Comet LLM Client V0.0.5-Alpha

高性能LLM客户端，聚合OpenAI和Anthropic API，提供统一接口和优化性能，使得开发者无需单独适配服务商，尽可能提高效率

特性

• 🚀 高性能异步支持
• 🔧 统一API接口（兼容OpenAI和Anthropic）
• 💾 智能HTTP连接池
• 🔄 自动重试和错误处理
• 📊 完整的类型提示
• 🔒 Pydantic数据验证
• 🛠️ 工具调用支持
• 📡 流式响应支持
• ✨ 优化的接口，无需手动适配每个服务商

安装

pip install -e .

快速开始

使用OpenAI

from CometLLM import CometLLM

# 方法1: 使用工厂方法
client = CometLLM.openai(api_key="your-api-key", model="gpt-4")

# 方法2: 使用配置字典
client = CometLLM({
    "provider": "openai",
    "api_key": "your-api-key",
    "model": "gpt-4",
    "temperature": 0.7,
})

# 简单完成
response = client.complete("Hello, how are you?")
print(response)

# 聊天完成
from CometLLM import Message

messages = [
    {"role": "system", "content": "You are a helpful assistant"},
    {"role": "user", "content": "What is Python?"},
]

response = client.chat(messages)
print(response.choices[0].message.content)

使用Anthropic (Claude)

from CometLLM import CometLLM

client = CometLLM.anthropic(api_key="your-api-key", model="claude-3-opus-20240229")

response = client.complete("Hello, how are you?")
print(response)

流式响应

# 同步流式
for chunk in client.stream([{"role": "user", "content": "Tell me a story"}]):
    content = chunk.choices[0].delta.get("content", "")
    print(content, end="", flush=True)

# 异步流式
async for chunk in client.astream([{"role": "user", "content": "Tell me a story"}]):
    content = chunk.choices[0].delta.get("content", "")
    print(content, end="", flush=True)

异步使用

import asyncio
from CometLLM import CometLLM

async def main():
    client = CometLLM.openai(api_key="your-api-key")
    
    # 异步完成
    response = await client.acomplete("Hello!")
    print(response)
    
    # 异步聊天
    response = await client.achat([
        {"role": "user", "content": "What is AI?"}
    ])
    print(response.choices[0].message.content)
    
    await client.aclose()

asyncio.run(main())

工具调用

from CometLLM import CometLLM, Tool, ToolFunction

client = CometLLM.openai(api_key="your-api-key", model="gpt-4")

tools = [
    Tool(
        type="function",
        function=ToolFunction(
            name="get_weather",
            description="Get weather for a location",
            parameters={
                "type": "object",
                "properties": {
                    "location": {"type": "string"},
                },
                "required": ["location"],
            },
        )
    )
]

response = client.chat(
    messages=[{"role": "user", "content": "What's the weather in Beijing?"}],
    tools=tools,
)

# 检查工具调用
if response.choices[0].message.tool_calls:
    for tool_call in response.choices[0].message.tool_calls:
        print(f"Function: {tool_call['function']['name']}")
        print(f"Arguments: {tool_call['function']['arguments']}")

配置选项

参数	类型	默认值	说明
provider	str	必填	提供商: openai 或 anthropic
api_key	str	必填	API密钥
model	str	必填	模型名称
base_url	str	None	自定义API地址
temperature	float	0.7	采样温度 (0-2)
max_tokens	int	None	最大生成token数
top_p	float	1.0	核采样参数
timeout	float	60.0	请求超时时间(秒)
max_retries	int	3	最大重试次数
retry_delay	float	1.0	重试延迟(秒)

错误处理

from CometLLM import CometLLM
from CometLLM.exceptions import AuthenticationError, RateLimitError, APIError

client = CometLLM.openai(api_key="your-api-key")

try:
    response = client.complete("Hello")
except AuthenticationError as e:
    print(f"认证失败: {e.message}")
except RateLimitError as e:
    print(f"速率限制: {e.message}")
except APIError as e:
    print(f"API错误: {e.message} (状态码: {e.status_code})")

项目结构

CometLLM/
├── __init__.py          # 包入口
├── client.py            # 统一客户端
├── types.py             # 类型定义
├── exceptions.py        # 异常定义
├── base.py              # 基础客户端
└── providers/
    ├── __init__.py
    ├── openai.py        # OpenAI实现
    └── anthropic.py     # Anthropic实现

许可证

MIT