### Python Example: Text-to-Video Generation with Polling

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/videos.md

This example demonstrates how to start a text-to-video generation and then poll for its completion using `retrieve_videos_result`. It prints the video URL and thumbnail upon success or failure.

```python
from zai import ZaiClient
import time

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

# Start video generation
response = client.videos.generations(
    model="cogvideox-3",
    prompt="A cat is playing with a ball in a sunny garden",
    quality="quality",
    size="1920x1080",
    fps=30
)

task_id = response.id
print(f"Generation started, task_id: {task_id}")

# Poll for completion
while True:
    result = client.videos.retrieve_videos_result(id=task_id)
    print(f"Status: {result.task_status}")
    
    if result.task_status == "SUCCESS":
        for video in result.video_result:
            print(f"Video URL: {video.url}")
            print(f"Thumbnail: {video.cover_image_url}")
        break
    elif result.task_status == "FAIL":
        print("Video generation failed")
        break
    
    time.sleep(10)  # Poll every 10 seconds
```

--------------------------------

### Python Example: Image-to-Video Generation

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/videos.md

This example shows how to initiate an image-to-video generation. It prints the task ID and suggests polling for the result using `retrieve_videos_result`.

```python
response = client.videos.generations(
    model="cogvideox-3",
    image_url="https://example.com/image.jpg",
    quality="speed",  # Faster generation
    size="1280x720",
    fps=30,
    movement_amplitude="medium"
)

print(f"Task ID: {response.id}")
# Poll for result using retrieve_videos_result()
```

--------------------------------

### ZhipuAiClient Initialization Example

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/CLIENT.md

Example demonstrating how to initialize and use the ZhipuAiClient for interactions with the Z.ai API in mainland China.

```APIDOC
## Example

```python
from zai import ZhipuAiClient

# Initialize for mainland China
client = ZhipuAiClient(api_key="your-api-key-here")

# All API resources available same as ZaiClient
response = client.chat.completions.create(
    model="glm-5.2",
    messages=[{"role": "user", "content": "Hello"}]
)
```
```

--------------------------------

### Python Example: High-Quality 4K Video Generation

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/videos.md

This example demonstrates generating a high-quality 4K video with a high frame rate and audio. It prints the task ID for tracking.

```python
response = client.videos.generations(
    model="cogvideox-3",
    prompt="A drone flying over snow-covered mountains at sunrise",
    quality="quality",  # Prioritize quality
    size="3840x2160",  # 4K resolution
    fps=60,  # High frame rate
    with_audio=True
)

task_id = response.id
print(f"High-quality 4K generation started: {task_id}")
```

--------------------------------

### Install Z.ai SDK via pip

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/README.md

Install the Z.ai SDK using pip. Ensure Python 3.8+ and pip are installed.

```bash
pip install zai-sdk
```

--------------------------------

### Python Example: Multiple Images to Video Generation

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/videos.md

This example shows how to generate a video from a list of image URLs. It specifies the model, image URLs, quality, size, and frame rate.

```python
image_urls = [
    "https://example.com/img1.jpg",
    "https://example.com/img2.jpg",
    "https://example.com/img3.jpg"
]

response = client.videos.generations(
    model="cogvideox-3",
    image_url=image_urls,  # Multiple images
    quality="speed",
    size="1024x1024",
    fps=30
)
```

--------------------------------

### Python Example: Video Generation with Tracking and Polling

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/videos.md

This example shows how to initiate video generation with custom `request_id` and `user_id`, and then poll for the result with a timeout. It handles success, failure, and timeout scenarios.

```python
import time

response = client.videos.generations(
    model="cogvideox-3",
    prompt="A sunset over the ocean",
    request_id="video_001",
    user_id="user_12345",
    quality="quality",
    size="1920x1080"
)

task_id = response.id
max_wait = 600  # 10 minutes
start_time = time.time()

# Poll until completion or timeout
while time.time() - start_time < max_wait:
    result = client.videos.retrieve_videos_result(id=task_id)
    
    if result.task_status == "SUCCESS":
        print("✓ Generation complete")
        for video in result.video_result:
            print(f"  URL: {video.url}")
        return result
    elif result.task_status == "FAIL":
        print("✗ Generation failed")
        return None
    
    print(f"  Still processing... ({int(time.time() - start_time)}s)")
    time.sleep(5)

print("✗ Generation timeout")
return None
```

--------------------------------

### Python Example: Different Aspect Ratios for Video Generation

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/videos.md

This example demonstrates generating videos with different aspect ratios: portrait (9:16), landscape (21:9), and square (1:1). It shows how to specify `aspect_ratio` and `size` accordingly.

```python
# Portrait (mobile-friendly)
response1 = client.videos.generations(
    model="cogvideox-3",
    prompt="A person dancing",
    aspect_ratio="9:16",
    size="1080x1920"
)

# Landscape (cinema-style)
response2 = client.videos.generations(
    model="cogvideox-3",
    prompt="Cinematic landscape shot",
    aspect_ratio="21:9",
    size="2520x1080"
)

# Square (social media)
response3 = client.videos.generations(
    model="cogvideox-3",
    prompt="Artistic animation",
    aspect_ratio="1:1",
    size="1024x1024"
)
```

