ZhipuAI Python SDK

You can configure the ZhipuAI client using environment variables or directly in your code.

Environment Variables:

• ZHIPUAI_API_KEY: Your API key.
• ZHIPUAI_BASE_URL: Custom API endpoint (optional, defaults to https://open.bigmodel.cn/api/paas/v4/).

Code Configuration: Pass the api_key to the ZhipuAI constructor. For advanced settings, you can also configure timeout, max_retries, and base_url.

import httpx
from zhipuai import ZhipuAI

client = ZhipuAI(
    api_key="your_api_key_here",
    timeout=httpx.Timeout(timeout=300.0, connect=8.0),
    max_retries=3,
    base_url="https://open.bigmodel.cn/api/paas/v4/"
)

You can initialize the ZhipuAI client using either environment variables or direct code configuration.

Environment Variables:

• ZHIPUAI_API_KEY: Your API key.
• ZHIPUAI_BASE_URL: (Optional) The API endpoint, e.g., https://open.bigmodel.cn/api/paas/v4/.

Code Configuration: Pass api_key and an optional base_url directly to the constructor.

from zhipuai import ZhipuAI

# Using explicit configuration
client = ZhipuAI(
    api_key="your-api-key",
    base_url="https://open.bigmodel.cn/api/paas/v4/"
)

Install the SDK using pip.

Requirements:

• Python 3.9 or higher
• pip package manager

Core Dependencies:

• httpx: HTTP client
• pydantic: Data validation and serialization
• typing-extensions: Type annotation extensions

pip install zhipuai

Customize client behavior such as timeouts and retries using httpx.Timeout and the max_retries parameter.

from zhipuai import ZhipuAI
import httpx

client = ZhipuAI(
    api_key="your-api-key",
    timeout=httpx.Timeout(timeout=300.0, connect=8.0),  # Request timeout
    max_retries=3,  # Retry attempts
    base_url="https://open.bigmodel.cn/api/paas/v4/"
)

Use client.videos.generations to create videos with models like cogvideox-2.

Key Parameters:

• prompt: The text description of the video.
• quality: 'quality' (priority on quality) or 'speed' (priority on speed).
• with_audio: Boolean to include background audio.
• size: Resolution (e.g., "1920x1080").
• fps: Frame rate (e.g., 30 or 60).

After initiating generation, use client.videos.retrieve_videos_result(id=response.id) to check the status and retrieve the result.

from zhipuai import ZhipuAI

client = ZhipuAI(api_key="your-api-key")
response = client.videos.generations(
    model="cogvideox-2",
    prompt="一个美丽的日落海滩场景",
    quality="quality",
    with_audio=True,
    size="1920x1080",
    fps=30,
    user_id="user_12345"
)

result = client.videos.retrieve_videos_result(id=response.id)
print(result)

To receive real-time streaming responses, set stream=True in the create method. The response becomes an iterator of chunks.

from zhipuai import ZhipuAI

client = ZhipuAI(api_key="your-api-key")
response = client.chat.completions.create(
    model="glm-4",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Tell me a story about AI."}
    ],
    stream=True
)

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

Use client.assistant.conversation to interact with a specific assistant. You must provide an assistant_id and a model like glm-4-assistant. This method supports streaming.

from zhipuai import ZhipuAI

client = ZhipuAI(api_key="your-api-key")

response = client.assistant.conversation(
    assistant_id="your_assistant_id",
    model="glm-4-assistant",
    messages=[
        {
            "role": "user",
            "content": [{"type": "text", "text": "帮我搜索智谱AI的最新产品信息"}]
        }
    ],
    stream=True
)

for chunk in response:
    print(chunk)

For role-playing models like charglm-3, use the meta parameter to define user_info, bot_info, bot_name, and user_name to shape the conversation context.

from zhipuai import ZhipuAI

client = ZhipuAI(api_key="your-api-key")
response = client.chat.completions.create(
    model="charglm-3",
    messages=[
        {
            "role": "user",
            "content": "你好，最近在忙什么呢？"
        }
    ],
    meta={
        "user_info": "我是一位电影导演，擅长拍摄音乐题材的电影。",
        "bot_info": "你是一位当红的国内女歌手及演员，拥有出众的音乐才华。",
        "bot_name": "小雅",
        "user_name": "导演"
    },
)
print(response.choices[0].message.content)

To receive responses in real-time, set stream=True in the client.chat.completions.create method. The response becomes an iterator that yields chunks of data.

from zhipuai import ZhipuAI

client = ZhipuAI(api_key="your-api-key")
response = client.chat.completions.create(
    model="glm-4",
    messages=[
        {"role": "system", "content": "你是一个人工智能助手，你叫ChatGLM"},
        {"role": "user", "content": "你好！你叫什么名字"},
    ],
    stream=True,
)

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

Use client.chat.completions.create to interact with text models like glm-4. You can provide a list of messages with roles (e.g., user, system) and use extra_body for parameters like temperature or max_tokens. You can also enable tools like web_search.

from zhipuai import ZhipuAI

client = ZhipuAI(api_key="your-api-key")
response = client.chat.completions.create(
    model="glm-4",
    messages=[
        {"role": "user", "content": "你好，请介绍一下智谱AI"}
    ],
    tools=[
        {
            "type": "web_search",
            "web_search": {
                "search_query": "Search the Zhipu",
                "search_result": True,
            }
        }
    ],
    extra_body={"temperature": 0.5, "max_tokens": 50}
)
print(response.choices[0].message.content)