Files
gpt4free/docs/async_client.md
kqlio67 b198d900aa update providers and documentation with image handling improvements (#2451)
* refactor(g4f/Provider/Airforce.py): Enhance Airforce provider with dynamic model fetching

* refactor(g4f/Provider/Blackbox.py): Enhance Blackbox AI provider configuration and streamline code

* feat(g4f/Provider/RobocodersAPI.py): Add RobocodersAPI new async chat provider

* refactor(g4f/client/__init__.py): Improve provider handling in async_generate method

* refactor(g4f/models.py): Update provider configurations for multiple models

* refactor(g4f/Provider/Blackbox.py): Streamline model configuration and improve response handling

* feat(g4f/Provider/DDG.py): Enhance model support and improve conversation handling

* refactor(g4f/Provider/Copilot.py): Enhance Copilot provider with model support

* refactor(g4f/Provider/AmigoChat.py): update models and improve code structure

* chore(g4f/Provider/not_working/AIUncensored.): move AIUncensored to not_working directory

* chore(g4f/Provider/not_working/Allyfy.py): remove Allyfy provider

* Update (g4f/Provider/not_working/AIUncensored.py g4f/Provider/not_working/__init__.py)

* refactor(g4f/Provider/ChatGptEs.py): Implement format_prompt for message handling

* refactor(g4f/Provider/Blackbox.py): Update message formatting and improve code structure

* refactor(g4f/Provider/LLMPlayground.py): Enhance text generation and error handling

* refactor(g4f/Provider/needs_auth/PollinationsAI.py): move PollinationsAI to needs_auth directory

* refactor(g4f/Provider/Liaobots.py): Update Liaobots provider models and aliases

* feat(g4f/Provider/DeepInfraChat.py): Add new DeepInfra models and aliases

* Update (g4f/Provider/__init__.py)

* Update (g4f/models.py)

* g4f/models.py

* Update g4f/models.py

* Update g4f/Provider/LLMPlayground.py

* Update (g4f/models.py g4f/Provider/Airforce.py
g4f/Provider/__init__.py g4f/Provider/LLMPlayground.py)

* Update g4f/Provider/__init__.py

* refactor(g4f/Provider/Airforce.py): Enhance text generation with retry and timeout

* Update g4f/Provider/AmigoChat.py g4f/Provider/__init__.py

* refactor(g4f/Provider/Blackbox.py): update model prefixes and image handling

Fixes #2445

- Update model prefixes for gpt-4o, gemini-pro, and claude-sonnet-3.5
- Add 'gpt-3.5-turbo' alias for 'blackboxai' model
- Modify image handling in create_async_generator method
- Add 'imageGenerationMode' and 'webSearchModePrompt' flags to API request
- Remove redundant 'imageBase64' field from image data structure

* New provider (g4f/Provider/Blackbox2.py)

Support for model llama-3.1-70b text generation

* docs(docs/async_client.md): update AsyncClient API guide with minor improvements

- Improve formatting and readability of code examples
- Add line breaks for better visual separation of sections
- Fix minor typos and inconsistencies in text
- Enhance clarity of explanations in various sections
- Remove unnecessary whitespace

* feat(docs/client.md): add response_format parameter

- Add 'response_format' parameter to image generation examples
- Specify 'url' format for standard image generation
- Include 'b64_json' format for base64 encoded image response
- Update documentation to reflect new parameter usage
- Improve code examples for clarity and consistency

* docs(README.md): update usage examples and add image generation

- Update text generation example to use new Client API
- Add image generation example with Client API
- Update configuration section with new cookie setting instructions
- Add response_format parameter to image generation example
- Remove outdated information and reorganize sections
- Update contributors list

* refactor(g4f/client/__init__.py): optimize image processing and response handling

- Modify _process_image_response to handle 'url' format without local saving
- Update ImagesResponse construction to include 'created' timestamp
- Simplify image processing logic for different response formats
- Improve error handling and logging for image generation
- Enhance type hints and docstrings for better code clarity

* feat(g4f/models.py): update model providers and add new models

- Add Blackbox2 to Provider imports
- Update gpt-3.5-turbo best provider to Blackbox
- Add Blackbox2 to llama-3.1-70b best providers
- Rename dalle_3 to dall_e_3 and update its best providers
- Add new models: solar_mini, openhermes_2_5, lfm_40b, zephyr_7b, neural_7b, mythomax_13b
- Update ModelUtils.convert with new models and changes
- Remove duplicate 'dalle-3' entry in ModelUtils.convert

* refactor(Airforce): improve API handling and add authentication

- Implement API key authentication with check_api_key method
- Refactor image generation to use new imagine2 endpoint
- Improve text generation with better error handling and streaming
- Update model aliases and add new image models
- Enhance content filtering for various model outputs
- Replace StreamSession with aiohttp's ClientSession for async operations
- Simplify model fetching logic and remove redundant code
- Add is_image_model method for better model type checking
- Update class attributes for better organization and clarity