--------------------------------

### Python Example: Fast Generation Mode

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/videos.md

This example prioritizes speed for real-time applications by setting the quality to 'speed' and enabling `off_peak` usage. It's suitable for quick animations.

```python
# Prioritize speed for real-time applications
response = client.videos.generations(
    model="cogvideox-3",
    prompt="Quick animation: a ball bouncing",
    quality="speed",
    size="1280x720",
    fps=30,
    off_peak=True  # Use off-peak slot if available
)
```

--------------------------------

### Resource Management with Context Manager

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/CLIENT.md

Example showing how to use the ZaiClient with a context manager (`with` statement) for automatic resource cleanup.

```APIDOC
### Resource Management

Both clients implement proper cleanup:

```python
with ZaiClient(api_key="your-api-key") as client:
    response = client.chat.completions.create(
        model="glm-5.2",
        messages=[{"role": "user", "content": "Test"}]
    )
    # Client automatically cleaned up after with block
```
```

--------------------------------

### Manual Resource Management

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/CLIENT.md

Example demonstrating manual cleanup of the ZaiClient instance using the `close()` method.

```APIDOC
Or manual cleanup:

```python
client = ZaiClient(api_key="your-api-key")
try:
    # Use client...
finally:
    client.close()
```
```

--------------------------------

### Example Usage of ZhipuAiClient

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/CLIENT.md

Demonstrates initializing the ZhipuAiClient and making a chat completion request. This client is configured for mainland China.

```python
from zai import ZhipuAiClient

# Initialize for mainland China
client = ZhipuAiClient(api_key="your-api-key-here")

# All API resources available same as ZaiClient
response = client.chat.completions.create(
    model="glm-5.2",
    messages=[{"role": "user", "content": "Hello"}]
)
```

--------------------------------

### Error Handling Example

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/CLIENT.md

Example demonstrating how to handle potential API errors when using ZaiClient, including authentication, timeout, and general status errors.

```APIDOC
### Error Handling Example

```python
from zai import ZaiClient
import zai

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

try:
    response = client.chat.completions.create(
        model="glm-5.2",
        messages=[{"role": "user", "content": "Hello"}]
    )
except zai.core.APIAuthenticationError as e:
    print(f"Authentication failed: {e}")
except zai.core.APITimeoutError as e:
    print(f"Request timed out: {e}")
except zai.core.APIStatusError as e:
    print(f"API error: {e.status_code}")
```
```

--------------------------------

### Sample SDK Debug Output

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/configuration.md

Example of debug log output from the Zai SDK, showing request method, body, response status, and timing information.

```log
DEBUG:zai.core._http_client:POST /chat/completions
DEBUG:zai.core._http_client:Request body: {...}
DEBUG:zai.core._http_client:Response status: 200
DEBUG:zai.core._http_client:Response time: 1.234s
```

--------------------------------

### Video Generation and Retrieval

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/README.md

This example demonstrates video generation using the Z-AI SDK. It includes parameters for model selection, prompt, quality, audio inclusion, size, and frame rate. After generation, it shows how to retrieve the video result using the response ID.

```python
from zai import ZaiClient
client = ZaiClient(api_key="your-api-key")

# Generate video
response = client.videos.generations(
    model="cogvideox-3",
    prompt="A cat is playing with a ball.",
    quality="quality",  # Output mode, "quality" for quality priority, "speed" for speed priority
    with_audio=True, # Whether to include audio
    size="1920x1080",  # Video resolution, supports up to 4K (e.g., "3840x2160")
    fps=30,  # Frame rate, can be 30 or 60
    max_wait_time=300,  # Maximum wait time (seconds)
)
print(response)

# Get video result
result = client.videos.retrieve_videos_result(id=response.id)
print(result)
```

--------------------------------

### Create Batch from File

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/batches.md

Example of creating a batch job by uploading a file containing batch requests. It first uploads the file and then uses the file ID to create the batch.

```python
from zai import ZaiClient

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

# First, upload batch requests file
file_response = client.files.create(
    file=open("batch_requests.jsonl", "rb"),
    purpose="batch"
)

file_id = file_response.id

# Create batch job
batch = client.batches.create(
    endpoint="/v1/chat/completions",
    input_file_id=file_id,
    completion_window="one-day"
)

print(f"Batch created: {batch.id}")
print(f"Status: {batch.status}")
```

--------------------------------

### Chat Completion with Tool Call

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/README.md

This example demonstrates how to create a chat completion that can utilize tools. The 'tools' parameter is used to specify available tools, such as 'web_search', along with their configurations. The response will contain the tool call if applicable.

```python
from zai import ZaiClient

# Initialize client
client = ZaiClient(api_key="your-api-key")

# Create chat completion
response = client.chat.completions.create(
    model='glm-5.1',
    messages=[
        {'role': 'system', 'content': 'You are a helpful assistant.'},
        {'role': 'user', 'content': 'What is artificial intelligence?'},
    ],
    tools=[
        {
            'type': 'web_search',
            'web_search': {
                'search_query': 'What is artificial intelligence?',
                'search_result': True,
            },
        }
    ],
    temperature=0.5,
    max_tokens=2000,
)

print(response)
```

