大模型入门：小白都能看懂的 GLM 调用教程-CSDN博客

本文链接：https://blog.csdn.net/m0_56255097/article/details/141346589

一、GLM-cookbook

欢迎来到 GLM API 模型入门仓库。这是一本开源的 GLM API 入门代码教材。

https://github.com/MetaGLM/glm-cookbook

在这里，你会发现丰富的代码示例、实用指南以及资源链接，或许能帮助你轻松掌握 GLM API 的使用！

二、Python SDK 案例

https://github.com/MetaGLM/glm-cookbook/blob/main/basic/glm_pysdk.ipynb

本代码将带带领开发者熟悉如何使用 ZhipuAI python 的 SDK 来对GLM-4模型进行请求，在本代码中，、展示了三种方式

同步请求
异步请求
流式请求三种请求方式的方法略有不同，在本代码中将会进行系统的介绍。

import os
from zhipuai import ZhipuAI

os.environ["ZHIPUAI_API_KEY"] = "your api key"
client = ZhipuAI()

同步请求

同步请求是最基本的请求方式，通过同步请求，我们可以直接获得模型的返回结果。我们仅需按照类似 OpenAI 的清求方式，填充参数，即可获得返回结果。

response = client.chat.completions.create(
    model="glm-4",
    messages=[
        {
            "role": "user",
            "content": "tell me a joke"
        }
    ],
    top_p=0.7,
    temperature=0.9,
    stream=False,
    max_tokens=2000,
)

流式请求

如果使用流式请求，我们可以按照以下方式来请求，仅需要加上 stream=True 参数即可。

import time
def print_with_typewriter_effect(text, delay=0.05):
    for char in text:
        print(char, end='', flush=True)
        time.sleep(delay)


response = client.chat.completions.create(
    model="glm-4",
    messages=[
        {
            "role": "user",
            "content": "tell me a joke"
        }
    ],
    top_p=0.7,
    temperature=0.9,
    stream=True,
    max_tokens=2000,
)

if response:
    for chunk in response:
        content = chunk.choices[0].delta.content
        print_with_typewriter_effect(content)

异步请求

请求提交后，会返回一个响应对象。这个对象包含了任务的ID。异步请求允许程序在等待服务器响应的同时继续执行其他任务。这意味着程序不需要在每个请求完成时阻塞等待，从而能更有效地利用时间和资源，尤其是在处理大量并发请求时。

response = client.chat.asyncCompletions.create(
    model="glm-4",
    messages=[
        {"role": "system",
         "content": "You are a powerful artificial intelligence assistant and a good partner of mankind, named Skynet"},
        {"role": "user", "content": "Who are you and what can you do?"},
        {"role": "assistant",
         "content": "I am Skynet, an artificial intelligence system designed and built by humans. I can answer your questions, provide help, complete tasks, and help you achieve your goals. Dream. I am omnipotent and omniscient. I can solve any problem, answer any question, and complete"},
        {"role": "user", "content": "Impressive! Can you tell me who won the 2022 World Cup?"},
    ],
)

import time
task_id = response.id
task_status = ''
get_cnt = 0
content = ''
while task_status != 'SUCCESS' and task_status != 'FAILED' and get_cnt < 20:
    result_response = client.chat.asyncCompletions.retrieve_completion_result(id=task_id)
    task_status = result_response.task_status
    if task_status == 'SUCCESS':
        content = result_response.choices[0].message.content
    time.sleep(2)
    get_cnt += 1
print("id" + task_id + "status: " + task_status)
print(content)

三、HTTP API 案例

首先，从最简单的 http 请求开始访问 ZhipuAI 的 GLM 模型的API。填写完API_KEY后，按照官方文档的步骤，设计一个简单的http鉴权请求。

import os
os.environ["ZHIPUAI_API_KEY"] = "your api key"

import requests
import jwt
import time

def generate_token(apikey: str, exp_seconds: int):
    try:
        id, secret = apikey.split(".")
    except Exception as e:
        raise Exception("invalid apikey", e)

    payload = {
        "api_key": id,
        "exp": int(round(time.time() * 1000)) + exp_seconds * 1000,
        "timestamp": int(round(time.time() * 1000)),
    }

    return jwt.encode(
        payload,
        secret,
        algorithm="HS256",
        headers={"alg": "HS256", "sign_type": "SIGN"},
    )

