By using this repository or any code related to it, you agree to the legal notice. The author is not responsible for any copies, forks, reuploads made by other users, or anything else related to gpt4free. This is the author's only account and repository. To prevent impersonation or irresponsible actions, please comply with the GNU GPL license this Repository uses.

New

• pypi package:

pip install -U g4f

Table of Contents

• Getting Started
  • Prerequisites
  • Setting up the project
    • Install using pypi
    • Install using docker
• Usage
  • The g4f Package
  • interference openai-proxy api (use with openai python package)
• Providers
  • gpt-3.5 / gpt-4
  • Other Models
• Related gpt4free projects
• Contribute
• Contributors
• Copyright
• Star History

Getting Started

Prerequisites:

1. Download and install Python (Version 3.10+ is recommended).

Setting up the project:

Install using pypi

pip install -U g4f

or

1. Clone the GitHub repository:

git clone https://github.com/xtekky/gpt4free.git

2. Navigate to the project directory:

cd gpt4free

3. (Recommended) Create a virtual environment to manage Python packages for your project:

python3 -m venv venv

4. Activate the virtual environment:
   • On Windows:

   .\venv\Scripts\activate
   

   • On macOS and Linux:

   source venv/bin/activate
   

5. Install the required Python packages from requirements.txt:

pip install -r requirements.txt

6. Create a test.py file in the root folder and start using the repo, further Instructions are below

import g4f

...

Setting up with Docker:

If you have Docker installed, you can easily set up and run the project without manually installing dependencies.

1. First, ensure you have both Docker and Docker Compose installed.

   • Install Docker
   • Install Docker Compose

2. Clone the GitHub repo:

git clone https://github.com/xtekky/gpt4free.git

3. Navigate to the project directory:

cd gpt4free

4. Build the Docker image:

docker compose build

5. Start the service using Docker Compose:

docker compose up

You server will now be running at http://localhost:1337. You can interact with the API or run your tests as you would normally.

To stop the Docker containers, simply run:

docker compose down

Note: When using Docker, any changes you make to your local files will be reflected in the Docker container thanks to the volume mapping in the docker-compose.yml file. If you add or remove dependencies, however, you'll need to rebuild the Docker image using docker compose build.

Usage

The g4f Package

ChatCompletion

import g4f


print(g4f.Provider.Ails.params)  # supported args

# Automatic selection of provider

# streamed completion
response = g4f.ChatCompletion.create(
    model="gpt-3.5-turbo",
    messages=[{"role": "user", "content": "Hello world"}],
    stream=True,
)

for message in response:
    print(message, flush=True, end='')

# normal response
response = g4f.ChatCompletion.create(
    model=g4f.models.gpt_4,
    messages=[{"role": "user", "content": "hi"}],
)  # alterative model setting

print(response)


# Set with provider
response = g4f.ChatCompletion.create(
    model="gpt-3.5-turbo",
    provider=g4f.Provider.DeepAi,
    messages=[{"role": "user", "content": "Hello world"}],
    stream=True,
)

for message in response:
    print(message)

Completion

import g4f

allowed_models = [
    'code-davinci-002',
    'text-ada-001',
    'text-babbage-001',
    'text-curie-001',
    'text-davinci-002',
    'text-davinci-003'
]

response = g4f.Completion.create(
    model  = 'text-davinci-003',
    prompt = 'say this is a test')

print(response)

Providers:

from g4f.Provider import (
    AItianhu,
    Acytoo,
    Aichat,
    Ails,
    Aivvm,
    Bard,
    Bing,
    ChatBase,
    ChatgptAi,
    ChatgptLogin,
    CodeLinkAva,
    DeepAi,
    H2o,
    HuggingChat,
    Opchatgpts,
    OpenAssistant,
    OpenaiChat,
    Raycast,
    Theb,
    Vercel,
    Vitalentum,
    Wewordle,
    Ylokh,
    You,
    Yqcloud,
)
# Usage:
response = g4f.ChatCompletion.create(..., provider=ProviderName)

Cookies Required:

Cookies are essential for the proper functioning of some service providers. It is imperative to maintain an active session, typically achieved by logging into your account.

When running the g4f package locally, the package automatically retrieves cookies from your web browser using the get_cookies function. However, if you're not running it locally, you'll need to provide the cookies manually by passing them as parameters using the cookies parameter.

import g4f
from g4f.Provider import (
    Bard,
    Bing,
    HuggingChat,
    OpenAssistant,
    OpenaiChat,
)
# Usage:
response = g4f.ChatCompletion.create(
    model=g4f.models.default,
    messages=[{"role": "user", "content": "Hello"}],
    provider=Bard,
    #cookies=g4f.get_cookies(".google.com"),
    cookies={"cookie_name": "value", "cookie_name2": "value2"},
    auth=True
)

Async Support:

To enhance speed and overall performance, execute providers asynchronously. The total execution time will be determined by the duration of the slowest provider's execution.