--------------------------------

### Module Structure Refactoring Example

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/Release-Note.md

Illustrates the reorganized directory structure for API resources within the SDK. This change aims to improve code organization and maintainability.

```treeview
src/zai/api_resource/
  ├── embeddings/
  │   ├── __init__.py
  │   └── embeddings.py
  ├── files/
  │   ├── __init__.py
  │   └── files.py
  ├── images/
  │   ├── __init__.py
  │   └── images.py
  └── batch/
      ├── __init__.py
      └── batches.py
```

--------------------------------

### ZaiClient Error Handling Example

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/CLIENT.md

Shows how to handle potential API errors, including authentication, timeouts, and general status errors, when using the ZaiClient.

```python
from zai import ZaiClient
import zai

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

try:
    response = client.chat.completions.create(
        model="glm-5.2",
        messages=[{"role": "user", "content": "Hello"}]
    )
except zai.core.APIAuthenticationError as e:
    print(f"Authentication failed: {e}")
except zai.core.APITimeoutError as e:
    print(f"Request timed out: {e}")
except zai.core.APIStatusError as e:
    print(f"API error: {e.status_code}")
```

--------------------------------

### Type Definitions Reorganization Example

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/Release-Note.md

Shows the refactored structure for type definitions, organizing related types into dedicated submodules. This enhances clarity and modularity.

```treeview
src/zai/types/
  ├── batch/
  │   ├── __init__.py
  │   ├── batch.py
  │   ├── batch_create_params.py
  │   ├── batch_error.py
  │   ├── batch_list_params.py
  │   └── batch_request_counts.py
  ├── image/
  │   ├── __init__.py
  │   └── image.py
  ├── embeddings/
  │   ├── __init__.py
  │   └── embeddings.py
  └── [other organized modules...]
```

--------------------------------

### Streaming Chat Response

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/chat.md

This example shows how to receive chat completions as a stream of data. It's useful for real-time applications where immediate feedback is desired.

```python
client = ZaiClient(api_key="your-api-key")

stream = client.chat.completions.create(
    model="glm-5.1",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Tell me a story."}
    ],
    stream=True
)

for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end='', flush=True)
```

--------------------------------

### Set Environment Variables for API Configuration

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/configuration.md

Set ZAI_API_KEY and ZAI_BASE_URL environment variables for authentication and custom API endpoints. Examples provided for Unix/Linux/macOS, Windows PowerShell, and Windows Command Prompt.

```bash
# Unix/Linux/macOS
export ZAI_API_KEY="your-api-key"
export ZAI_BASE_URL="https://api.z.ai/api/paas/v4"
```

```powershell
# Windows (PowerShell)
$env:ZAI_API_KEY="your-api-key"
$env:ZAI_BASE_URL="https://api.z.ai/api/paas/v4"
```

```cmd
# Windows (Command Prompt)
set ZAI_API_KEY=your-api-key
set ZAI_BASE_URL=https://api.z.ai/api/paas/v4
```

--------------------------------

### Web Search with Tracking Parameters

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/web-search.md

Executes a web search and includes tracking parameters like request_id and user_id. This is useful for logging and monitoring search requests. The example also demonstrates logging the completion time and result count.

```python
import time

response = client.web_search.web_search(
    search_query="my research topic",
    request_id="search_001",
    user_id="user_12345",
    count=10
)

# Log search with timing
search_timestamp = int(time.time())
print(f"Search completed at {search_timestamp}")
print(f"Request ID: search_001")
print(f"Results: {len(response.search_result)}")
```

--------------------------------

### List All Batch Jobs

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/batches.md

This method lists all batch jobs. It supports pagination using the 'after' parameter to specify a starting point and 'limit' to control the number of results. Optional extra headers, body parameters, and timeouts are also supported.

```python
def list(
    self,
    *, 
    after: str | NotGiven = NOT_GIVEN,
    limit: int | NotGiven = NOT_GIVEN,
    extra_headers: Headers | None = None,
    extra_body: Body | None = None,
    timeout: float | httpx.Timeout | None | NotGiven = NOT_GIVEN,
) -> SyncCursorPage[Batch]:
```

--------------------------------

### Set Up Client

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/INDEX.md

Initializes the Z-AI client with an API key.

```APIDOC
## Set Up Client

### Description
Initializes the Z-AI client with an API key.

### Method
```python
from zai import ZaiClient
client = ZaiClient(api_key="your-api-key")
```
```

--------------------------------

### Set Up Z-AI Client

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/INDEX.md

Initialize the ZaiClient with your API key. This is the first step before making any requests.

```python
from zai import ZaiClient
client = ZaiClient(api_key="your-api-key")
```

--------------------------------

### Initialize ZaiClient with Options

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/configuration.md

Demonstrates how to initialize the ZaiClient with various configuration parameters including API key, base URL, timeout, max retries, custom headers, and source channel.

```python
from zai import ZaiClient

client = ZaiClient(
    api_key="your-api-key",
    base_url="https://api.z.ai/api/paas/v4",
    timeout=30.0,
    max_retries=3,
    http_client=None,
    custom_headers={"Custom-Header": "value"},
    disable_token_cache=True,
    source_channel="python-sdk"
)
```

