Messages

Anthropic 原生消息接口。Claude 模型用它，相比 OpenAI 的 Chat Completions，区别在 system 顶层独立、content 支持 block 数组、工具调用走 tool_use / tool_result。

协议边界

这是 Anthropic 请求/响应格式。Claude 与 Deepseek 分组是主要使用场景；部分 OpenAI 分组也可能通过网关转换支持 /v1/messages，但是否开放取决于分组配置。

Endpoint

POST https://api.sakrylle.com/v1/messages

请求头

名称	必填	说明
`Authorization`	是	`Bearer sk-xxxxxxxxxxxxxxxx`
`Content-Type`	是	`application/json`
`anthropic-version`	否	与 Anthropic 官方一致，例如 `2023-06-01`。直传上游。
`Accept`	否	流式可写 `text/event-stream`

请求体参数

字段	类型	必填	说明
`model`	string	是	Claude 模型 ID，例如 `claude-haiku-4-5-20251001`
`messages`	array	是	对话历史，每项 `role` 为 `user` 或 `assistant`，`content` 为字符串或 block 数组
`max_tokens`	integer	是	最大输出 token 数
`system`	string \| array	否	系统提示，顶层字段（不放进 messages）
`stream`	boolean	否	默认 `false`；为 `true` 时返回 SSE
`temperature`	number	否	0–1
`top_p`	number	否	核采样概率
`top_k`	integer	否	top-k 采样
`stop_sequences`	array	否	停止序列
`tools`	array	否	工具定义，每项含 `name`、`description`、`input_schema`
`tool_choice`	object	否	`{"type":"auto"}` / `{"type":"any"}` / `{"type":"tool","name":"..."}`
`metadata`	object	否	例如 `{"user_id":"..."}`

网关只校验 model 必填，其余字段透传上游。

请求体示例

json

{
  "model": "claude-haiku-4-5-20251001",
  "max_tokens": 1024,
  "system": "你是一个简洁的中文助手。",
  "messages": [
    { "role": "user", "content": "你好，介绍一下你自己。" }
  ]
}

响应示例

json

{
  "id": "msg_xxxxxxxxxxxxxxxxxxxxxxxx",
  "type": "message",
  "role": "assistant",
  "model": "claude-haiku-4-5-20251001",
  "content": [
    { "type": "text", "text": "你好！我是 Claude，一个由 Anthropic 训练的助手。" }
  ],
  "stop_reason": "end_turn",
  "stop_sequence": null,
  "usage": {
    "input_tokens": 24,
    "cache_creation_input_tokens": 0,
    "cache_read_input_tokens": 0,
    "output_tokens": 18
  }
}

工具调用示例

请求里声明 tools：

json

{
  "model": "claude-haiku-4-5-20251001",
  "max_tokens": 1024,
  "tools": [
    {
      "name": "get_weather",
      "description": "查询某个城市的当前天气",
      "input_schema": {
        "type": "object",
        "properties": {
          "city": { "type": "string" }
        },
        "required": ["city"]
      }
    }
  ],
  "messages": [
    { "role": "user", "content": "上海现在天气怎么样？" }
  ]
}

模型决定调用工具时，响应 content 含 tool_use block：

json

{
  "id": "msg_xxxxxxxxxxxxxxxxxxxxxxxx",
  "type": "message",
  "role": "assistant",
  "model": "claude-haiku-4-5-20251001",
  "content": [
    {
      "type": "tool_use",
      "id": "toolu_xxxxxxxxxxxxxxxx",
      "name": "get_weather",
      "input": { "city": "上海" }
    }
  ],
  "stop_reason": "tool_use",
  "usage": { "input_tokens": 312, "output_tokens": 47 }
}

把工具结果以 tool_result block 写回 messages 继续对话：

json

{
  "model": "claude-haiku-4-5-20251001",
  "max_tokens": 1024,
  "messages": [
    { "role": "user", "content": "上海现在天气怎么样？" },
    {
      "role": "assistant",
      "content": [
        {
          "type": "tool_use",
          "id": "toolu_xxxxxxxxxxxxxxxx",
          "name": "get_weather",
          "input": { "city": "上海" }
        }
      ]
    },
    {
      "role": "user",
      "content": [
        {
          "type": "tool_result",
          "tool_use_id": "toolu_xxxxxxxxxxxxxxxx",
          "content": "晴，26°C"
        }
      ]
    }
  ]
}

流式响应示例

stream: true 时返回 SSE，事件类型固定。最简骨架：

event: message_start
data: {"type":"message_start","message":{"id":"msg_xxx","type":"message","role":"assistant","model":"claude-haiku-4-5-20251001","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":24,"output_tokens":0}}}

event: content_block_start
data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"你好"}}

event: content_block_stop
data: {"type":"content_block_stop","index":0}

event: message_delta
data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":24,"output_tokens":18}}

event: message_stop
data: {"type":"message_stop"}

错误

400 invalid_request_error — 缺 model / 请求体非 JSON / 解析失败
401 authentication_error — API Key 无效
403 billing_error — 余额或订阅限额不足
403 permission_error — 分组限制 Claude Code 客户端时拒绝其他客户端
429 rate_limit_error — 触发并发或速率限制
502 upstream_error — 上游连续失败，请稍后重试

