OpenAI Model Migration (Relocation Plan)

This guide explains how to quickly migrate from OpenAI models to ZhipuAI, using examples to help you complete the migration process. Additionally, ZhipuAI offers a “Special Relocation Plan” for OpenAI API users, making it easy to switch to domestic large models. Get the benefits now

	Relocation Plan
Special Benefits	150 million Tokens (50 million GLM-4 and 100 million GLM-4-Air); Series migration training from OpenAI to GLM;
High Usage Benefits	OpenAI-equivalent Token giveaway plan (no upper limit); OpenAI-equivalent concurrency scale; Matching high-level membership policies, up to 60% off; Exclusive relocation consultants and 5-person/day technical expert support; Assistance with filing and training.

Basic Toolkit

The OpenAI SDK provides a ready-to-use tool for calling, and we have backend-compatible with all OpenAI endpoints, offering a convenient migration method. You only need to replace the api_key and base_url to use our models.

Switch API Endpoint

from openai import OpenAI 
 
client = OpenAI(
    api_key="your zhipuai api key",
    base_url="https://open.bigmodel.cn/api/paas/v4/"
) 
 
response = client.chat.completions.create(
    model="glm-4",  
    messages=[    
        {"role": "system", "content": "You are a smart and creative novel writer"},    
        {"role": "user", "content": "Please write a short fairy tale as the king of fairy tales."} 
    ],
    top_p=0.7,
    temperature=0.9
 ) 
 
 print(response.choices[0].message)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

Use ZhipuAI Official SDK

Some features of ZhipuAI require you to call through the official SDK, which you can install via pypi.

pip install zhipuai

Example Calls

from zhipuai import ZhipuAI
client = ZhipuAI(api_key="") # Fill in your own APIKey
response = client.chat.completions.create(
    model="glm-4",  # Fill in the model name you need to call
    messages=[
        {"role": "user", "content": "As a marketing expert, please create an attractive slogan for the ZhipuAI open platform"},
        {"role": "assistant", "content": "Of course, to create an attractive slogan, please tell me some information about your product"},
        {"role": "user", "content": "ZhipuAI Open Platform"},
        {"role": "assistant", "content": "Intelligence for the future, infinite drawing - ZhipuAI, making innovation within reach!"},
        {"role": "user", "content": "Create a more precise and attractive slogan"}
    ],
)
print(response.choices[0].message)

1
2
3
4
5
6
7
8
9
10
11
12
13

Open Source Frameworks

In terms of agent and related task framework migration, we support the quick migration of the langchain framework. We provide the langchain-zhipuai extension tool, offering support adaptation for BaseChatModel and Embeddings. LangChain users can quickly complete the migration using the following examples:

Install langchain-zhipuai

Download link: https://github.com/MetaGLM/langchain-zhipuai/releases

Before using, please set the environment variable ZHIPUAI_API_KEY to your ZhipuAI API Key.

Migrate to ChatZhipuAI

from langchain_zhipuai.agents.zhipuai_all_tools.base import _get_assistants_tool
from langchain_zhipuai.chat_models import ChatZhipuAI
from langchain.agents import tool
from langchain.tools.shell import ShellTool
from pydantic.v1 import BaseModel, Extra, Field
from langchain import hub
from langchain_zhipuai.agents.all_tools_bind.base import create_zhipuai_tools_agent
from langchain_zhipuai.agent_toolkits import BaseToolOutput
from langchain_zhipuai.agents.all_tools_agent import ZhipuAiAllToolsAgentExecutor
 
 
@tool
def shell(query: str = Field(description="The command to execute")):
    """Use Shell to execute system shell commands"""
    tool = ShellTool()
    return BaseToolOutput(tool.run(tool_input=query))
 
llm = ChatZhipuAI(api_key="") # You can specify the apikey here
 
tools = [
    _get_assistants_tool(shell),
    {"type": "code_interpreter", "code_interpreter": {"sandbox": "none"}},
    {"type": "web_browser"},
    {"type": "drawing_tool"},
]
llm_with_all_tools = llm.bind(
    tools=tools
)
 
prompt = hub.pull("zhipuai-all-tools-chat/zhipuai-all-tools-agent")
agent = create_zhipuai_tools_agent(
    prompt=prompt, llm_with_all_tools=llm_with_all_tools
)
 
agent_executor = ZhipuAiAllToolsAgentExecutor(
    agent=agent,
    tools=[shell] ,
    verbose=True,
    return_intermediate_steps=True,
)
 
agent_executor.invoke(
    {
        "input": "Hello",
        "chat_history": [],
    }
)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47

Compatible with Vector Databases

We provide a compatible Embedding calling method with OpenAI. When using Embeddings related to vector libraries, you only need to replace the Embedding instance with ZhipuAIEmbeddings.

from langchain_zhipuai.embeddings.base import ZhipuAIEmbeddings
 
"""Test zhipuai embeddings."""
documents = ["foo bar"]
embedding = ZhipuAIEmbeddings()
output = embedding.embed_documents(documents)
# len(output) == 1
# len(output[0]) == 1024

1
2
3
4
5
6
7
8