import g4f, asyncio

_providers = [
    g4f.Provider.Aichat,
    g4f.Provider.Aivvm,
    g4f.Provider.ChatBase,
    g4f.Provider.Bing,
    g4f.Provider.CodeLinkAva,
    g4f.Provider.DeepAi,
    g4f.Provider.GptGo,
    g4f.Provider.Wewordle,
    g4f.Provider.You,
    g4f.Provider.Yqcloud,
]

async def run_provider(provider: g4f.Provider.AsyncProvider):
    try:
        response = await provider.create_async(
            model=g4f.models.default.name,
            messages=[{"role": "user", "content": "Hello"}],
        )
        print(f"{provider.__name__}:", response)
    except Exception as e:
        print(f"{provider.__name__}:", e)
        
async def run_all():
    calls = [
        run_provider(provider) for provider in _providers
    ]
    await asyncio.gather(*calls)

asyncio.run(run_all())

interference openai-proxy api (use with openai python package)

If you want to use the embedding function, you need to get a huggingface token. You can get one at https://huggingface.co/settings/tokens make sure your role is set to write. If you have your token, just use it instead of the OpenAI api-key.

get requirements:

pip install -r interference/requirements.txt

run server:

python3 -m interference.app

import openai

openai.api_key = "Empty if you don't use embeddings, otherwise your hugginface token"
openai.api_base = "http://localhost:1337"


def main():
    chat_completion = openai.ChatCompletion.create(
        model="gpt-3.5-turbo",
        messages=[{"role": "user", "content": "write a poem about a tree"}],
        stream=True,
    )

    if isinstance(chat_completion, dict):
        # not stream
        print(chat_completion.choices[0].message.content)
    else:
        # stream
        for token in chat_completion:
            content = token["choices"][0]["delta"].get("content")
            if content != None:
                print(content, end="", flush=True)


if __name__ == "__main__":
    main()

Models

gpt-3.5 / gpt-4

Website	Provider	gpt-3.5	gpt-4	Streaming	Asynchron	Status	Auth
www.aitianhu.com	g4f.Provider.AItianhu	✔️	❌	❌	✔️		❌
chat.acytoo.com	g4f.Provider.Acytoo	✔️	❌	✔️	✔️		❌
chat-gpt.org	g4f.Provider.Aichat	✔️	❌	❌	✔️		❌
ai.ls	g4f.Provider.Ails	✔️	❌	✔️	✔️		❌
chat.aivvm.com	g4f.Provider.Aivvm	✔️	✔️	✔️	❌		❌
bard.google.com	g4f.Provider.Bard	❌	❌	❌	✔️		✔️
bing.com	g4f.Provider.Bing	❌	✔️	✔️	✔️		❌
www.chatbase.co	g4f.Provider.ChatBase	✔️	✔️	✔️	✔️		❌
chatgpt.ai	g4f.Provider.ChatgptAi	✔️	❌	✔️	✔️		❌
opchatgpts.net	g4f.Provider.ChatgptLogin	✔️	❌	❌	✔️		❌
ava-ai-ef611.web.app	g4f.Provider.CodeLinkAva	✔️	❌	✔️	✔️		❌
deepai.org	g4f.Provider.DeepAi	✔️	❌	✔️	✔️		❌
gptgo.ai	g4f.Provider.GptGo	✔️	❌	✔️	✔️		❌
gpt-gm.h2o.ai	g4f.Provider.H2o	❌	❌	✔️	✔️		❌
huggingface.co	g4f.Provider.HuggingChat	❌	❌	✔️	✔️		✔️
opchatgpts.net	g4f.Provider.Opchatgpts	✔️	❌	❌	✔️		❌
open-assistant.io	g4f.Provider.OpenAssistant	❌	❌	✔️	✔️		✔️
chat.openai.com	g4f.Provider.OpenaiChat	✔️	❌	❌	✔️		✔️
www.perplexity.ai	g4f.Provider.PerplexityAi	✔️	❌	❌	✔️		❌
raycast.com	g4f.Provider.Raycast	✔️	✔️	✔️	❌		✔️
theb.ai	g4f.Provider.Theb	✔️	❌	✔️	❌		✔️
sdk.vercel.ai	g4f.Provider.Vercel	✔️	❌	✔️	❌		❌
app.vitalentum.io	g4f.Provider.Vitalentum	✔️	❌	✔️	✔️		❌
wewordle.org	g4f.Provider.Wewordle	✔️	❌	❌	✔️		❌
chat.ylokh.xyz	g4f.Provider.Ylokh	✔️	❌	✔️	✔️		❌
you.com	g4f.Provider.You	✔️	❌	❌	✔️		❌
chat9.yqcloud.top	g4f.Provider.Yqcloud	✔️	❌	✔️	✔️		❌
aiservice.vercel.app	g4f.Provider.AiService	✔️	❌	❌	❌		❌
chat.dfehub.com	g4f.Provider.DfeHub	✔️	❌	✔️	❌		❌
free.easychat.work	g4f.Provider.EasyChat	✔️	❌	✔️	❌		❌
next.eqing.tech	g4f.Provider.Equing	✔️	❌	✔️	❌		❌
chat9.fastgpt.me	g4f.Provider.FastGpt	✔️	❌	✔️	❌		❌
forefront.com	g4f.Provider.Forefront	✔️	❌	✔️	❌		❌
chat.getgpt.world	g4f.Provider.GetGpt	✔️	❌	✔️	❌		❌
liaobots.com	g4f.Provider.Liaobots	✔️	✔️	✔️	✔️		❌
p5.v50.ltd	g4f.Provider.V50	✔️	❌	❌	❌		❌
chat.wuguokai.xyz	g4f.Provider.Wuguokai	✔️	❌	❌	❌		❌