--------------------------------

### Embeddings API Reference

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/MANIFEST.txt

Reference for generating text embeddings, with details on batch processing and examples for semantic search.

```APIDOC
## Embeddings API

### Description
Enables the creation of text embeddings, which are numerical representations of text data. Supports batch processing for efficiency and provides examples for semantic search applications.

### Methods

- **create()**: Generates embeddings for provided text.
- **Batch Processing**: Information on processing multiple text inputs simultaneously.
- **Semantic Search Examples**: Demonstrations of how to use embeddings for search functionalities.
```

--------------------------------

### Load API Key and Base URL from .env File

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/README.md

Configure the ZaiClient by loading API credentials and base URL from a local .env file. Ensure the .env file is in the project root.

```dotenv
ZAI_API_KEY=your-api-key
ZAI_BASE_URL=https://api.z.ai/api/paas/v4
```

```python
from dotenv import load_dotenv
load_dotenv()

from zai import ZaiClient
client = ZaiClient()
```

--------------------------------

### Handling APIServerFlowExceedError

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/errors.md

Catch APIServerFlowExceedError when the server is overloaded (503 Service Unavailable). This example indicates that a back-off strategy should be implemented.

```python
try:
    response = client.chat.completions.create(...)
except zai.core.APIServerFlowExceedError as e:
    print("Server overloaded, please retry later")
    # Implement back-off strategy
```

--------------------------------

### Upload File for Fine-tuning

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/files.md

Upload a file from a local path for fine-tuning purposes. Ensure the file is opened in binary read mode.

```python
from zai import ZaiClient

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

# Upload from file path
file_response = client.files.create(
    file=open("training_data.jsonl", "rb"),
    purpose="fine-tune"
)

print(f"File ID: {file_response.id}")
print(f"Status: {file_response.status}")
```

--------------------------------

### Initialize ZaiClient with Explicit API Key

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/CLIENT.md

Use this snippet to initialize the ZaiClient when you have your API key readily available. This is the most straightforward way to authenticate.

```python
from zai import ZaiClient

# Initialize with explicit API key
client = ZaiClient(api_key="your-api-key-here")
```

--------------------------------

### Initialize ZaiClient and Create Chat Completion

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/Release-Note.md

Initialize the ZaiClient with your API key and create a chat completion using the glm-4 model. Replace 'your-api-key' with your actual Z.ai API key.

```python
from zai import ZaiClient

# Initialize client
client = ZaiClient(api_key="your-api-key")

# Create chat completion
response = client.chat.completions.create(
    model="glm-4",
    messages=[{"role": "user", "content": "Hello, Z.ai!"}]
)

print(response.choices[0].message.content)
```

--------------------------------

### Semantic Search Example

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/embeddings.md

Perform semantic search by generating embeddings for a query and a list of documents, then calculating similarity. Requires NumPy for vector operations.

```python
client = ZaiClient(api_key="your-api-key")

# Generate embedding for search query
query = "best practices for Python"
query_response = client.embeddings.create(
    input=query,
    model="embedding-3"
)
query_embedding = query_response.data[0].embedding

# Generate embeddings for documents
documents = [
    "Python coding standards and best practices",
    "JavaScript tutorial for beginners",
    "Database design patterns",
    "Python performance optimization"
]

doc_response = client.embeddings.create(
    input=documents,
    model="embedding-3"
)

# Calculate similarity (dot product or cosine)
import numpy as np

query_vec = np.array(query_embedding)
doc_vecs = [np.array(doc.embedding) for doc in doc_response.data]

# Cosine similarity
similarities = [
    np.dot(query_vec, doc_vec) / 
    (np.linalg.norm(query_vec) * np.linalg.norm(doc_vec))
    for doc_vec in doc_vecs
]

# Rank documents
ranked = sorted(zip(documents, similarities), key=lambda x: x[1], reverse=True)
print("Top results:")
for doc, score in ranked:
    print(f"  {score:.3f}: {doc}")
```

--------------------------------

### Basic Z.ai SDK Usage

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/README.md

Demonstrates how to create a client and make a chat completion request. Use ZaiClient for overseas regions and ZhipuAiClient for Mainland China regions.

```python
from zai import ZaiClient, ZhipuAiClient

# For Overseas users, create the ZaiClient
client = ZaiClient(api_key="your-api-key")

# For Chinese users, create the ZhipuAiClient
client = ZhipuAiClient(api_key="your-api-key")

# Create chat completion
response = client.chat.completions.create(
    model="glm-5.2",
    messages=[
        {"role": "user", "content": "Hello, Z.ai!"}
    ]
)
print(response.choices[0].message.content)
```

--------------------------------

### Adjusting Speech Speed and Volume

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/audio.md

Allows for fine-tuning speech output by adjusting speed and volume parameters. Examples show generating slow/quiet and fast/loud speech.

```python
# Slow and quiet
slow_response = client.audio.speech(
    model="glm-tts",
    input="Speak slowly",
    voice="tongtong",
    speed=0.7,  # 70% of normal speed
    volume=0.5  # 50% volume
)

# Fast and loud
fast_response = client.audio.speech(
    model="glm-tts",
    input="Speak quickly",
    voice="tongtong",
    speed=1.5,  # 150% of normal speed
    volume=2.0  # 200% volume
)

with open("slow.pcm", "wb") as f:
    f.write(slow_response.content)
with open("fast.pcm", "wb") as f:
    f.write(fast_response.content)
```