* feat(g4f/Provider/HuggingChat.py): update HuggingChat model list and aliases

Request by @TheFirstNoob
- Add 'Qwen/Qwen2.5-72B-Instruct' as the first model in the list
- Update model aliases to include 'qwen-2.5-72b'
- Reorder existing models in the list for consistency
- Remove duplicate entry for 'Qwen/Qwen2.5-72B-Instruct' in models list

* refactor(g4f/Provider/ReplicateHome.py): remove unused text models

Request by @TheFirstNoob
- Removed the 'meta/meta-llama-3-70b-instruct' and 'mistralai/mixtral-8x7b-instruct-v0.1' text models from the  list
- Updated the  list to only include the remaining text and image models
- This change simplifies the model configuration and reduces the number of available models, focusing on the core text and image models provided by Replicate

* refactor(g4f/Provider/HuggingChat.py): Move HuggingChat to needs_auth directory

Request by @TheFirstNoob

* Update (g4f/Provider/needs_auth/HuggingChat.py)

* Update g4f/models.py

* Update g4f/Provider/Airforce.py

* Update g4f/models.py g4f/Provider/needs_auth/HuggingChat.py

* Added 'Airforce' provider to the 'o1-mini' model (g4f/models.py)

* Update (g4f/Provider/Airforce.py g4f/Provider/AmigoChat.py)

* Update g4f/models.py g4f/Provider/DeepInfraChat.py g4f/Provider/Airforce.py

* Update g4f/Provider/DeepInfraChat.py

* Update (g4f/Provider/DeepInfraChat.py)

* Update g4f/Provider/Blackbox.py

* Update (docs/client.md docs/async_client.md g4f/client/__init__.py)

* Update (docs/async_client.md docs/client.md)

* Update (g4f/client/__init__.py)

---------

Co-authored-by: kqlio67 <kqlio67@users.noreply.github.com>
Co-authored-by: kqlio67 <>
Co-authored-by: H Lohaus <hlohaus@users.noreply.github.com>
2024-12-05 01:07:59 +01:00

11 KiB
Raw Blame History

G4F - AsyncClient API Guide

The G4F AsyncClient API is a powerful asynchronous interface for interacting with various AI models. This guide provides comprehensive information on how to use the API effectively, including setup, usage examples, best practices, and important considerations for optimal performance.

Compatibility Note

The G4F AsyncClient API is designed to be compatible with the OpenAI API, making it easy for developers familiar with OpenAI's interface to transition to G4F.

Table of Contents

Introduction

The G4F AsyncClient API is an asynchronous version of the standard G4F Client API. It offers the same functionality as the synchronous API but with improved performance due to its asynchronous nature. This guide will walk you through the key features and usage of the G4F AsyncClient API.

Key Features

  • Custom Providers: Use custom providers for enhanced flexibility.
  • ChatCompletion Interface: Interact with chat models through the ChatCompletion class.
  • Streaming Responses: Get responses iteratively as they are received.
  • Non-Streaming Responses: Generate complete responses in a single call.
  • Image Generation and Vision Models: Support for image-related tasks.

Getting Started

Initializing the AsyncClient

To use the G4F AsyncClient, create a new instance:

from g4f.client import AsyncClient
from g4f.Provider import OpenaiChat, Gemini

client = AsyncClient(
    provider=OpenaiChat,
    image_provider=Gemini,
    # Add other parameters as needed
)

Creating Chat Completions

Heres an improved example of creating chat completions:

response = await client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[
        {
            "role": "user",
            "content": "Say this is a test"
        }
    ]
     # Add other parameters as needed
)

This example:

  • Asks a specific question Say this is a test
  • Configures various parameters like temperature and max_tokens for more control over the output
  • Disables streaming for a complete response

You can adjust these parameters based on your specific needs.

Configuration

Configure the AsyncClient with additional settings:

client = AsyncClient(
    api_key="your_api_key_here",
    proxies="http://user:pass@host",
    # Add other parameters as needed
)

Usage Examples

Text Completions

Generate text completions using the ChatCompletions endpoint:

import asyncio
from g4f.client import AsyncClient

async def main():
    client = AsyncClient()
    
    response = await client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[
            {
                "role": "user",
                "content": "Say this is a test"
            }
        ]
    )
    
    print(response.choices[0].message.content)

asyncio.run(main())

Streaming Completions

Process responses incrementally as they are generated:

import asyncio
from g4f.client import AsyncClient

async def main():
    client = AsyncClient()

    stream = client.chat.completions.create(
        model="gpt-4",
        messages=[
            {
                "role": "user",
                "content": "Say this is a test"
            }
        ],
        stream=True,
    )
    
    async for chunk in stream:
        if chunk.choices and chunk.choices[0].delta.content:
            print(chunk.choices[0].delta.content, end="")

asyncio.run(main())

Using a Vision Model

Analyze an image and generate a description:

import g4f
import requests
import asyncio
from g4f.client import AsyncClient

async def main():
    client = AsyncClient(
        provider=g4f.Provider.CopilotAccount
    )

    image = requests.get("https://raw.githubusercontent.com/xtekky/gpt4free/refs/heads/main/docs/cat.jpeg", stream=True).raw

    response = await client.chat.completions.create(
        model=g4f.models.default,
        messages=[
            {
                "role": "user",
                "content": "What's in this image?"
            }
        ],
        image=image
    )

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

asyncio.run(main())

Image Generation

The response_format parameter is optional and can have the following values:

  • If not specified (default): The image will be saved locally, and a local path will be returned (e.g., "/images/1733331238_cf9d6aa9-f606-4fea-ba4b-f06576cba309.jpg").
  • "url": Returns a URL to the generated image.
  • "b64_json": Returns the image as a base64-encoded JSON string.

