openAI API简易使用教程

准备

1. 创建openAI 账号（https://platform.openai.com/overview），右上角personal，创建API key。

2. 安装包

pip install openai
pip install --upgrade tiktoken
1
2

tiktoken 是用来计算每次查询时的token数，因为openAI是根据token数计费，不是必须安装。

API调用

api key 可以直接明文写在代码中，也可以通过环境变量方式获取

import os
import openai

# OPENAI_API_KEY是自己设定的环境变量名
openai.api_key = os.getenv("OPENAI_API_KEY")
# 明文
openai.api_key = *************
1
2
3
4
5
6
7

openAI提供了几种不同场景的模型，主要有text completion、code completion、chat completion、image completion，例如chat completion，则调用方式为

openai.ChatCompletion.create(
  model="gpt-3.5-turbo",
  messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Who won the world series in 2020?"},
        {"role": "assistant", "content": "The Los Angeles Dodgers won the World Series in 2020."},
        {"role": "user", "content": "Where was it played?"}
    ]
)
1
2
3
4
5
6
7
8
9

其中
model 是具体的模型，gpt-3.5-turbo是openAI最先进的语言模型，当然也可以用其他模型。
role，三种固定值，
system：类似一种前提，表示后续的对话以此情景为基础；
user：提问者；
assistant：当对话需要结合上下文时，通过它让模型知道之前的对话内容。

content是具体的对话内容

发起一次请求到响应会存在几秒钟的延迟，response 格式如下所示

{
 'id': 'chatcmpl-6p9XYPYSTTRi0xEviKjjilqrWU2Ve',
 'object': 'chat.completion',
 'created': 1677649420,
 'model': 'gpt-3.5-turbo',
 'usage': {'prompt_tokens': 56, 'completion_tokens': 31, 'total_tokens': 87},
 'choices': [
   {
    'message': {
      'role': 'assistant',
      'content': 'The 2020 World Series was played in Arlington, Texas at the Globe Life Field, which was the new home stadium for the Texas Rangers.'},
    'finish_reason': 'stop',
    'index': 0
   }
  ]
}
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

提取回复内容response['choices'][0]['message']['content']

每次response中不同finish-reason值代表不同状态：

• stop：API返回完整内容
• length：由于max_token限制，回答不完整。
• content_filter：回复被过滤
• null：API还在思考答案

计算token数量

openAI的gpt-3.5-turbo-0301模型token最多4096，超过限制只能缩短请求内容。
而且请求的token和回复的token数会被加一起计费，例如说输入了10个token，openAI回复了20个token，那么最终收费是按照30个token进行收费。

import tiktoken
encoding = tiktoken.encoding_for_model("gpt-3.5-turbo")

def num_tokens_from_string(string: str, encoding_name: str) -> int:
    """Returns the number of tokens in a text string."""
    encoding = tiktoken.get_encoding(encoding_name)
    num_tokens = len(encoding.encode(string))
    return num_tokens

num_tokens_from_string("tiktoken is great!", "cl100k_base")
1
2
3
4
5
6
7
8
9
10

how to count tokens with tiktoken

示例

1. 中文请求

# 中文
Reponse = openai.ChatCompletion.create(
  model="gpt-3.5-turbo",
  messages=[
        {"role": "user", "content": "上海在哪里"},
    ]
)
1
2
3
4
5
6
7

回复

{
  "choices": [
    {
      "finish_reason": "stop",
      "index": 0,
      "message": {
        "content": "\n\n\u4e0a\u6d77\u4f4d\u4e8e\u4e2d\u56fd\u4e1c\u90e8\u6cbf\u6d77\u5730\u5e26\uff0c\u6bd7\u90bb\u6c5f\u82cf\u548c\u6d59\u6c5f\u4e24\u7701\uff0c\u5904\u4e8e\u957f\u6c5f\u53e3\u548c\u676d\u5dde\u6e7e\u4e4b\u95f4\uff0c\u5730\u7406\u5750\u6807\u4e3a31.23\u00b0N, 121.47\u00b0E\u3002",
        "role": "assistant"
      }
    }
  ],
  "created": 1678794854,
  "id": "chatcmpl-6txWIqBPu6GIbaN7fwnosvKzfkLEE",
  "model": "gpt-3.5-turbo-0301",
  "object": "chat.completion",
  "usage": {
    "completion_tokens": 63,
    "prompt_tokens": 13,
    "total_tokens": 76
  }
}
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21

2. 翻译

reponse = openai.ChatCompletion.create(
  model="gpt-3.5-turbo",
  messages=[
  {"role": "system", "content": "You are a helpful assistant that translates English to French."},
  {"role": "user", "content": 'Translate the following English text to French: "{text}"'}
]
)
1
2
3
4
5
6
7

回复

{
  "choices": [
    {
      "finish_reason": "stop",
      "index": 0,
      "message": {
        "content": "Je suis d\u00e9sol\u00e9, je ne peux pas traduire cette demande car il n'y a pas de texte fourni entre les accolades. Veuillez ajouter du texte \u00e0 traduire.",
        "role": "assistant"
      }
    }
  ],
  "created": 1678798886,
  "id": "chatcmpl-6tyZKMOnfUUeRIszSmgovSKvzsNfA",
  "model": "gpt-3.5-turbo-0301",
  "object": "chat.completion",
  "usage": {
    "completion_tokens": 41,
    "prompt_tokens": 34,
    "total_tokens": 75
  }
}
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21

也可以不加system

reponse = openai.ChatCompletion.create(
  model="gpt-3.5-turbo",
  messages=[
  {"role": "user", "content": 'Translate the following English text to French: "{text}"'}
]
)
1
2
3
4
5
6

回复：

{
  "choices": [
    {
      "finish_reason": "stop",
      "index": 0,
      "message": {
        "content": "\n\n\"{text}\" is already in English and does not need to be translated.",
        "role": "assistant"
      }
    }
  ],
  "created": 1678799017,
  "id": "chatcmpl-6tybRxkYc92IE1mXELPuWnw2OFmgE",
  "model": "gpt-3.5-turbo-0301",
  "object": "chat.completion",
  "usage": {
    "completion_tokens": 17,
    "prompt_tokens": 18,
    "total_tokens": 35
  }
}
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21