--------------------------------

### Configure Client

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/INDEX.md

Initializes the Z-AI client with an API key, timeout, and maximum retries.

```APIDOC
## Configure Client

### Description
Initializes the Z-AI client with an API key, timeout, and maximum retries.

### Method
```python
client = ZaiClient(
    api_key="key",
    timeout=30.0,
    max_retries=3
)
```
```

--------------------------------

### Handling APIAuthenticationError

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/errors.md

Catch APIAuthenticationError when API credentials are invalid or missing, indicated by a 401 Unauthorized status. The example demonstrates initializing the client with an invalid API key.

```python
try:
    client = zai.ZaiClient(api_key="invalid-key")
    response = client.chat.completions.create(
        model="glm-5.2",
        messages=[{"role": "user", "content": "Hello"}]
    )
except zai.core.APIAuthenticationError as e:
    print("Authentication failed - check your API key")
```

--------------------------------

### Generate Videos using Z-AI SDK

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/INDEX.md

Create videos from a text prompt using a video generation model. Requires specifying the model and prompt.

```python
response = client.videos.generations(
    model="cogvideox-3",
    prompt="A cat playing"
)
```

--------------------------------

### Handling APIRequestFailedError

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/errors.md

Catch APIRequestFailedError for issues with invalid request parameters or malformed data, indicated by a 400 Bad Request status. The example shows how to trigger this error with an invalid model name.

```python
try:
    response = client.chat.completions.create(
        model="invalid-model",
        messages=[{"role": "user", "content": "Hello"}]
    )
except zai.core.APIRequestFailedError as e:
    print(f"Invalid request: {e}")
    print(f"Status code: {e.status_code}")
```

--------------------------------

### Retrieve Batch Job Details

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/batches.md

Use this method to get detailed information about a specific batch job using its ID. It supports optional extra headers and body parameters, as well as request timeouts.

```python
def retrieve(
    self,
    batch_id: str,
    *, 
    extra_headers: Headers | None = None,
    extra_body: Body | None = None,
    timeout: float | httpx.Timeout | None | NotGiven = NOT_GIVEN,
) -> Batch:
```

--------------------------------

### API Key Management with .env

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/Release-Note.md

Demonstrates the recommended way to load API keys using environment variables and the python-dotenv library. Ensure your API key is stored in a .env file in your project's root directory.

```python
# Before (still works)
from zai import ZaiClient
client = ZaiClient(api_key="your-key")

# Enhanced (recommended)
import os
from zai import ZaiClient
from dotenv import load_dotenv

load_dotenv()
api_key = os.getenv('ZAI_API_KEY')
client = ZaiClient(api_key=api_key)
```

--------------------------------

### ZaiClient Resource Management with Context Manager

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/CLIENT.md

Demonstrates using the ZaiClient with a `with` statement for automatic resource cleanup. The client is automatically closed after the block.

```python
with ZaiClient(api_key="your-api-key") as client:
    response = client.chat.completions.create(
        model="glm-5.2",
        messages=[{"role": "user", "content": "Test"}]
    )
    # Client automatically cleaned up after with block
```

--------------------------------

### Configure Z.ai SDK with Environment Variables

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/README.md

Set the API key and optionally the base URL using environment variables. This is a common method for configuring SDKs.

```bash
export ZAI_API_KEY="your-api-key"
export ZAI_BASE_URL="https://api.z.ai/api/paas/v4/"  # Optional
```

--------------------------------

### Image Generation with Different Sizes and Styles

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/images.md

Demonstrates generating images with various size, style, and quality parameters to achieve different aesthetic outcomes, from professional headshots to epic landscapes.

```python
# Standard high-definition
response1 = client.images.generations(
    prompt="Professional headshot portrait",
    size="512x512",
    style="natural",
    quality="hd"
)

# Wide landscape
response2 = client.images.generations(
    prompt="Epic fantasy landscape",
    size="1536x1024",
    style="vivid"
)

# Portrait
response3 = client.images.generations(
    prompt="Beautiful woman in elegant dress",
    size="768x1024",
    quality="standard"
)
```

--------------------------------

### Initialize ZaiClient from Environment Variable

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/CLIENT.md

Initialize ZaiClient without explicitly providing an API key. The client will automatically attempt to read the key from the ZAI_API_KEY environment variable. Ensure the environment variable is set before running.

```python
client = ZaiClient()  # reads ZAI_API_KEY from environment
```

--------------------------------

### Generating Speech with Different Voices

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/audio.md

Demonstrates how to generate speech using a list of available voices. Each generated speech is saved as a WAV file.

```python
voices = ["tongtong", "chuichui", "xiaochen"]

for voice in voices:
    response = client.audio.speech(
        model="glm-tts",
        input="Hello everyone!",
        voice=voice,
        response_format="wav"
    )
    
    with open(f"speech_{voice}.wav", "wb") as f:
        f.write(response.content)
    print(f"Generated speech_{voice}.wav")
```

