1 接口描述

2 请求参数

2.1 请求头参数

2.2 请求参数

请求参数示例

{
  "model": "gpt-4o-mini",
  "stream": false,
  "messages": [
    {
      "role": "user",
      "content": "Hello!"
    }
  ]
}

3、请求返回

3.1 非流式返回

3.1.1 非流式正常返回

返回结果示例

{
  "id": "chatcmpl-123",
  "object": "chat.completion",
  "created": 1677652288,
  "model": "gpt-4o-mini",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "\n\nHello there, how may I assist you today?"
    },
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 9,
    "completion_tokens": 12,
    "total_tokens": 21
  }
}

3.1.2 非流式异常返回

非流式异常返回时：

• http code 返回非200。
• http body 中返回 error 结构，error结构中包含code、type、message、param等信息，具体可见OpenAI接口文档中的error结构描述及错误码部分介绍。

错误结果示例

{
  "error": {
    "message": "You exceeded your current quota, please check your plan and billing details.",
    "type": "rate_limit_error",
    "param": null,
    "code": "rate_limit_exceeded"
  }
}

3.2 流式返回

3.2.1 流式正常返回

返回结果示例

{"id":"chatcmpl-123","object":"chat.completion.chunk","created":1694268190,"model":"gpt-4o-mini", "choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}
{"id":"chatcmpl-123","object":"chat.completion.chunk","created":1694268190,"model":"gpt-4o-mini", "choices":[{"index":0,"delta":{"content":"Hello"},"finish_reason":null}]}
....
{"id":"chatcmpl-123","object":"chat.completion.chunk","created":1694268190,"model":"gpt-4o-mini", "choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

​**stream_options.include_usage 为 True 时多返回一条包含usage流式消息**​。

{"id":"chatcmpl-123","object":"chat.completion.chunk","created":1694268199,"model":"gpt-4o-mini", "choices":[], "usage": {"prompt_tokens": 9,"completion_tokens": 120,"total_tokens": 129}}

3.2.2 流式异常返回

流式异常分为两种：

• 如果在流式请求接收处理之前发生了异常，如鉴权、参数校验等问题，与普通的非流式一样返回http code，并带有error结构。
• 如果在流式请求已经接收，会先对外返回流式请求连接建立的信息，此时http code为200，而在后续模型流式返回过程中发生了异常，会在流失返回的chunk返回异常信息，并终止当前的流式请求。

4、请求示例代码

4.1 curl方式请求

curl --request POST \
  --url $OPEN_AI_HOST/v1/chat/completions \
  --header 'Accept: */*' \
  --header 'Accept-Encoding: gzip, deflate, br' \
  --header 'Authorization: Bearer $OPEN_API_KEY' \
  --header 'Content-Type: application/json' \
  --data '{
	"model": "gpt-4o-mini",
	"messages": [
		{
			"role": "user",
			"content": "Hello"
		}
	]
}'

4.2 python方式请求

import json
import requests
URL = "$OPEN_AI_HOST/v1/chat/completions"
headers = {
    "Authorization": "Bearer $OPEN_API_KEY",
    "Content-Type": "application/json"
}
data = {
    "model": "gpt-4o-mini",
    "messages": [
        {"role": "user", "content": "Hello"}
    ]
    ,"stream": True
}
try:
    response = requests.post(URL, json=data, stream=True)
    if response.status_code != 200:
        print(response.json())
    else:
        # 流式
        for line in response.iter_lines(chunk_size=8192, decode_unicode=True):
            // 处理请求
        # 非流式直接处理
        message = response.json()["choices"][0]["message"]["content"]
except Exception as e:
    print(f"Exception: {e}")

4.3 openai python客户端示例代码

import openai
from openai import OpenAI
client = OpenAI(base_url="$OPEN_AI_HOST/v1", api_key="$OPEN_API_KEY")
messages = [
    {"role": "user", "content": "Hello"}
]
try:
    stream = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=messages,
        stream=True 
    )
    # 流式
    for chunk in stream:
        print(chunk.choices[0].delta.content or "", end="", flush=True)
except openai.APIStatusError as e:
    print(f"APIStatusError: {e.status_code}, {e.message}, {e.body}")
except openai.APIError as e:
    print(f"APIError: {e.body}")
except Exception as e:
    print(f"Exception: {e}")

5 参考文档

• OpenAI 官网API接口：/docs/api-reference/introduction
• OpenAI 官网文档：/api-reference/chat
• GitHub openai cookbook：/openai/openai-cookbook