AGI Open Network China Models

A simple yet powerful framework for accessing Chinese AI models. Currently supports the full range of SiliconFlow models, with plans to support more Chinese AI service providers in the future.

Features

• 🚀 Simple and intuitive API interface
• 🎯 Support for multiple model types (Chat, Text, Image, Audio, Video)
• 🔧 Flexible configuration options
• 📚 Comprehensive documentation and examples
• 🛠 Complete type hints

Installation

pip install agi-open-network-cn

Quick Start

SiliconFlow Models

from agi_open_network_cn import (
    SiliconFlowClient,
    SiliconFlowChatModel,
    SiliconFlowImageModel,
    SiliconFlowAudioModel,
)

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

# Use ChatGLM
chat_model = SiliconFlowChatModel(client, model_name="chatglm-turbo")
response = chat_model.simple_chat("Tell me about ChatGLM")
print(response)

# Use Stable Diffusion to generate images
image_model = SiliconFlowImageModel(client)
image_url = image_model.simple_generate("A cute Chinese dragon")
print(image_url)

# Speech to text
audio_model = SiliconFlowAudioModel(client)
text = audio_model.simple_transcribe("speech.mp3")
print(text)

Supported Models and Features

SiliconFlow

Chat Models

• ChatGLM Series
  • chatglm-turbo: General-purpose model with balanced performance
  • chatglm-pro: Professional version with enhanced capabilities
  • chatglm-std: Standard version with good cost-performance ratio
  • chatglm-lite: Lightweight version for faster responses
• Qwen Series
  • qwen-turbo: Qwen general version
  • qwen-plus: Qwen enhanced version
• GPT Series
  • gpt-3.5-turbo
  • gpt-4

Image Models

• Stable Diffusion Series
  • stable-diffusion-3-5-large-turbo: Latest version, faster generation
  • stable-diffusion-xl: Large model for higher quality
• FLUX Series
  • FLUX.1-schnell: High-performance image generation
  • Pro/black-forest-labs/FLUX.1-schnell: Professional version

Audio Features

• Speech to Text: Supports multiple languages and scenarios
• Text to Speech: High naturalness with emotional expression
• Custom Voice: Support for voice cloning

Video Features

• Text to Video: Supports various styles and scenarios
• Async Generation: Support for long video generation
• Auto Status Query: Convenient progress tracking

Advanced Usage

Custom Model Parameters

# Using advanced parameters
response = chat_model.chat(
    messages=[
        {"role": "system", "content": "You are a professional Python teacher"},
        {"role": "user", "content": "Explain decorators"},
    ],
    temperature=0.7,
    max_tokens=2000,
    top_p=0.9,
)

Batch Processing

# Batch image generation
prompts = [
    "Chinese ink painting: Mountains and waters",
    "Chinese ink painting: Plum blossoms",
    "Chinese ink painting: Bamboo",
]

for prompt in prompts:
    image_url = image_model.simple_generate(prompt)
    print(f"{prompt}: {image_url}")

Async Video Generation

from agi_open_network_cn import SiliconFlowVideoModel

video_model = SiliconFlowVideoModel(client)
response = video_model.generate("A video showcasing Chinese traditional culture")
request_id = response["request_id"]

# Poll for results
while True:
    status = video_model.get_status(request_id)
    if status["status"] == "completed":
        print(f"Video URL: {status['url']}")
        break
    time.sleep(10)

Error Handling

from agi_open_network_cn.exceptions import AGIOpenNetworkError

try:
    response = chat_model.simple_chat("Hello")
except AGIOpenNetworkError as e:
    print(f"Error occurred: {e}")

Contributing

We welcome all forms of contributions, including but not limited to:

• Submitting issues and suggestions
• Improving documentation
• Adding new features
• Fixing bugs
• Adding new model providers

License

MIT License

Contact Us

• Website: https://www.agiopen.network
• GitHub: https://github.com/agiopennetwork/agi-open-network-cn
• Email: info@agiopen.network