跳到主要內容
API 參考
API 參考

Anthropic Messages

透過 RunAPI 相容 Anthropic 的 Messages 操作傳送請求。

檢視 Markdown
01

概覽

使用 RunAPI 的與 Anthropic 相容的 Messages 操作，搭配相容模型及該協定的請求與回應格式。

快速入門

  1. 建立 API 金鑰並將其設定為 RUNAPI_API_KEY。
  2. 選擇相容的模型，然後傳送文件中指定的路徑參數與 JSON 本文。
  3. 讀取此操作顯示的 JSON 回應或伺服器傳送事件，並處理協定錯誤。

端點

POST /v1/messages
基礎 URL
https://runapi.ai
合約版本
v1
建議的身分驗證方式
x-api-key: YOUR_API_TOKEN
02

身分驗證載體

RunAPI 接受以下列出的每種傳輸協定。產生的範例使用該協定偏好的標頭。

header
Authorization: Bearer YOUR_API_TOKEN
header - 建議
x-api-key: YOUR_API_TOKEN
header
x-goog-api-key: YOUR_API_TOKEN
query
?key=YOUR_API_TOKEN
03

相容模型

開啟模型頁面以查看目前的定價、速率限制及商業用途詳情。

顯示 58 個相容模型
04

請求合約

JSON 請求內文

僅列出同時具備公開協定與 RunAPI 支援依據的欄位。其他 JSON 本文欄位將直接傳遞。

JSON 主體10 個欄位
max_tokensinteger
必填

停止前最多生成的 token 數。

messagesarray
必填

Anthropic Messages 格式的對話輪次。

messages[].content["string", "array"]
必填
messages[].rolestring
必填
modelstring
必填

由相容模型目錄回傳的 RunAPI 模型識別碼。

streamboolean
選填

以伺服器傳送事件的形式回傳具型別的 Messages 事件。

預設值: false
system["string", "array"]
選填

與對話訊息分開提供的系統指令。

temperaturenumber
選填

取樣溫度；模型支援程度可能有所不同。

toolsarray
選填

模型可呼叫的工具。

top_pnumber
選填

Nucleus 取樣機率；模型支援程度可能有所不同。

請求範例

JSON 
{
  "max_tokens": 64,
  "messages": [
    {
      "content": "Summarize one practical improvement for an API deployment.",
      "role": "user"
    }
  ],
  "model": "claude-sonnet-4-6"
}
05

同步回應

HTTP 200

結構描述

JSON 
{
  "additionalProperties": true,
  "properties": {
    "content": {
      "type": "array"
    },
    "id": {
      "type": "string"
    },
    "model": {
      "type": "string"
    },
    "role": {
      "const": "assistant",
      "type": "string"
    },
    "stop_reason": {
      "type": [
        "string",
        "null"
      ]
    },
    "type": {
      "const": "message",
      "type": "string"
    },
    "usage": {
      "additionalProperties": true,
      "properties": {
        "input_tokens": {
          "type": "integer"
        },
        "output_tokens": {
          "type": "integer"
        }
      },
      "required": [
        "input_tokens",
        "output_tokens"
      ],
      "type": "object"
    }
  },
  "required": [
    "id",
    "type",
    "role",
    "content",
    "model",
    "stop_reason",
    "usage"
  ],
  "type": "object"
}

範例

JSON 
{
  "content": [
    {
      "text": "Hello! How can I help today?",
      "type": "text"
    }
  ],
  "id": "msg_example",
  "model": "claude-sonnet-4-6",
  "role": "assistant",
  "stop_reason": "end_turn",
  "stop_sequence": null,
  "type": "message",
  "usage": {
    "input_tokens": 12,
    "output_tokens": 8
  }
}
06

SSE 事件：message_start

text/event-stream

結構描述

JSON 
{
  "additionalProperties": true,
  "properties": {
    "message": {
      "type": "object"
    },
    "type": {
      "const": "message_start",
      "type": "string"
    }
  },
  "required": [
    "type",
    "message"
  ],
  "type": "object"
}