--------------------------------

### Batch Video Generation and Polling

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/videos.md

Shows how to initiate multiple video generation tasks concurrently and then poll for their completion status, retrieving video URLs upon success.

```python
prompts = [
    "A cat playing with yarn",
    "Dog running through a field",
    "Bird flying in the sky",
]

# Start all generations
tasks = []
for i, prompt in enumerate(prompts):
    response = client.videos.generations(
        model="cogvideox-3",
        prompt=prompt,
        size="1280x720",
        request_id=f"batch_001_{i+1}"
    )
    tasks.append((prompt, response.id))
    print(f"Started: {prompt} -> {response.id}")

# Poll all tasks until complete
import time
completed = {}

while len(completed) < len(tasks):
    for prompt, task_id in tasks:
        if task_id in completed:
            continue
        
        result = client.videos.retrieve_videos_result(id=task_id)
        if result.task_status == "SUCCESS":
            completed[task_id] = result.video_result[0].url
            print(f"✓ {prompt}")
        elif result.task_status == "FAIL":
            completed[task_id] = None
            print(f"✗ {prompt}")
    
    if len(completed) < len(tasks):
        time.sleep(10)

print("\nResults:")
for prompt, task_id in tasks:
    url = completed.get(task_id)
    print(f"  {prompt}: {url}")
```

--------------------------------

### Upload Voice Sample for Cloning

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/files.md

Upload a voice sample file for voice cloning. The file must be opened in binary read mode.

```python
# Upload voice for voice cloning
file_response = client.files.create(
    file=open("voice_sample.wav", "rb"),
    purpose="voice-clone-input"
)

print(f"Voice sample uploaded: {file_response.id}")
```

--------------------------------

### Batch Video Generation

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/videos.md

Shows how to generate multiple videos concurrently by submitting a list of prompts and then polling for their results.

```APIDOC
## Batch Video Generation

### Description
This example demonstrates how to generate multiple videos in a batch. It first initiates all generation tasks and then continuously polls for their completion status, printing the URLs of successfully generated videos.

### Method
`client.videos.generations()` and `client.videos.retrieve_videos_result()`

### Parameters for `generations()`
- **model** (string) - Required - The model to use for video generation.
- **prompt** (string) - Required - The text prompt describing the desired video content.
- **size** (string) - Required - The desired resolution of the video (e.g., "1280x720").
- **request_id** (string) - Optional - A unique identifier for the request, useful for tracking.

### Parameters for `retrieve_videos_result()`
- **id** (string) - Required - The ID of the video generation task to retrieve results for.

### Request Example
```python
prompts = [
    "A cat playing with yarn",
    "Dog running through a field",
    "Bird flying in the sky",
]

# Start all generations
tasks = []
for i, prompt in enumerate(prompts):
    response = client.videos.generations(
        model="cogvideox-3",
        prompt=prompt,
        size="1280x720",
        request_id=f"batch_001_{i+1}"
    )
    tasks.append((prompt, response.id))
    print(f"Started: {prompt} -> {response.id}")

# Poll all tasks until complete
import time
completed = {}

while len(completed) < len(tasks):
    for prompt, task_id in tasks:
        if task_id in completed:
            continue
        
        result = client.videos.retrieve_videos_result(id=task_id)
        if result.task_status == "SUCCESS":
            completed[task_id] = result.video_result[0].url
            print(f"✓ {prompt}")
        elif result.task_status == "FAIL":
            completed[task_id] = None
            print(f"✗ {prompt}")
    
    if len(completed) < len(tasks):
        time.sleep(10)

print("\nResults:")
for prompt, task_id in tasks:
    url = completed.get(task_id)
    print(f"  {prompt}: {url}")
```

### Response
#### Success Response (from `retrieve_videos_result`)
- **task_status** (string) - The status of the video generation task (e.g., "SUCCESS", "FAIL").
- **video_result** (array) - A list containing the video generation results. For successful tasks, it typically contains one item.
  - **url** (string) - The URL of the generated video if the task was successful.
```

--------------------------------

### Configure Z-AI Client

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/INDEX.md

Configure the ZaiClient with additional parameters like timeout and maximum retries for enhanced control.

```python
client = ZaiClient(
    api_key="key",
    timeout=30.0,
    max_retries=3
)
```

--------------------------------

### Saving Speech in Multiple Formats

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/audio.md

Shows how to generate speech output in different formats, specifically PCM and WAV, and save them to files.

```python
text = "This is a test message."

# Generate as PCM (default)
response_pcm = client.audio.speech(
    model="glm-tts",
    input=text,
    voice="tongtong",
    response_format="pcm"
)

# Generate as WAV
response_wav = client.audio.speech(
    model="glm-tts",
    input=text,
    voice="tongtong",
    response_format="wav"
)

with open("output.pcm", "wb") as f:
    f.write(response_pcm.content)
with open("output.wav", "wb") as f:
    f.write(response_wav.content)
```

--------------------------------

### Set API Key using Environment Variable

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/README.md

Configure the ZaiClient by setting the ZAI_API_KEY environment variable. The ZaiClient automatically reads this variable.

```bash
export ZAI_API_KEY="your-api-key"
```

