作为刚学go的一个练手项目，自用
- 将chatgpt前端进行逆向，实现绕过cloudflare
- 对官方api进行代理
- 实现前端接口转标准api(通过access_token实现标准api传入访问)
端口列表
/backend-api/* (前端逆向接口)
/backend-api/conversation/ws (前端WS逆向为原数据流格式)
/api/* (前端逆向接口)
/public-api/* (前端逆向接口)
/v1/* (官方api代理)
/r/v1/chat/completions (前端接口转标准api,支持流式)
/r/v1/chat/completions/ws (前端WS转标准api,支持流式)
/r/ws/v1/chat/completions (功能相同，提供兼容)
/r/v1/images/generations (前端接口转标准api,不支持流式,只支持gpt-4的账户)
/getsession (实现__Secure-next-auth.session-token刷新session，返回session,或输入username与password输出session)
/token (获取ArkoseToken)
目前ws逆向仅支持3.5，4有些问题，看ip纯度，纯度高就能输出
/r/v1/chat/completions method:["GET", "POST", "OPTIONS"]
input:
{
"model": "gpt-3.5-turbo-16k",
"messages": [
{
"role": "user",
"content": "what can you do"
}
]
}
output:
{
"id": "chatcmpl-m3mYrjKTZuoNARfQerON95UJlA9XSWBi",
"object": "chat.completion",
"created": 1701011706,
"model": "gpt-3.5-turbo-16k",
"choices": [
{
"index": 0,
"message": {
"role": "assistant",
"content": "I can do a wide range of tasks and provide information on various topics. Here are some of the things I can do:\n\n1. Answer Questions: I can provide information on a wide range of topics, including science, history, technology, mathematics, and more.\n\n2. Generate Text: I can generate text for various purposes, such as writing essays, creating stories, composing emails, and more.\n\n3. Language Translation: I can translate text from one language to another.\n\n4. Math Assistance: I can help with mathematical calculations, equations, and explanations.\n\n5. Programming Help: I can assist with coding and programming-related questions and problems.\n\n6. Writing Assistance: I can help with grammar and writing suggestions, including editing and proofreading.\n\n7. General Knowledge: I can provide general knowledge and facts on a wide variety of subjects.\n\n8. Recommendations: I can offer recommendations for books, movies, music, travel destinations, and more.\n\n9. Conversation and Chat: I can engage in casual conversation and chat on a variety of topics.\n\n10. Learning and Education: I can assist with learning and provide explanations on academic subjects.\n\n11. Problem Solving: I can help you brainstorm ideas, solve problems, and make decisions.\n\n12. Trivia and Quizzes: I can create and participate in trivia quizzes and answer trivia questions.\n\nPlease keep in mind that I do not have access to real-time information beyond my last knowledge update in January 2022, so some information may be outdated, and I cannot provide current news or events. If you have a specific task or question in mind, feel free to ask, and I'll do my best to assist you!"
},
"finish_reason": "stop"
}
],
"usage": {
"prompt_tokens": 0,
"completion_tokens": 0,
"total_tokens": 0
}
}
/r/v1/images/generations method:["GET", "POST", "OPTIONS"]
input:
{
"model": "dall-e-3",
"prompt": "A cute baby sea otter",
"n": 1,
"size": "1024x1024"
}
output:
{
"created": 1701014049,
"data": [
{
"revised_prompt": "A cute baby sea otter, looking fluffy and adorable, with big, curious eyes, floating on its back in a calm blue ocean. The otter is holding a small shell in its tiny paws, and its fur is wet, giving it a shiny appearance under the sun. The background features a serene seascape with gentle waves and a clear sky.",
"url": "https://files.oaiusercontent.com/file-fqEmsHBijHGkBKo0CnYIAfCJ?se=2023-11-26T16%3A54%3A09Z&sp=r&sv=2021-08-06&sr=b&rscc=max-age%3D31536000%2C%20immutable&rscd=attachment%3B%20filename%3Daa87dac2-8142-419d-9fe1-afa90c0a376e.webp&sig=xjwmZhzC3fZSF7V6TJ5hTWkmxBOMiVQKs0v/wTJRvAM%3D"
}
]
}
/getsession methods:["POST"]
input:
{
"refreshCookie":""
}
or
{
"username":"",
"password":""
}
output:
{
"user": {
"id": "",
"name": "",
"email": "",
"image": "",
"picture": "",
"idp": "auth0",
"iat": 1701014297,
"mfa": false,
"groups": [],
"intercom_hash": ""
},
"expires": "2024-02-24T15:58:17.821Z",
"accessToken": "",
"authProvider": "auth0",
"models": [
{
"slug": "text-davinci-002-render-sha",
"max_tokens": 8191,
"title": "Default (GPT-3.5)",
"description": "Our fastest model, great for most everyday tasks.",
"tags": [
"gpt3.5"
],
"capabilities": {},
"product_features": {}
}
],
"refreshCookie": ""
}
/token/:id methods:["GET"]
eg: /token/0A1D34FC-659D-4E23-B17B-694DCFCF6A6C
output:
{
"token": ""
}
- 在harPool目录中加入har文件，实现登录验证与gpt4对话验证(获取har教程)
- 复制一份.env.temp，并修改名称为.env，修改配置项后保存
proxy = "http://127.0.0.1:10809" #代理地址 （选填）
port = 5000 #程序运行端口
host = '127.0.0.1' #可访问ip，0.0.0.0允许所有ip
verify = false #是否对访问进行验证
auth_key = "" #若开启访问验证，则需要在Header中添加AuthKey字段，且值为auth_key的值才能访问 （选填）
arkose_must = false #是否强行gpt3.5进行验证
OpenAI_HOST = "chat.openai.com" #openai网页api接口地址 （选填）
openai_api_host = "api.openai.com" #openai官方api接口 （选填）
proxy_pool_url="" #ipidea代理池链接 （选填）
#示例http://api.proxy.ipidea.io/getProxyIp?num=10&return_type=json&lb=1&sb=0&flow=1®ions=us&protocol=http，根据访问频次设置num值
log_level = "debug" #日志等级
redis_address = "127.0.0.1:6379" #redis地址（若不开启代理池可选填）
redis_passwd = "" #redis密码
redis_db = 0 #选择的redis数据库其中proxy_pool_url使用的是ipidea的代理池，注册送100M流量，无限ip，一个月，测试足够 使用代理池后需要填写redis信息，redis版本需要7以上
go build && ./WarpGPT
首先克隆代码
git clone https://github.com/oliverkirk-sudo/WarpGPT.git
cd WarpGPT正确配置.env文件，在harPool中放入har文件 （其中host应该为0.0.0.0）
docker build -t warpgpt .
docker run -d -p 5000:5000 warpgptApache-2.0