更多错误码见错误处理。

代码示例

bash

curl https://api.sakrylle.com/v1/messages \
  -H "Authorization: Bearer sk-xxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "claude-haiku-4-5-20251001",
    "max_tokens": 1024,
    "messages": [
      {"role": "user", "content": "你好"}
    ]
  }'

python

from anthropic import Anthropic

client = Anthropic(
    api_key="sk-xxxxxxxxxxxxxxxx",
    base_url="https://api.sakrylle.com",
)

msg = client.messages.create(
    model="claude-haiku-4-5-20251001",
    max_tokens=1024,
    system="你是一个简洁的中文助手。",
    messages=[{"role": "user", "content": "你好"}],
)
print(msg.content[0].text)

javascript

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({
  apiKey: "sk-xxxxxxxxxxxxxxxx",
  baseURL: "https://api.sakrylle.com",
});

const msg = await client.messages.create({
  model: "claude-haiku-4-5-20251001",
  max_tokens: 1024,
  messages: [{ role: "user", content: "你好" }],
});
console.log(msg.content[0].text);

package main

import (
    "context"
    "fmt"

    "github.com/anthropics/anthropic-sdk-go"
    "github.com/anthropics/anthropic-sdk-go/option"
)

func main() {
    client := anthropic.NewClient(
        option.WithAPIKey("sk-xxxxxxxxxxxxxxxx"),
        option.WithBaseURL("https://api.sakrylle.com/"),
    )
    msg, err := client.Messages.New(context.Background(), anthropic.MessageNewParams{
        Model:     anthropic.F(anthropic.Model("claude-haiku-4-5-20251001")),
        MaxTokens: anthropic.F(int64(1024)),
        Messages: anthropic.F([]anthropic.MessageParam{
            anthropic.NewUserMessage(anthropic.NewTextBlock("你好")),
        }),
    })
    if err != nil {
        panic(err)
    }
    fmt.Println(msg.Content[0].Text)
}

rust

use reqwest::Client;
use serde_json::json;

#[tokio::main]
async fn main() -> anyhow::Result<()> {
    let client = Client::new();
    let resp: serde_json::Value = client
        .post("https://api.sakrylle.com/v1/messages")
        .bearer_auth("sk-xxxxxxxxxxxxxxxx")
        .header("anthropic-version", "2023-06-01")
        .json(&json!({
            "model": "claude-haiku-4-5-20251001",
            "max_tokens": 1024,
            "messages": [{"role": "user", "content": "你好"}]
        }))
        .send()
        .await?
        .json()
        .await?;
    println!("{resp:#}");
    Ok(())
}

java

import java.net.URI;
import java.net.http.*;

public class Messages {
  public static void main(String[] args) throws Exception {
    String body = """
      {"model":"claude-haiku-4-5-20251001","max_tokens":1024,
       "messages":[{"role":"user","content":"你好"}]}
      """;
    HttpRequest req = HttpRequest.newBuilder()
      .uri(URI.create("https://api.sakrylle.com/v1/messages"))
      .header("Authorization", "Bearer sk-xxxxxxxxxxxxxxxx")
      .header("Content-Type", "application/json")
      .header("anthropic-version", "2023-06-01")
      .POST(HttpRequest.BodyPublishers.ofString(body))
      .build();
    HttpResponse<String> resp = HttpClient.newHttpClient()
      .send(req, HttpResponse.BodyHandlers.ofString());
    System.out.println(resp.body());
  }
}

csharp

using System.Net.Http.Headers;
using System.Text;

var http = new HttpClient();
http.DefaultRequestHeaders.Authorization =
    new AuthenticationHeaderValue("Bearer", "sk-xxxxxxxxxxxxxxxx");
http.DefaultRequestHeaders.Add("anthropic-version", "2023-06-01");

var payload = """
  {"model":"claude-haiku-4-5-20251001","max_tokens":1024,
   "messages":[{"role":"user","content":"你好"}]}
  """;
var resp = await http.PostAsync(
    "https://api.sakrylle.com/v1/messages",
    new StringContent(payload, Encoding.UTF8, "application/json"));
Console.WriteLine(await resp.Content.ReadAsStringAsync());

Messages ​

Endpoint ​

请求头 ​

请求体参数 ​

请求体示例 ​

响应示例 ​

工具调用示例 ​

流式响应示例 ​

错误 ​

代码示例 ​

Messages

Endpoint

请求头

请求体参数

请求体示例

响应示例

工具调用示例

流式响应示例

错误

代码示例