g4f-sdk 0.1.0

The missing resilient and intelligent SDK for the g4f (GPT4Free) library.

G4F-SDK: The Resilient G4F Client

G4F-SDK is the missing fault-tolerant SDK for the powerful g4f (GPT4Free) library. While g4f provides access to a wide range of free AI models, its providers can often be unstable, have undocumented rate limits, and varying context length restrictions.

This module acts as an intelligent wrapper, turning g4f into a reliable tool for serious projects by adding layers of resilience, intelligence, and unified API access for Chat, Image Generation, and Audio processing.

✨ Key Features

• 🧠 Intelligent Failover & Retries: Automatically retries requests on timeouts or errors.
• 📏 Adaptive Context Management: Dynamically detects provider-specific context length limits and reduces the context window to prevent overflow errors.
• ✂️ Smart History Trimming: Automatically truncates chat history to fit within the model's context window, prioritizing system prompts and recent messages.
• 🔮 Hybrid Model Database: Combines dynamic discovery of g4f models with a rich static database to provide crucial metadata (token limits, vision/web support) even for new providers.
• 🛡️ Feature Awareness: Prevents errors by checking if a chosen provider supports features like Vision or Web Search before sending the request.
• 📦 Modular & Unified API: Provides a clean, top-level client (G4F) for all functionalities.

🔧 Installation

The G4F-SDK is available on PyPI.

1. Install the package using pip:

   pip install g4f-sdk
   

   (Note: This command automatically installs the core dependencies: g4f and tiktoken.)

2. (Optional) Create a config.json file in your project root to customize settings.

🚀 Quick Start

Initialize the client and run a resilient chat completion.

import asyncio
# Import the main client class G4F from the installed package
from ai import G4F

client = G4F()

async def main():
    print("--- Starting a simple chat ---")

    # The generate method is resilient to failures
    response_content, updated_context = await client.chat.generate(
        msg="Hello! Can you tell me a fun fact about programming?"
    )

    if response_content:
        print("\nAI Response:")
        print(response_content)
    else:
        print("\nFailed to get a response after several retries.")

if __name__ == "__main__":
    asyncio.run(main())

📚 Full Usage Guide

Initializing the Client

The main client class is G4F.

1. Default (searches for config.json):

from ai import G4F
client = G4F()

2. With a dictionary:

custom_config = {"timeout": 60, "max_retries": 3}
client = G4F(config_input=custom_config)

Chat Completions

The client.chat object handles all text and vision tasks.

Simple Text Generation:

response, context = await client.chat.generate(msg="What is the capital of France?")
print(response)

Using Vision (Image Input):

from g4f.models import gpt_4o
import base64

# Use client.new_chat() for a specific model/context
vision_chat = client.new_chat(model=gpt_4o)

with open("image.jpg", "rb") as f:
    image_base64 = base64.b64encode(f.read()).decode("utf-8")

response, _ = await vision_chat.generate(
    msg="What is in this image?",
    images=[f"data:image/jpeg;base64,{image_base64}"]
)
print(response)

Using Web Search (RAG):

response, _ = await client.chat.generate(
    msg="What are the latest news on AI?",
    web_search=True
)
print(response)

Image Generation

Use the client.images object.

image_url = await client.images.generate(
    prompt="A cute robot programming on a laptop, digital art",
    nologo=True 
)
if image_url:
    print(f"Image generated: {image_url}")

Audio Processing

Use the client.audio object for Text-to-Speech and Speech-to-Text.

Text-to-Speech (TTS):

audio_bytes = await client.audio. text_to_speech(
    text="Hello world! This is a test of the text-to-speech system."
)
if audio_bytes:
    with open("output.mp3", "wb") as f:
        f.write(audio_bytes)
    print("Saved speech to output.mp3")

Speech-to-Text (STT):

transcribed_text = await client.audio.speech_to_text(file="output.mp3")
if transcribed_text:
    print(f"Transcribed text: '{transcribed_text}'")

Managing Chat Context

You can create multiple independent chat sessions and manage their history.

# Create two separate conversations
chat_1 = client.new_chat()
chat_2 = client.new_chat()

await chat_1.generate(msg="My name is Bob.")
await chat_2.generate(msg="My name is Alice.")

# Ask chat 1 about its context
response, _ = await chat_1.generate(msg="What is my name?")
print(f"Chat 1 response: {response}")

# Ask chat 2 about its context
response, _ = await chat_2.generate(msg="What is my name?")
print(f"Chat 2 response: {response}")

# You can also manually get or set the context
current_history = chat_1.get_context()
print(current_history)

📂 Project Structure

.
├── ai/
│ ├── __init__.py # Main G4F Facade Class (imported as 'from ai import G4F')
│ ├── config.py # Config Class and Base Handler
│ ├── chat.py # ChatHandler (Text and Vision Logic)
│ ├── media.py # ImageHandler and AudioHandler
│ ├── models_info.py # Hybrid Model Database System
│ └── default_config.py # Fallback default settings
├── .gitignore # Files to ignore
├── README.md # You are here!
├── requirements.txt # Project dependencies
└── setup.py # Used to build and publish to PyPI

⚙️ Configuration

Create a config.json file in your project's root to customize behavior:

Example config.json:

{
    "api_key": null,
    "max_retries": 3,
    "timeout": 60,
    "context_reduction_factor": 0.5
}

🤝 Contributing & License

Contributions are welcome! This project is licensed under the MIT License.