文本补全接口

创建文本补全请求，适用于单轮文本生成任务。

接口信息

  • 接口地址: POST /v1/completions
  • 认证方式: Bearer Token
  • Content-Type: application/json

请求参数

必填参数

参数类型说明
modelstring模型名称
promptstring/array提示文本

可选参数

参数类型默认值说明
suffixstringnull文本后缀
max_tokensinteger16最大生成 token 数
temperaturenumber1.0采样温度，范围 0-2
top_pnumber1.0核采样参数
ninteger1生成数量
streambooleanfalse是否流式返回
logprobsintegernull返回的 log 概率数量
echobooleanfalse是否返回 prompt
stopstring/arraynull停止生成的标记
presence_penaltynumber0存在惩罚
frequency_penaltynumber0频率惩罚
best_ofinteger1生成候选项数量
logit_biasobjectnulltoken 偏好设置
userstringnull用户标识

请求示例

基础请求

curl https://aigc.aochengcloud.com/v1/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" \
  -d '{
    "model": "gpt-3.5-turbo-instruct",
    "prompt": "写一首关于春天的诗：",
    "max_tokens": 100
  }'

带参数的请求

curl https://aigc.aochengcloud.com/v1/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" \
  -d '{
    "model": "gpt-3.5-turbo-instruct",
    "prompt": "将以下英文翻译成中文：\n\nHello, how are you?",
    "temperature": 0.3,
    "max_tokens": 50,
    "stop": ["\n\n"]
  }'

流式请求

curl https://aigc.aochengcloud.com/v1/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" \
  -d '{
    "model": "gpt-3.5-turbo-instruct",
    "prompt": "Once upon a time",
    "stream": true,
    "max_tokens": 200
  }'

响应格式

非流式响应

{
  "id": "cmpl-8VYRlGZS4uPpnS3FyW32vM",
  "object": "text_completion",
  "created": 1704067200,
  "model": "gpt-3.5-turbo-instruct",
  "choices": [
    {
      "text": "\n\n春风拂面暖阳斜，\n花开满园香万家。",
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 10,
    "completion_tokens": 20,
    "total_tokens": 30
  }
}

流式响应

data: {"id":"cmpl-8VYRlGZS4uPpnS3FyW32vM","object":"text_completion","created":1704067200,"model":"gpt-3.5-turbo-instruct","choices":[{"text":"\n\n","index":0,"logprobs":null,"finish_reason":null}]}

data: {"id":"cmpl-8VYRlGZS4uPpnS3FyW32vM","object":"text_completion","created":1704067200,"model":"gpt-3.5-turbo-instruct","choices":[{"text":"春","index":0,"logprobs":null,"finish_reason":null}]}

data: {"id":"cmpl-8VYRlGZS4uPpnS3FyW32vM","object":"text_completion","created":1704067200,"model":"gpt-3.5-turbo-instruct","choices":[{"text":"风","index":0,"logprobs":null,"finish_reason":null}]}

data: [DONE]

响应字段说明

字段类型说明
idstring请求的唯一标识
objectstring对象类型
createdinteger创建时间戳
modelstring使用的模型
choicesarray生成的文本列表
usageobjecttoken 用量统计

choices 字段

字段类型说明
textstring生成的文本
indexinteger文本索引
logprobsobjectlog 概率
finish_reasonstring结束原因

SDK 示例

Python

import openai

client = openai.OpenAI(
    api_key="sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx",
    base_url="https://aigc.aochengcloud.com/v1"
)

# 非流式请求
response = client.completions.create(
    model="gpt-3.5-turbo-instruct",
    prompt="写一首关于春天的诗：",
    max_tokens=100
)
print(response.choices[0].text)

# 流式请求
stream = client.completions.create(
    model="gpt-3.5-turbo-instruct",
    prompt="Once upon a time",
    stream=True,
    max_tokens=200
)
for chunk in stream:
    print(chunk.choices[0].text, end="")

Node.js

import OpenAI from 'openai';

const openai = new OpenAI({
  apiKey: 'sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx',
  baseURL: 'https://aigc.aochengcloud.com/v1'
});

// 非流式请求
const response = await openai.completions.create({
  model: 'gpt-3.5-turbo-instruct',
  prompt: '写一首关于春天的诗：',
  max_tokens: 100
});
console.log(response.choices[0].text);

使用场景

文本续写

response = client.completions.create(
    model="gpt-3.5-turbo-instruct",
    prompt="人工智能的发展趋势是",
    max_tokens=200
)

文本翻译

response = client.completions.create(
    model="gpt-3.5-turbo-instruct",
    prompt="将以下英文翻译成中文：\n\nHello, how are you?",
    temperature=0.3,
    max_tokens=50
)

代码生成

response = client.completions.create(
    model="gpt-3.5-turbo-instruct",
    prompt="用Python写一个快速排序函数：",
    max_tokens=300
)

文本摘要

response = client.completions.create(
    model="gpt-3.5-turbo-instruct",
    prompt="请将以下文本总结为一句话：\n\n" + long_text,
    max_tokens=100
)

参数调优

temperature

控制输出的随机性：

  • 0: 确定性输出，适合事实性任务
  • 0.5: 平衡创造性和准确性
  • 1.0: 适中
  • 2.0: 高创造性，可能不准确

max_tokens

控制生成文本的长度：

  • 短文本：50-100
  • 中等长度：200-500
  • 长文本：1000-2000

stop

设置停止生成的标记：

{
  "stop": ["\n\n", "END"]
}

错误处理

错误码说明处理方式
400请求参数错误检查请求参数
401认证失败检查 API Key
429请求频率超限降低请求频率
500服务器错误稍后重试
最后更新: