search

Niansuh / chat2api

A service that can convert ChatGPT on the web to OpenAI API format.
MIT License
9 stars 4 forks source link

readme

Original Author: https://github.com/lanqian528/chat2api

CHAT2API

🤖 A simple ChatGPT TO API proxy

🌟 Free, unlimited GPT-3.5 available without an account

💥 Support AccessToken account, support GPT-4, GPT-4o, GPTs

🔍 The reply format is completely consistent with the real API and adapts to almost all clients.

Function

completed

  • [x] Streaming, non-streaming
  • [x] Login-free GPT3.5 conversation
  • [x] GPT-3.5 dialogue (the incoming model name does not contain gpt-4, then gpt-3.5 will be used by default, which is text-davinci-002-render-sha)
  • [x] GPT-4 dialogue (incoming model names include: gpt-4, gpt-4o, gpt-4-moblie to use the corresponding model, you need to pass in AccessToken)
  • [x] GPT-4 drawing, coding, networking
  • [x] Support GPTs (incoming model name: gpt-4-gizmo-g-*)
  • [x] Upload images and files (the format is API corresponding format, supports url and base64)
  • [x] WebUI (http://127.0.0.1:5005, does not support login)
  • [x] Can be used as a gateway and can be deployed on multiple machines
  • [x] Multi-account polling, supporting AccessToken and RefreshToken at the same time
  • [x] Tokens management, supports uploading and clearing

TODO

  • [ ] None yet, welcome to submit an issue

Tokens management

First configure the environment variable AUTHORIZATION, and then run the program

Visit /tokens or /api_prefix/tokens to view the number of existing Tokens, upload new Tokens, or clear Tokens

When requesting, pass in the value you configured in AUTHORIZATION to poll multiple accounts. AUTHORIZATION can be configured with multiple values separated by commas.

tokens.png

Environment variables

Each environment variable has a default value. If you do not understand the meaning of the environment variable, please do not set it, let alone pass an empty value.

# Security related
API_PREFIX=your_prefix // API prefix password. If not set, it is easy to be accessed. After setting, you need to request /your_prefix/v1/chat/completions
AUTHORIZATION=sk-xxxxxxxx,sk-yyyyyyyy // Go first /tokens Upload ac or rt, pass it in when requesting AUTHORIZATION Multiple accounts can be polled
# Request related
CHATGPT_BASE_URL=https://chatgpt.com // ChatGPT gateway address. After setting, the requested website will be changed. Multiple gateways are separated by commas.
PROXY_URL=your_first_proxy, your_second_proxy //Proxy URL, multiple proxies are separated by commas
ARKOSE_TOKEN_URL=https://arkose.example.com/token // Get the address of Arkose token, instructions are provided above
# Function related
HISTORY_DISABLED=true // Whether not to save the chat record and return conversation_id, true means not to save and not return
POW_DIFFICULTY=000032 // The difficulty of proof of work to be solved. The smaller the string, the longer the calculation time. It is recommended to 000032
RETRY_TIMES=3 //Number of retries on error
ENABLE_GATEWAY=true // Whether to enable gateway mode (WEBUI), true means enabled
CONVERSATION_ONLY=false // The gateway used supports being enabled when the server handles pow and arkose. If enabled, the conversation interface will be used directly.

Deploy

Deploy directly

git clone https://github.com/Niansuh/chat2api
cd chat2api
pip install -r requirements.txt
python app.py

Docker deployment

You need to install Docker and Docker Compose.

docker run -d \
  --name chat2api \
  -p 5005:5005 \
  niansuh/chat2api:latest

(Recommended, PLUS account available) Docker Compose deployment

Create a new directory, such as chat2api, and enter that directory:

mkdir chat2api
cd chat2api

Download the docker-compose.yml file in the library in this directory:

wget https://raw.githubusercontent.com/Niansuh/chat2api/main/docker-compose.yml

Modify the environment variables in the docker-compose.yml file and save:

docker-compose up -d

Use

  1. To use it on the web, directly access the following address, which only supports binding GPT3.5:
http://127.0.0.1:5005
  1. Using API, supports passing AccessToken or RefreshToken, available GPT-4, GPT-4o, GPTs:
curl --location 'http://127.0.0.1:5005/v1/chat/completions' \
--header 'Content-Type: application/json' \
--header 'Authorization: Bearer {{OpenAI APIKEY}}' \
--data '{
     "model": "gpt-3.5-turbo",
     "messages": [{"role": "user", "content": "Say this is a test!"}],
     "stream": true
   }'

Pass in your account's AccessToken or RefreshToken as OpenAI APIKEY

If the AUTHORIZATION environment variable is set, the set value can be passed in as OpenAI APIKEY for multi-Tokens polling

  • Obtain AccessToken: After logging in to chatgpt official website, open https://chatgpt.com/api/auth/session to obtain the value of accessToken
  • RefreshToken acquisition: No acquisition method is provided here.
  • Login-free gpt3.5 No need to pass in Token

ArkoseToken

Currently supports external services providing ArkoseToken

It is recommended to use docker-compose for deployment, which has built-in Arkose service

  1. Set the environment variable ARKOSE_TOKEN_URL

  2. When ArkoseToken is needed, chat2api will send a POST request to ARKOSE_TOKEN_URL

  3. Please provide external services in the following format:

{"blob": "rFYaxQNEApDlx/Db.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"}
{"token": "45017c7bb17115f36.7290869304|r=ap-southeast-1|meta=3|metabgclr=transparent|metaiconclr=%23757575|guitextcolor=%23000000|pk=0A1D34FC-659D-4E23-B17B-694DCFCF6A6C|at=40|sup=1|rid=3|ag=101|cdn_url=https%3A%2F%2Ftcr9i.openai.com%2Fcdn%2Ffc|lurl=https%3A%2F%2Faudio-ap-southeast-1.arkoselabs.com|surl=https%3A%2F%2Ftcr9i.openai.com|smurl=https%3A%2F%2Ftcr9i.openai.com%2Fcdn%2Ffc%2Fassets%2Fstyle-manager"}

Common problem

  • Error code:

    • 401: The current IP does not support login-free, please try changing the IP address, or setting a proxy in the environment variable PROXY_URL, or your authentication fails.
    • 403: Please check the specific error information in the log
    • 429: The current IP request has exceeded the limit within 1 hour. Please try again later or change the IP.
    • 500: Server internal error, request failed.
    • 502: The server gateway is wrong, or the network is unavailable. Please try changing the network environment.

  • What is known:

    • Many Japanese IPs do not support Bintang. It is recommended to use American IPs for Bintang 3.5.
    • 99% of accounts support free GPT-4o, but it is opened according to the IP region. Currently, Japan and Singapore IP are known to have a higher probability of being opened.

  • What is the environment variable AUTHORIZATION?

    • It is an authentication that you set for chat2api. After setting it, you can use the saved Tokens for polling. When requesting, it is passed in as APIKEY

  • How to obtain AccessToken?

  • PLUS account reports error 403?

    • PLUS account needs to configure ArkoseToken, please configure it according to the above

  • ArkoseToken What is it and how to get it?

Sponsor

Capsolver

License

MIT License