跳到主要内容

Anthropic格式API

Constreet提供完全兼容Anthropic Messages API的接口，专为Claude系列模型优化。如果你使用Anthropic官方SDK，只需修改base_url即可无缝切换。

接口地址​

https://api.constreet.cc/v1

认证方式​

支持两种认证方式：

方式一：x-api-key（推荐）​

x-api-key: YOUR_API_KEY

方式二：Authorization​

Authorization: Bearer YOUR_API_KEY

Messages API​

接口说明​

Claude的核心对话接口

请求地址​

POST /v1/messages

请求示例​

curl https://api.constreet.cc/v1/messages \  -H "content-type: application/json" \  -H "x-api-key: YOUR_API_KEY" \  -d '{    "model": "claude-sonnet-4-6",    "max_tokens": 1024,    "messages": [      {        "role": "user",        "content": "用Python写一个二分查找算法"      }    ]  }'

请求参数​

Messages格式​

与OpenAI不同，Anthropic的system提示词是独立参数：

{  "model": "claude-sonnet-4-6",  "max_tokens": 1024,  "system": "你是一个Python编程专家",  "messages": [    {      "role": "user",      "content": "写一个快速排序"    },    {      "role": "assistant",      "content": "好的，我来写一个Python的快速排序..."    },    {      "role": "user",      "content": "加上详细注释"    }  ]}

注意：

  • messages中只能有user和assistant角色
  • 必须以user消息开始
  • user和assistant消息必须交替出现

响应格式​

{  "id": "msg_123",  "type": "message",  "role": "assistant",  "content": [    {      "type": "text",      "text": "这是Claude的回复内容"    }  ],  "model": "claude-sonnet-4-6",  "stop_reason": "end_turn",  "usage": {    "input_tokens": 25,    "output_tokens": 150  }}

流式输出​

设置stream: true启用：

curl https://api.constreet.cc/v1/messages \  -H "content-type: application/json" \  -H "x-api-key: YOUR_API_KEY" \  -d '{    "model": "claude-sonnet-4-6",    "max_tokens": 1024,    "messages": [{"role": "user", "content": "你好"}],    "stream": true  }'

流式响应事件：

event: message_startdata: {"type":"message_start","message":{"id":"msg_123",...}}event: content_block_deltadata: {"type":"content_block_delta","delta":{"type":"text_delta","text":"你"}}event: content_block_deltadata: {"type":"content_block_delta","delta":{"type":"text_delta","text":"好"}}event: message_stopdata: {"type":"message_stop"}

支持的模型​

多模态支持​

Claude支持图片输入：

图片格式​

{  "model": "claude-sonnet-4-6",  "max_tokens": 1024,  "messages": [    {      "role": "user",      "content": [        {          "type": "image",          "source": {            "type": "base64",            "media_type": "image/jpeg",            "data": "/9j/4AAQSkZJRg..."          }        },        {          "type": "text",          "text": "这张图片里有什么？"        }      ]    }  ]}

支持的图片格式：

  • JPEG
  • PNG
  • GIF
  • WebP

最大图片大小：5MB

代码示例​

Python（官方SDK）​

from anthropic import Anthropicnlient = Anthropic(    api_key="YOUR_API_KEY",    base_url="https://api.constreet.cc/v1")message = client.messages.create(    model="claude-sonnet-4-6",    max_tokens=1024,    messages=[        {"role": "user", "content": "写一个Python装饰器"}    ])print(message.content[0].text)

Python（流式输出）​

with client.messages.stream(    model="claude-sonnet-4-6",    max_tokens=1024,    messages=[{"role": "user", "content": "写一篇文章"}]) as stream:    for text in stream.text_stream:        print(text, end="", flush=True)

Node.js（官方SDK）​

import Anthropic from '@anthropic-ai/sdk';const client = new Anthropic({  apiKey: 'YOUR_API_KEY',  baseURL: 'https://api.constreet.cc/v1'});const message = await client.messages.create({  model: 'claude-sonnet-4-6',  max_tokens: 1024,  messages: [    { role: 'user', content: '写一个快速排序' }  ]});console.log(message.content[0].text);

cURL​

curl https://api.constreet.cc/v1/messages \  -H "content-type: application/json" \  -H "x-api-key: YOUR_API_KEY" \  -d '{    "model": "claude-sonnet-4-6",    "max_tokens": 1024,    "messages": [      {"role": "user", "content": "Hello, Claude!"}    ]  }'

错误处理​

错误响应格式​

{  "type": "error",  "error": {    "type": "invalid_request_error",    "message": "错误描述"  }}

常见错误类型​

错误处理示例​

from anthropic import Anthropic, APIErrorclient = Anthropic(    api_key="YOUR_API_KEY",    base_url="https://api.constreet.cc/v1")try:    message = client.messages.create(        model="claude-sonnet-4-6",        max_tokens=1024,        messages=[{"role": "user", "content": "Hello"}]    )except APIError as e:    print(f"错误类型: {e.type}")    print(f"错误信息: {e.message}")

高级功能​

系统提示词​

message = client.messages.create(    model="claude-sonnet-4-6",    max_tokens=1024,    system="你是一个Python专家，擅长数据分析和机器学习。",    messages=[        {"role": "user", "content": "如何使用pandas处理缺失值？"}    ])

多轮对话​

conversation = [    {"role": "user", "content": "写一个用户类"},    {"role": "assistant", "content": "好的，我来写一个User类..."},    {"role": "user", "content": "加上密码加密功能"}]message = client.messages.create(    model="claude-sonnet-4-6",    max_tokens=1024,    messages=conversation)

图片分析​

import base64with open("image.jpg", "rb") as f:    image_data = base64.b64encode(f.read()).decode()message = client.messages.create(    model="claude-sonnet-4-6",    max_tokens=1024,    messages=[        {            "role": "user",            "content": [                {                    "type": "image",                    "source": {                        "type": "base64",                        "media_type": "image/jpeg",                        "data": image_data                    }                },                {                    "type": "text",                    "text": "分析这张架构图"                }            ]        }    ])

最佳实践​

合理设置max_tokens​

不同模型的最大输出不同：

  • Opus: 32K
  • Sonnet: 64K
  • Haiku: 8K

根据需求设置合适的值，避免浪费。

使用system提示词​

将角色设定放在system参数中，而不是messages里：

# ✅ 推荐system="你是一个Python专家"messages=[{"role": "user", "content": "写代码"}]# ❌ 不推荐messages=[    {"role": "user", "content": "你是Python专家。写代码"}]

处理长对话​

Claude有200K上下文，但建议：

  • 定期总结对话
  • 删除不重要的历史
  • 使用缓存机制（如果支持）

错误重试​

from tenacity import retry, stop_after_attempt, wait_exponential@retry(    stop=stop_after_attempt(3),    wait=wait_exponential(multiplier=1, min=4, max=10))def call_claude():    return client.messages.create(...)

与OpenAI格式对比​

速率限制​

下一步​