--------------------------------

### Create Batch with Metadata

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/batches.md

Creates a batch job and attaches custom metadata to it for organization or tracking.

```python
batch = client.batches.create(
    endpoint="/v1/chat/completions",
    input_file_id="file_xyz",
    metadata={
        "project": "research-2024",
        "experiment": "summarization-v2",
        "team": "nlp"
    }
)

print(f"Batch {batch.id} created with metadata")
```

--------------------------------

### create

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/files.md

Upload a file to the platform. This method allows for uploading files with specified purposes and optional details for retrieval or fine-tuning.

```APIDOC
## Method: create

### Description
Upload a file to the platform. This method allows for uploading files with specified purposes and optional details for retrieval or fine-tuning.

### Method Signature
```python
def create(
    self,
    *,
    file: FileTypes = None,
    upload_detail: List[UploadDetail] = None,
    purpose: Literal['fine-tune', 'retrieval', 'batch', 'voice-clone-input'],
    knowledge_id: str = None,
    sentence_size: int = None,
    extra_headers: Headers | None = None,
    extra_body: Body | None = None,
    timeout: float | httpx.Timeout | None | NotGiven = NOT_GIVEN,
) -> FileObject
```

### Parameters
#### Required Parameters
- **purpose** (Literal['fine-tune', 'retrieval', 'batch', 'voice-clone-input']) - Purpose of the file upload. Determines how file is used.

#### Optional Parameters
- **file** (FileTypes) - File to upload. Can be file object or file path. Required if `upload_detail` not provided.
- **upload_detail** (List[UploadDetail]) - Detailed upload information. Required if `file` not provided.
- **knowledge_id** (str) - Knowledge base ID for retrieval purpose.
- **sentence_size** (int) - Sentence size for text splitting in retrieval.
- **extra_headers** (Headers) - Additional HTTP headers.
- **extra_body** (Body) - Additional body parameters.
- **timeout** (float | httpx.Timeout) - Request timeout.

### Return Type
`FileObject` - Uploaded file information:
- `id` (str): Unique file identifier
- `object` (str): Object type ('file')
- `created_at` (int): Unix timestamp of creation
- `filename` (str): Original filename
- `purpose` (str): Upload purpose
- `size` (int): File size in bytes
- `status` (str): File processing status ('uploaded', 'processed', 'failed')
- `status_details` (Optional[str]): Additional status information

### Throws
- `APIStatusError`: Base API error with status code
- `APIAuthenticationError`: Invalid or missing API key (401)
- `APIRequestFailedError`: Invalid request parameters (400)
- `APIReachLimitError`: Rate limit exceeded (429)
- `APIInternalError`: Server error (500)
- `APIServerFlowExceedError`: Server overloaded (503)
- `APITimeoutError`: Request timeout
- `ValueError`: If neither `file` nor `upload_detail` provided
```

--------------------------------

### Environment Variables for Configuration

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/CLIENT.md

Details on environment variables that can be used to configure the ZhipuAiClient and ZaiClient, including API key and base URL.

```APIDOC
## Configuration

### Environment Variables

Both client classes read from environment variables for configuration:

| Variable | Description | Example |
|----------|-------------|---------|
| `ZAI_API_KEY` | API key for authentication | `sk_xxxxx` |
| `ZAI_BASE_URL` | Custom base URL for API endpoint | `https://api.z.ai/api/paas/v4` |
```

--------------------------------

### Configure Custom HTTP Client

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/configuration.md

Use your own httpx.Client for advanced configuration such as timeouts, connection limits, and custom headers. The SDK manages the lifecycle of the provided client; do not call close() on it.

```python
import httpx
from zai import ZaiClient

# Create custom HTTP client
http_client = httpx.Client(
    timeout=30.0,
    limits=httpx.Limits(max_connections=100, max_keepalive_connections=50),
    headers={"User-Agent": "MyApp/1.0"}
)

# Use with SDK
client = ZaiClient(
    api_key="key",
    http_client=http_client
)

# Client will use your http_client
# Don't call close() on client - you manage the http_client
```

--------------------------------

### Create Batch Requests File

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/batches.md

Prepares and writes chat completion requests to a JSONL file, then uploads the file and creates a batch job. This is useful for preparing custom batch requests.

```python
import json

# Prepare batch requests
requests = [
    {
        "custom_id": "summary-1",
        "params": {
            "model": "glm-5.2",
            "messages": [{"role": "user", "content": "Summarize: The quick brown fox..."}],
            "max_tokens": 100
        }
    },
    {
        "custom_id": "summary-2",
        "params": {
            "model": "glm-5.2",
            "messages": [{"role": "user", "content": "Summarize: Python is a programming..."}],
            "max_tokens": 100
        }
    },
]

# Write to JSONL file
with open("batch_requests.jsonl", "w") as f:
    for request in requests:
        f.write(json.dumps(request) + "\n")

# Upload and create batch
file_response = client.files.create(
    file=open("batch_requests.jsonl", "rb"),
    purpose="batch"
)

batch = client.batches.create(
    endpoint="/v1/chat/completions",
    input_file_id=file_response.id
)

print(f"Batch {batch.id} submitted with {len(requests)} requests")
```