Generate images using a specified prompt:

import asyncio
from g4f.client import AsyncClient

async def main():
    client = AsyncClient()
    
    response = await client.images.generate(
        prompt="a white siamese cat",
        model="flux",
        response_format="url"
        # Add any other necessary parameters
    )
    
    image_url = response.data[0].url
    print(f"Generated image URL: {image_url}")

asyncio.run(main())

Base64 Response Format

import asyncio
from g4f.client import AsyncClient

async def main():
    client = AsyncClient()
    
    response = await client.images.generate(
        prompt="a white siamese cat",
        model="flux",
        response_format="b64_json"
        # Add any other necessary parameters
    )
    
    base64_text = response.data[0].b64_json
    print(base64_text)

asyncio.run(main())

Concurrent Tasks with asyncio.gather

Execute multiple tasks concurrently:

import asyncio
from g4f.client import AsyncClient

async def main():
    client = AsyncClient()
    
    task1 = client.chat.completions.create(
        model=None,
        messages=[
            {
                "role": "user",
                "content": "Say this is a test"
            }
        ]
    )
    
    task2 = client.images.generate(
        model="flux",
        prompt="a white siamese cat",
        response_format="url"
    )
    
    try:
        chat_response, image_response = await asyncio.gather(task1, task2)
        
        print("Chat Response:")
        print(chat_response.choices[0].message.content)
        
        print("\nImage Response:")
        print(image_response.data[0].url)
    except Exception as e:
        print(f"An error occurred: {e}")

asyncio.run(main())

Available Models and Providers

The G4F AsyncClient supports a wide range of AI models and providers, allowing you to choose the best option for your specific use case. Here's a brief overview of the available models and providers:

Models

  • GPT-3.5-Turbo
  • GPT-4o-Mini
  • GPT-4
  • DALL-E 3
  • Gemini
  • Claude (Anthropic)
  • And more...

Providers

  • OpenAI
  • Google (for Gemini)
  • Anthropic
  • Microsoft Copilot
  • Custom providers

To use a specific model or provider, specify it when creating the client or in the API call:

client = AsyncClient(provider=g4f.Provider.OpenaiChat)

# or

response = await client.chat.completions.create(
    model="gpt-4",
    provider=g4f.Provider.CopilotAccount,
    messages=[
        {
            "role": "user",
            "content": "Hello, world!"
        }
    ]
)

Error Handling and Best Practices

Implementing proper error handling and following best practices is crucial when working with the G4F AsyncClient API. This ensures your application remains robust and can gracefully handle various scenarios. Here are some key practices to follow:

  1. Use try-except blocks to catch and handle exceptions:
try:
    response = await client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[
            {
                "role": "user",
                "content": "Hello, world!"
            }
        ]
    )
except Exception as e:
    print(f"An error occurred: {e}")
  1. Check the response status and handle different scenarios:
if response.choices:
    print(response.choices[0].message.content)
else:
    print("No response generated")
  1. Implement retries for transient errors:
import asyncio
from tenacity import retry, stop_after_attempt, wait_exponential

@retry(stop=stop_after_attempt(3), wait=wait_exponential(multiplier=1, min=4, max=10))
async def make_api_call():
    # Your API call here
    pass

Rate Limiting and API Usage

When working with the G4F AsyncClient API, it's important to implement rate limiting and monitor your API usage. This helps ensure fair usage, prevents overloading the service, and optimizes your application's performance. Here are some key strategies to consider:

  1. Implement rate limiting in your application:
import asyncio
from aiolimiter import AsyncLimiter

rate_limit = AsyncLimiter(max_rate=10, time_period=1)  # 10 requests per second

async def make_api_call():
    async with rate_limit:
        # Your API call here
        pass
  1. Monitor your API usage and implement logging:
import logging

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

async def make_api_call():
    try:
        response = await client.chat.completions.create(...)
        logger.info(f"API call successful. Tokens used: {response.usage.total_tokens}")
    except Exception as e:
        logger.error(f"API call failed: {e}")
  1. Use caching to reduce API calls for repeated queries:
from functools import lru_cache

@lru_cache(maxsize=100)
def get_cached_response(query):
    # Your API call here
    pass

Conclusion

The G4F AsyncClient API provides a powerful and flexible way to interact with various AI models asynchronously. By leveraging its features and following best practices, you can build efficient and responsive applications that harness the power of AI for text generation, image analysis, and image creation.

Remember to handle errors gracefully, implement rate limiting, and monitor your API usage to ensure optimal performance and reliability in your applications.


Return to Home