Other Models

Model	Base Provider	Provider	Website
palm	Google	g4f.Provider.Bard	bard.google.com
h2ogpt-gm-oasst1-en-2048-falcon-7b-v3	Huggingface	g4f.Provider.H2o	www.h2o.ai
h2ogpt-gm-oasst1-en-2048-falcon-40b-v1	Huggingface	g4f.Provider.H2o	www.h2o.ai
h2ogpt-gm-oasst1-en-2048-open-llama-13b	Huggingface	g4f.Provider.H2o	www.h2o.ai
claude-instant-v1	Anthropic	g4f.Provider.Vercel	sdk.vercel.ai
claude-v1	Anthropic	g4f.Provider.Vercel	sdk.vercel.ai
claude-v2	Anthropic	g4f.Provider.Vercel	sdk.vercel.ai
command-light-nightly	Cohere	g4f.Provider.Vercel	sdk.vercel.ai
command-nightly	Cohere	g4f.Provider.Vercel	sdk.vercel.ai
gpt-neox-20b	Huggingface	g4f.Provider.Vercel	sdk.vercel.ai
oasst-sft-1-pythia-12b	Huggingface	g4f.Provider.Vercel	sdk.vercel.ai
oasst-sft-4-pythia-12b-epoch-3.5	Huggingface	g4f.Provider.Vercel	sdk.vercel.ai
santacoder	Huggingface	g4f.Provider.Vercel	sdk.vercel.ai
bloom	Huggingface	g4f.Provider.Vercel	sdk.vercel.ai
flan-t5-xxl	Huggingface	g4f.Provider.Vercel	sdk.vercel.ai
code-davinci-002	OpenAI	g4f.Provider.Vercel	sdk.vercel.ai
gpt-3.5-turbo-16k	OpenAI	g4f.Provider.Vercel	sdk.vercel.ai
gpt-3.5-turbo-16k-0613	OpenAI	g4f.Provider.Vercel	sdk.vercel.ai
gpt-4-0613	OpenAI	g4f.Provider.Vercel	sdk.vercel.ai
text-ada-001	OpenAI	g4f.Provider.Vercel	sdk.vercel.ai
text-babbage-001	OpenAI	g4f.Provider.Vercel	sdk.vercel.ai
text-curie-001	OpenAI	g4f.Provider.Vercel	sdk.vercel.ai
text-davinci-002	OpenAI	g4f.Provider.Vercel	sdk.vercel.ai
text-davinci-003	OpenAI	g4f.Provider.Vercel	sdk.vercel.ai
llama13b-v2-chat	Replicate	g4f.Provider.Vercel	sdk.vercel.ai
llama7b-v2-chat	Replicate	g4f.Provider.Vercel	sdk.vercel.ai

Related gpt4free projects

🎁 Projects	⭐ Stars	📚 Forks	🛎 Issues	📬 Pull requests
gpt4free
gpt4free-ts
ChatGPT-Clone
ChatGpt Discord Bot
LangChain gpt4free
ChatGpt Telegram Bot
Action Translate Readme
Langchain Document GPT

Contribute

to add another provider, its very simple:

1. create a new file in g4f/provider with the name of the Provider
2. Implement a class that extends BaseProvider.

from .base_provider import BaseProvider
from ..typing import CreateResult, Any


class HogeService(BaseProvider):
    url = "http://hoge.com"
    working = True
    supports_gpt_35_turbo = True

    @staticmethod
    def create_completion(
        model: str,
        messages: list[dict[str, str]],
        stream: bool,
        **kwargs: Any,
    ) -> CreateResult:
        pass

3. Here, you can adjust the settings, for example if the website does support streaming, set working to True...
4. Write code to request the provider in create_completion and yield the response, even if its a one-time response, do not hesitate to look at other providers for inspiration
5. Add the Provider Name in g4f/provider/init.py

from .base_provider import BaseProvider
from .HogeService import HogeService

__all__ = [
  HogeService,
]

6. You are done !, test the provider by calling it:

import g4f

response = g4f.ChatCompletion.create(model='gpt-3.5-turbo', provider=g4f.Provider.PROVIDERNAME,
                                    messages=[{"role": "user", "content": "test"}], stream=g4f.Provider.PROVIDERNAME.supports_stream)

for message in response:
    print(message, flush=True, end='')

Contributors

A list of the contributors is available here
The Vercel.py file contains code from vercel-llm-api by @ading2210, which is licenced under the GNU GPL v3

Copyright

This program is licensed under the GNU GPL v3

xtekky/gpt4free: Copyright (C) 2023 xtekky

This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with this program.  If not, see <https://www.gnu.org/licenses/>.

Star History