範例

JSON 
{
  "message": {
    "content": [],
    "id": "msg_example",
    "model": "claude-sonnet-4-6",
    "role": "assistant",
    "stop_reason": null,
    "stop_sequence": null,
    "type": "message",
    "usage": {
      "input_tokens": 12,
      "output_tokens": 1
    }
  },
  "type": "message_start"
}
07

SSE 事件：content_block_delta

text/event-stream

結構描述

JSON 
{
  "additionalProperties": true,
  "properties": {
    "delta": {
      "type": "object"
    },
    "index": {
      "type": "integer"
    },
    "type": {
      "const": "content_block_delta",
      "type": "string"
    }
  },
  "required": [
    "type",
    "index",
    "delta"
  ],
  "type": "object"
}

範例

JSON 
{
  "delta": {
    "text": "Hello",
    "type": "text_delta"
  },
  "index": 0,
  "type": "content_block_delta"
}
08

SSE 事件：message_delta

text/event-stream

結構描述

JSON 
{
  "additionalProperties": true,
  "properties": {
    "delta": {
      "type": "object"
    },
    "type": {
      "const": "message_delta",
      "type": "string"
    },
    "usage": {
      "additionalProperties": true,
      "properties": {
        "output_tokens": {
          "type": "integer"
        }
      },
      "required": [
        "output_tokens"
      ],
      "type": "object"
    }
  },
  "required": [
    "type",
    "delta",
    "usage"
  ],
  "type": "object"
}

範例

JSON 
{
  "delta": {
    "stop_reason": "end_turn",
    "stop_sequence": null
  },
  "type": "message_delta",
  "usage": {
    "output_tokens": 8
  }
}
09

SSE 事件：message_stop

text/event-stream

結構描述

JSON 
{
  "additionalProperties": true,
  "properties": {
    "type": {
      "const": "message_stop",
      "type": "string"
    }
  },
  "required": [
    "type"
  ],
  "type": "object"
}

範例

JSON 
{
  "type": "message_stop"
}
10

協定錯誤：invalid_request

HTTP 400

結構描述

JSON 
{
  "additionalProperties": true,
  "properties": {
    "error": {
      "additionalProperties": true,
      "properties": {
        "message": {
          "type": "string"
        },
        "type": {
          "type": "string"
        }
      },
      "required": [
        "type",
        "message"
      ],
      "type": "object"
    },
    "request_id": {
      "type": "string"
    },
    "type": {
      "const": "error",
      "type": "string"
    }
  },
  "required": [
    "type",
    "error"
  ],
  "type": "object"
}

範例

JSON 
{
  "error": {
    "message": "max_tokens is required",
    "type": "invalid_request_error"
  },
  "request_id": "req_example",
  "type": "error"
}
11

用量

結構描述

JSON 
{
  "additionalProperties": true,
  "properties": {
    "input_tokens": {
      "type": "integer"
    },
    "output_tokens": {
      "type": "integer"
    }
  },
  "required": [
    "input_tokens",
    "output_tokens"
  ],
  "type": "object"
}

範例

JSON 
{
  "input_tokens": 12,
  "output_tokens": 8
}
12

產生的程式碼範例

每個 cURL、JavaScript 和 Python 範例均會從此契約傳送經驗證的請求，無需特定協定的 SDK。

安裝

Shell 
pip install requests
PYTHON 
import requests

response = requests.post(
    "https://runapi.ai/v1/messages",
    headers={"x-api-key": "YOUR_API_TOKEN", "Content-Type": "application/json"},
    json={"model": "claude-sonnet-4-6", "max_tokens": 64, "messages": [{"role": "user", "content": "Summarize one practical improvement for an API deployment."}]}
)
response.raise_for_status()
print(response.json())