gpt4free/README.md
2023-10-10 22:04:42 +01:00

31 KiB
Raw Blame History

248433934-7886223b-c1d1-4260-82aa-da5741f303bb

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.

pip install -U g4f

New features

python -m g4f.gui.run

or

from g4f.gui import run_gui; run_gui()

preview:

image
  • run interference from pypi package:
python -m g4f.interference.run

or

from g4f.interference import run_interference; run_interference()

Table of Contents

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
  1. Navigate to the project directory:
cd gpt4free
  1. (Recommended) Create a virtual environment to manage Python packages for your project:
python3 -m venv venv
  1. Activate the virtual environment:
    • On Windows:
    .\venv\Scripts\activate
    
    • On macOS and Linux:
    source venv/bin/activate
    
  2. Install the required Python packages from requirements.txt:
pip install -r requirements.txt
  1. 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.

  2. Clone the GitHub repo:

git clone https://github.com/xtekky/gpt4free.git
  1. Navigate to the project directory:
cd gpt4free
  1. Build the Docker image:
docker compose build
  1. 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

g4f.logging = True # enable logging
g4f.check_version = False # Disable automatic version checking
print(g4f.version) # check version
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"}],
    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": "Hello"}],
)  # alterative model setting

print(response)
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:
import g4f
from g4f.Provider import (
    AItianhu,
    Acytoo,
    Aichat,
    Ails,
    Bard,
    Bing,
    ChatBase,
    ChatgptAi,
    H2o,
    HuggingChat,
    OpenAssistant,
    OpenaiChat,
    Raycast,
    Theb,
    Vercel,
    Vitalentum,
    Ylokh,
    You,
    Yqcloud,
)

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

for message in response:
    print(message)
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.ChatBase,
    g4f.Provider.Bing,
    g4f.Provider.GptGo,
    g4f.Provider.You,
    g4f.Provider.Yqcloud,
]

async def run_provider(provider: g4f.Provider.BaseProvider):
    try:
        response = await g4f.ChatCompletion.create_async(
            model=g4f.models.default,
            messages=[{"role": "user", "content": "Hello"}],
            provider=provider,
        )
        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())
Proxy Support:

All providers support specifying a proxy in the create function.

import g4f

response = await g4f.ChatCompletion.create(
    model=g4f.models.default,
    messages=[{"role": "user", "content": "Hello"}],
    proxy="http://host:port",
    # or socks5://user:pass@host:port
)

print(f"Result:", response)

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

run interference from pypi package:

from g4f.interference import run_interference

run_interference()

run interference from repo:

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 etc/interference/requirements.txt

run server:

python3 -m etc/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 ✔️ ✔️ Active
chat.acytoo.com g4f.Provider.Acytoo ✔️ ✔️ ✔️ Active
chat-gpt.org g4f.Provider.Aichat ✔️ ✔️ Unknown
ai.ls g4f.Provider.Ails ✔️ ✔️ ✔️ Active
bard.google.com g4f.Provider.Bard ✔️ Active ✔️
bing.com g4f.Provider.Bing ✔️ ✔️ ✔️ Active
www.chatbase.co g4f.Provider.ChatBase ✔️ ✔️ ✔️ ✔️ Active
chatgpt.ai g4f.Provider.ChatgptAi ✔️ ✔️ ✔️ Active
opchatgpts.net g4f.Provider.ChatgptLogin ✔️ ✔️ Active
ava-ai-ef611.web.app g4f.Provider.CodeLinkAva ✔️ ✔️ ✔️ Active
gptgo.ai g4f.Provider.GptGo ✔️ ✔️ ✔️ Active
gpt-gm.h2o.ai g4f.Provider.H2o ✔️ ✔️ Active
huggingface.co g4f.Provider.HuggingChat ✔️ ✔️ Active ✔️
opchatgpts.net g4f.Provider.Opchatgpts ✔️ ✔️ Active
open-assistant.io g4f.Provider.OpenAssistant ✔️ ✔️ Active ✔️
chat.openai.com g4f.Provider.OpenaiChat ✔️ ✔️ Active ✔️
www.perplexity.ai g4f.Provider.PerplexityAi ✔️ ✔️ Active
raycast.com g4f.Provider.Raycast ✔️ ✔️ ✔️ Active ✔️
theb.ai g4f.Provider.Theb ✔️ ✔️ Active ✔️
sdk.vercel.ai g4f.Provider.Vercel ✔️ ✔️ Active
app.vitalentum.io g4f.Provider.Vitalentum ✔️ ✔️ ✔️ Active
wewordle.org g4f.Provider.Wewordle ✔️ ✔️ Active
chat.ylokh.xyz g4f.Provider.Ylokh ✔️ ✔️ ✔️ Active
you.com g4f.Provider.You ✔️ ✔️ Active
chat9.yqcloud.top g4f.Provider.Yqcloud ✔️ ✔️ ✔️ Active
cromicle.top g4f.Provider.Cromicle ✔️ ✔️ ✔️ Active
aiservice.vercel.app g4f.Provider.AiService ✔️ Inactive
chat.dfehub.com g4f.Provider.DfeHub ✔️ ✔️ Inactive
free.easychat.work g4f.Provider.EasyChat ✔️ ✔️ Inactive
next.eqing.tech g4f.Provider.Equing ✔️ ✔️ Inactive
chat9.fastgpt.me g4f.Provider.FastGpt ✔️ ✔️ Inactive
forefront.com g4f.Provider.Forefront ✔️ ✔️ Inactive
chat.getgpt.world g4f.Provider.GetGpt ✔️ ✔️ Inactive
liaobots.com g4f.Provider.Liaobots ✔️ ✔️ ✔️ ✔️ Inactive
p5.v50.ltd g4f.Provider.V50 ✔️ Inactive
chat.wuguokai.xyz g4f.Provider.Wuguokai ✔️ Inactive

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
🎁 Projects Stars 📚 Forks 🛎 Issues 📬 Pull requests
gpt4free Stars Forks Issues Pull Requests
gpt4free-ts Stars Forks Issues Pull Requests
Free AI API's & Potential Providers List Stars Forks Issues Pull Requests
ChatGPT-Clone Stars Forks Issues Pull Requests
ChatGpt Discord Bot Stars Forks Issues Pull Requests
LangChain gpt4free Stars Forks Issues Pull Requests
ChatGpt Telegram Bot Stars Forks Issues Pull Requests
Action Translate Readme Stars Forks Issues Pull Requests
Langchain Document GPT Stars Forks Issues Pull Requests

Contribute

Create Provider with AI Tool

Call in your terminal the "create_provider" script:

python etc/tool/create_provider.py
  1. Enter your name for the new provider.
  2. Copy&Paste a cURL command from your browser developer tools.
  3. Let the AI create the provider for you.
  4. Customize the provider according to your needs.

Create Provider

  1. Check out the current list of potential providers, or find your own provider source!
  2. Create a new file in g4f/provider with the name of the Provider
  3. Implement a class that extends BaseProvider.
from __future__ import annotations

from ..typing import AsyncResult, Messages
from .base_provider import AsyncGeneratorProvider

class HogeService(AsyncGeneratorProvider):
    url                   = "https://chat-gpt.com"
    supports_gpt_35_turbo = True
    working               = True

    @classmethod
    async def create_async_generator(
        cls,
        model: str,
        messages: Messages,
        proxy: str = None,
        **kwargs
    ) -> AsyncResult:
        yield ""
  1. Here, you can adjust the settings, for example if the website does support streaming, set supports_stream to True...
  2. Write code to request the provider in create_async_generator and yield the response, even if its a one-time response, do not hesitate to look at other providers for inspiration
  3. Add the Provider Name in g4f/provider/init.py
from .HogeService import HogeService

__all__ = [
  HogeService,
]
  1. 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
Top 1 Contributor: @hlohaus

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

Star History Chart