国内大模型调用实战

国内大模型API调用实战

今天我们就来具体看下具体大模型厂商API的使用。

1.API

• deepseek
  deepseek支持requests直接调用和openai式的调用，新注册用户有500万token

    import requests
    import json
  
    url = "https://api.deepseek.com/chat/completions"
  
    payload = json.dumps({
       
    "messages": [
        {
       
        "content": "You are a helpful assistant",
        "role": "system"
        },
        {
       
        "content": "Hi",
        "role": "user"
        }
    ],
    "model": "deepseek-coder",
    "frequency_penalty": 0,
    "max_tokens": 2048,
    "presence_penalty": 0,
    "stop": None,
    "stream": False,
    "temperature": 1,
    "top_p": 1,
    "logprobs": False,
    "top_logprobs": None
    })
    headers = {
       
    'Content-Type': 'application/json',
    'Accept': 'application/json',
    'Authorization': 'Bearer <key>'
    }
    response = requests.request("POST", url, headers=headers, data=payload)
  
    print(response.text)
  
    from openai import OpenAI
    # for backward compatibility, you can still use `https://api.deepseek.com/v1` as `base_url`.
    client = OpenAI(api_key="sk-4f510fa4ca9d4ea7b56cdca99418fd9d", base_url="https://api.deepseek.com")
  
    response = client.chat.completions.create(
        model="deepseek-chat",
        messages=[
            {
       "role": "system", "content": "You are a helpful assistant"},
            {
       "role": "user", "content": "Hello"},
    ],
        max_tokens=1024,
        temperature=0.7,
        stream=False
    )
  
    print(response.choices[0].message.content)
  

• 智普AI
  需要安装zhipuai包，新注册用户有赠送token

  • pip install zhipuai

  • Python>=3.7

    from zhipuai import ZhipuAI
    client = ZhipuAI(api_key="dd") # 请填写您自己的APIKey
    response = client.chat.completions.create(
    model="glm-4",  # 填写需要调用的模型名称
        messages=[
            {
           "role": "user", "content": "你好！你叫什么名字"},
        ],
        stream=True,
        )
    for chunk in response:
        print(chunk.choices[0].delta)
    

• kimi
  kimi也是支持openai样式的调用，新注册用户的15元的赠送

  • pip install --upgrade 'openai>=1.0'

    from openai import OpenAI
    
    client = OpenAI(
        api_key = "sk-KbwNWKBaRwhkvmfIHeaV4xnlpp3haV7qLr0sSU7wIOI9ZSNh",
        base_url = "https://api.moonshot.cn/v1",
    )
    
    completion = client.chat.completions.create(
        model = "moonshot-v1-8k",
        messages = [
            {
           "role": "system", "content": "你是 Kimi，由 Moonshot AI 提供的人工智能助手，你更擅长中文和英文的对话。你会为用户提供安全，有帮助，准确的回答。同时，你会拒绝一切涉及恐怖主义，种族歧视，黄色暴力等问题的回答。Moonshot AI 为专有名词，不可翻译成其他语言。"},
            {
           "role": "user", "content": "你好，我叫李雷，1+1等于多少？"}
        ],
        temperature = 0.3,
    )
    
    print(completion.choices[0].message.content)
    

• 字节豆包
  豆包需要安装方舟包

  • pip install 'volcengine-python-sdk[ark]'

    from volcenginesdkarkruntime import Ark
    
    # Authentication
    # 1.If you authorize your endpoint using an API key, you can set your api key to environment variable "ARK_API_KEY"
    # or specify api key by Ark(api_key="${YOUR_API_KEY}").
    # Note: If you use an API key, this API key will not be refreshed.
    # To prevent the API from expiring and failing after some time, choose an API key with no expiration date.
    
    # 2.If you authorize your endpoint with Volcengine Identity and Access Management（IAM), set your api key to environment variable "VOLC_ACCESSKEY", "VOLC_SECRETKEY"
    # or specify ak&sk by Ark(ak="${YOUR_AK}", sk="${YOUR_SK}").
    # To get your ak&sk, please refer to this document([https://www.volcengine.com/docs/6291/65568](https://www.volcengine.com/docs/6291/65568))
    # For more information，please check this document（[https://www.volcengine.com/docs/82379/1263279](https://www.volcengine.com/docs/82379/1263279)）
    client = Ark()
    
    # Non-streaming:
    print("----- standard request -----")
    completion = client.chat.completions.create(
      model="${YOUR_ENDPOINT_ID}",
      messages = [
          {
           "role": "system", "content": "你是豆包，是由字节跳动开发的 AI 人工智能助手"},
          {
           "role": "user", "content": "常见的十字花科植物有哪些？"},
      ],
    )
    print(completion.choices[0].message.content)
    
    # Streaming:
    print("----- streaming request -----")
    stream = client.chat.completions.create(
      model="${YOUR_ENDPOINT_ID}",
      messages = [
          {
           "role": "system", "content": "你是豆包，是由字节跳动开发的 AI 人工智能助手"},
          {
           "role": "user", "content": "常见的十字花科植物有哪些？"},
      ],
      stream=True
    )
    for chunk in stream:
      if not chunk.choices:
          continue
      print(chunk.choices[0].delta.content, end="")
    print()
    

• 讯飞星火
  讯飞星火新注册用户都有免费额度，不同模型额度不要，3.5有10万token， 2.0的有200万一年token. 接口是采用websocket

  • pip install --upgrade spark_ai_python

  • python 3.8

    from sparkai.llm.llm import ChatSparkLLM, ChunkPrintHandler
    from sparkai.core.messages import ChatMessage
    
    #星火认知大模型Spark3.5 Max的URL值，其他版本大模型URL值请前往文档（https://www.xfyun.cn/doc/spark/Web.html）查看
    SPARKAI_URL = 'wss://spark-api.xf-yun.com/v3.5/chat'
    #星火认知大模型调用秘钥信息，请前往讯飞开放平台控制台（https://console.xfyun.cn/services/bm35）查看
    SPARKAI_APP_ID = ''
    SPARKAI_API_SECRET = ''
    SPARKAI_API_KEY = ''
    #星火认知大模型Spark3.5 Max的domain值，其他版本大模型domain值请前往文档（https://www.xfyun.cn/doc/spark/Web.html）查看
    SPARKAI_DOMAIN = 'generalv3.5'
    
    if __name__ == '__main__':
      spark = ChatSparkLLM(
          spark_api_url=SPARKAI_URL,
          spark_app_id=SPARKAI_APP_ID,
          spark_api_key=SPARKAI_API_KEY,
          spark_api_secret=SPARKAI_API_SECRET,
          spark_llm_domain=SPARKAI_DOMAIN,
          streaming=False,
      )
      messages = [ChatMessage(
          role="user",
          content='你好呀'
      )]
      handler = ChunkPrintHandler()
      a = spark.generate([messages], callbacks=[handler])
      print(a)
    

2. API使用总结

可以知道上面的API使用大同小异，有些直接兼容openai的调用格式：

• 需要注册获取api key，进行授权
• api接口基本都支持stream流式输出
• api接口都是典型chat模式，提供不同的角色信息，system，user和assistant， 多轮对话在history中拼接会话信息

大家快去使用吧。