--------------------------------

### Upload and Manage Files using Z-AI SDK

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/INDEX.md

Upload files for purposes like batch processing. The file must be opened in binary read mode.

```python
response = client.files.create(
    file=open("data.jsonl", "rb"),
    purpose="batch"
)
```

--------------------------------

### Initialize ZaiClient with Custom Endpoint

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/configuration.md

Configure the ZaiClient to use a custom API endpoint by providing the base_url parameter.

```python
from zai import ZaiClient

client = ZaiClient(
    api_key="your-api-key",
    base_url="https://custom.api.endpoint/v4"
)
```

--------------------------------

### Configure Z.ai SDK with Code

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/README.md

Configure the ZaiClient or ZhipuAiClient directly within your Python code by passing the API key and an optional base URL.

```python
from zai import ZaiClient, ZhipuAiClient

client = ZaiClient(
    api_key="your-api-key",
    base_url="https://api.z.ai/api/paas/v4/"  # Optional
)

# if you want to use Zhipu's domain service
zhipu_client = ZhipuAiClient(
    api_key="your-api-key",
    base_url="https://open.bigmodel.cn/api/paas/v4/"  # Optional
)
```

--------------------------------

### Initialize ZhipuAiClient for Mainland China

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/CLIENT.md

Initialize the ZhipuAiClient for use with mainland China API endpoints. Requires an API key.

```python
ZhipuAiClient(
    *, 
    api_key: str | None = None, 
    base_url: str | httpx.URL | None = None, 
    timeout: Union[float, Timeout, None, NotGiven] = NOT_GIVEN, 
    max_retries: int = ZAI_DEFAULT_MAX_RETRIES, 
    http_client: httpx.Client | None = None, 
    custom_headers: Mapping[str, str] | None = None, 
    disable_token_cache: bool = True, 
    _strict_response_validation: bool = False, 
    source_channel: str | None = None,
) -> None
```

--------------------------------

### List All Files

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/files.md

Retrieve a list of all uploaded files. The total count and details for each file are printed.

```python
files_list = client.files.list()

print(f"Total files: {len(files_list.data)}")
for file in files_list.data:
    print(f"  {file.id}: {file.filename} ({file.size} bytes)")
```

--------------------------------

### List Files with Filtering and Pagination

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/files.md

Use this method to retrieve a list of files. You can filter by purpose, paginate results using `limit` and `after`, and sort the output. Ensure to check `has_more` for subsequent requests.

```python
def list(
    self,
    *,
    purpose: str | NotGiven = NOT_GIVEN,
    limit: int | NotGiven = NOT_GIVEN,
    after: str | NotGiven = NOT_GIVEN,
    order: str | NotGiven = NOT_GIVEN,
    extra_headers: Headers | None = None,
    extra_body: Body | None = None,
    timeout: float | httpx.Timeout | None | NotGiven = NOT_GIVEN,
) -> ListOfFileObject:
    pass
```

--------------------------------

### Upload Files

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/README.md

Upload a file for batch processing. The file should be opened in binary read mode and its purpose specified.

```python
response = client.files.create(
    file=open("data.jsonl", "rb"),
    purpose="batch"
)

print(f"File ID: {response.id}")
```

--------------------------------

### List Batches with Pagination

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/api-reference/batches.md

Lists batches using pagination, collecting all batches and then summarizing them by status.

```python
all_batches = []
cursor = None

while True:
    if cursor:
        page = client.batches.list(after=cursor, limit=20)
    else:
        page = client.batches.list(limit=20)
    
    all_batches.extend(page.data)
    
    if not page.has_more:
        break
    
    cursor = page.data[-1].id

print(f"Total batches: {len(all_batches)}")

# Show status summary
by_status = {}
for batch in all_batches:
    status = batch.status
    by_status[status] = by_status.get(status, 0) + 1

for status, count in by_status.items():
    print(f"  {status}: {count}")
```

--------------------------------

### Production Configuration for ZaiClient

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/configuration.md

Configure ZaiClient for production with specific timeouts, max retries, and custom headers for environment and version tracking. Logging is set to WARNING level.

```python
import logging
from zai import ZaiClient
import httpx

logging.basicConfig(level=logging.WARNING)

client = ZaiClient(
    api_key="your-api-key",  # Use env var in production
    timeout=httpx.Timeout(
        timeout=30.0,
        connect=5.0,
        read=25.0
    ),
    max_retries=3,
    custom_headers={
        "X-App-Version": "1.0",
        "X-Environment": "production"
    }
)
```

--------------------------------

### Chat with AI using Z-AI SDK

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/INDEX.md

Use this snippet to initiate a chat conversation with an AI model. Requires a ZaiClient instance.

```python
response = client.chat.completions.create(
    model="glm-5.2",
    messages=[{"role": "user", "content": "Hello"}]
)
```

--------------------------------

### Use Context Manager for Automatic Cleanup

Source: https://github.com/zai-org/z-ai-sdk-python/blob/main/_autodocs/configuration.md

The recommended way to use ZaiClient is with a context manager. This ensures the client is automatically closed when exiting the block, preventing resource leaks.

```python
from zai import ZaiClient

with ZaiClient(api_key="key") as client:
    response = client.chat.completions.create(...)
    # Client automatically closed after block
```