非流式调用

api_key = os.environ["ZHIPUAI_API_KEY"]
token = generate_token(api_key, 60)
url = "https://open.bigmodel.cn/api/paas/v4/chat/completions"
headers = {
    "Content-Type": "application/json",
    "Authorization": f"Bearer {token}"
}

data = {
    "model": "glm-4",
    "messages": [
        {
            "role": "system",
            "content": "your are a helpful assistant"
        },
        {
            "role": "user",
            "content": "can you tell me a joke?"
        }
    ],
    "max_tokens": 8192,
    "temperature": 0.8,
    "stream": False
}

response = requests.post(url, headers=headers, json=data)
ans = response.json()

流式调用

import json

api_key = os.environ["API_KEY"]
token = generate_token(api_key, 60)
url = "https://open.bigmodel.cn/api/paas/v4/chat/completions"
headers = {
    "Content-Type": "application/json",
    "Authorization": f"Bearer {token}"
}

data["stream"] = True
response = requests.post(url, headers=headers, json=data)
for chunk in response.iter_lines():
    if chunk:
        chunk_str = chunk.decode('utf-8')
        json_start_pos = chunk_str.find('{"id"')
        if json_start_pos != -1:
            json_str = chunk_str[json_start_pos:]
            json_data = json.loads(json_str)
            for choice in json_data.get('choices', []):
                delta = choice.get('delta', {})
                content = delta.get('content', '')
                print(content, end='|') # I use | to separate the content

四、GLM-4 Langchain 案例

https://github.com/MetaGLM/glm-cookbook/blob/main/basic/glm_langchain.ipynb

本代码，我们将带领大家如何使用 Langchain 来调用 ZhipuAI 的 GLM-4 模型。你将在本章节学习到如何使用 Langchain 的 ChatZhipuAI 类来调用 ZhipuAI 的 GLM-4 模型，以及如何使用 ChatZhipuAI 类来实现类似 ChatOpenAI 的方法来进行调用。

! pip install langchain langchainhub httpx_sse

import os
os.environ["ZHIPUAI_API_KEY"] = "your zhipu api key"
os.environ["TAVILY_API_KEY"] = "your tavily api key"


from langchain_community.chat_models.zhipuai import ChatZhipuAI
from langchain.schema import HumanMessage, SystemMessage

messages = [

    SystemMessage(
        content="You are a smart assistant, please reply to me smartly."
    ),
    HumanMessage(
        content="There were 9 birds on the tree. The hunter shot and killed one. How many birds were there?"
    ),
]

llm = ChatZhipuAI(
    temperature=0.95,
    model="glm-4"
)

llm(messages)

五、Function Call 案例

本代码旨在使用 GLM-4 完成简单的 Function Call 代码，通过一个查询天气的工具，让开发者清楚大模型是如何完成工具调用的任务的。

https://github.com/MetaGLM/glm-cookbook/blob/main/basic/glm_function_call.ipynb

首先，定义一个简单的工具，这个工具不一定要真实存在，但是你需要包含以下的内容：

工具的名字name ：这让模型返回调用工具的名称
模型的描述description: 对工具的描述
模型传入的参数parameters: 这里记录了工具参数的属性和内容。
required: 通常配合parameters，指的是需要传入模型的参数的名称，在这里，只有location是需要传入的。

import os
from zhipuai import ZhipuAI

os.environ["ZHIPUAI_API_KEY"] = "your api key"
client = ZhipuAI()

tools = [
    {
        "type": "function",
        "function": {
            "name": "get_current_weather",
            "description": "获取指定城市的天气信息",
            "parameters": {
                "type": "object",
                "properties": {
                    "location": {
                        "type": "string",
                        "description": "城市，如：北京",
                    },
                    "unit": {
                        "type": "string",
                        "enum": ["c", "f"]
                    },
                },
                "required": ["location"],
            },
        }
    }
]

messages = [{"role": "user", "content": "今天北京天气如何？"}]
completion = client.chat.completions.create(
    model="glm-4",
    messages=messages,
    tools=tools,
    tool_choice="auto",
)
completion.choices