
# 方式1：使用API密钥
client = OpenClaw(api_key="your_api_key_here")
# 方式2：从环境变量读取
import os
client = OpenClaw(api_key=os.getenv("OPENCLAW_API_KEY"))

基础对话功能

# 单轮对话
response = client.chat("今天天气怎么样？")
print(response.content)
# 带参数的对话
response = client.chat(
    message="写一首关于春天的诗",
    model="claw-4",  # 指定模型版本
    temperature=0.7,  # 随机性控制
    max_tokens=500    # 最大生成长度
)

多轮对话

# 创建对话会话
conversation = client.create_conversation()
# 添加消息
conversation.add_message("user", "推荐一部好看的科幻电影")
response1 = conversation.get_response()
print(f"助理: {response1.content}")
# 继续对话
conversation.add_message("user", "能详细说说剧情吗？")
response2 = conversation.get_response()
print(f"助理: {response2.content}")
# 查看对话历史
for msg in conversation.history:
    print(f"{msg.role}: {msg.content}")

高级功能

流式输出

# 实时流式输出
response = client.chat_stream(
    message="解释一下量子计算",
    stream=True
)
for chunk in response:
    print(chunk.content, end="", flush=True)

文件处理

# 上传并处理文件
with open("document.pdf", "rb") as f:
    result = client.process_file(
        file=f,
        task="summarize",  # 翻译、分析等
        file_type="pdf"
    )
print(result.summary)
# 处理图片
from PIL import Image
image = Image.open("photo.jpg")
result = client.analyze_image(
    image=image,
    task="describe"  # 描述、识别物体等
)

函数调用

# 定义工具函数
def get_weather(city: str) -> str:
    """获取城市天气"""
    # 实际天气API调用
    return f"{city}的天气是晴天，25℃"
# 启用函数调用
response = client.chat_with_tools(
    message="北京天气如何？",
    tools=[get_weather],
    tool_choice="auto"
)
# 执行函数调用
if response.tool_calls:
    for call in response.tool_calls:
        result = call.function()
        print(f"函数结果: {result}")

批量处理

# 批量处理任务
tasks = [
    "总结第一段",
    "翻译成英文",
    "提取关键词"
]
results = client.batch_process(
    texts=["这是一个示例文本..."],
    tasks=tasks
)
for result in results:
    print(f"任务: {result.task}, 结果: {result.content}")

配置选项

from openclaw import OpenClaw, OpenClawConfig
# 自定义配置
config = OpenClawConfig(
    api_key="your_key",
    base_url="https://api.openclaw.ai/v1",  # 自定义API端点
    timeout=30.0,  # 超时设置
    max_retries=3,  # 重试次数
    request_timeout=60.0  # 请求超时
)
client = OpenClaw(config=config)
# 或直接在初始化时配置
client = OpenClaw(
    api_key="your_key",
    base_url="https://api.openclaw.ai/v1",
    timeout=30.0
)

异步支持

import asyncio
from openclaw import AsyncOpenClaw
async def main():
    # 异步客户端
    client = AsyncOpenClaw(api_key="your_key")
    # 异步对话
    response = await client.chat("异步测试")
    print(response.content)
    # 批量异步处理
    tasks = [client.chat(f"问题{i}") for i in range(5)]
    results = await asyncio.gather(*tasks)
    for i, result in enumerate(results):
        print(f"结果{i}: {result.content}")
# 运行异步函数
asyncio.run(main())

错误处理

from openclaw import OpenClaw, OpenClawError
client = OpenClaw(api_key="your_key")
try:
    response = client.chat("测试消息")
except OpenClawError as e:
    print(f"API错误: {e}")
except ConnectionError as e:
    print(f"连接错误: {e}")
except Exception as e:
    print(f"其他错误: {e}")
else:
    print(f"成功: {response.content}")
finally:
    # 清理资源
    client.close()

实用工具函数

# 计算Token数量
text = "这是一个测试文本"
token_count = client.count_tokens(text)
print(f"Token数量: {token_count}")
# 文本嵌入
embedding = client.embed_text(text)
print(f"嵌入维度: {len(embedding)}")
# 相似度计算
texts = ["文本1", "文本2", "文本3"]
similarities = client.calculate_similarity(texts[0], texts[1:])

完整示例

import os
from openclaw import OpenClaw
from dotenv import load_dotenv
# 加载环境变量
load_dotenv()
class OpenClawAssistant:
    def __init__(self):
        self.client = OpenClaw(
            api_key=os.getenv("OPENCLAW_API_KEY"),
            timeout=30.0
        )
        self.conversation = self.client.create_conversation()
    def chat(self, message: str) -> str:
        """处理用户消息"""
        self.conversation.add_message("user", message)
        response = self.conversation.get_response(
            temperature=0.7,
            max_tokens=1000
        )
        return response.content
    def reset(self):
        """重置对话"""
        self.conversation = self.client.create_conversation()
# 使用示例
if __name__ == "__main__":
    assistant = OpenClawAssistant()
    while True:
        user_input = input("你: ")
        if user_input.lower() == "退出":
            break
        response = assistant.chat(user_input)
        print(f"助理: {response}")