GPT4Free源码分析与使用指南<\/h1>

项目概述<\/h2>
GPT4Free是一个开源项目，允许用户免费使用GPT模型（包括GPT-3.5和GPT-4）的接口。该项目通过分析公开可用的GPT服务的安全漏洞和认证机制，实现了绕过认证直接使用这些服务的目的。<\/p>
项目地址：https:\/\/github.com\/xtekky\/gpt4free<\/p>

支持的接口<\/h2>

GPT-3.5接口<\/h3>
Bing<\/li>
ChatgptAi<\/li>
GptGo<\/li>
GeekGpt<\/li>
You<\/li>
Chatgpt4Online<\/li> <\/ul>
GPT-4接口<\/h3>
项目中也包含GPT-4的接口实现<\/li> <\/ul>
安装与运行<\/h2>

Web界面运行<\/h3>
python .<\/span>\g4f\gui\run.<\/span>py
<\/span><\/span><\/code><\/pre>默认使用GPT-3.5模型，切换到GPT-4可能会失败<\/p>
Python包使用方式<\/h3>
import<\/span> g4f
<\/span><\/span>
<\/span><\/span>g4f.<\/span>debug.<\/span>logging =<\/span> False<\/span>  # 禁用调试日志<\/span>
<\/span><\/span>g4f.<\/span>debug.<\/span>check_version =<\/span> False<\/span>  # 禁用自动版本检查<\/span>
<\/span><\/span>
<\/span><\/span># 流式响应<\/span>
<\/span><\/span>response =<\/span> g4f.<\/span>ChatCompletion.<\/span>create(
<\/span><\/span>    model=<\/span>"gpt-3.5-turbo"<\/span>,
<\/span><\/span>    messages=<\/span>[{"role"<\/span>: "user"<\/span>, "content"<\/span>: "昨天的当天是明天的什么"<\/span>}],
<\/span><\/span>    stream=<\/span>True<\/span>,
<\/span><\/span>)
<\/span><\/span>for<\/span> message in<\/span> response:
<\/span><\/span>    print(message, flush=<\/span>True<\/span>, end=<\/span>''<\/span>)
<\/span><\/span>
<\/span><\/span># GPT-4接口使用<\/span>
<\/span><\/span>response =<\/span> g4f.<\/span>ChatCompletion.<\/span>create(
<\/span><\/span>    model=<\/span>g4f.<\/span>models.<\/span>gpt_4,
<\/span><\/span>    messages=<\/span>[{"role"<\/span>: "user"<\/span>, "content"<\/span>: "昨天的当天是明天的什么"<\/span>}],
<\/span><\/span>)
<\/span><\/span>print(response)
<\/span><\/span><\/code><\/pre>技术实现分析<\/h2>
整体架构<\/h3>

前端：基于Flask框架<\/li>
后端API接口：\/backend-api\/v2\/conversation<\/code><\/li>
默认模型配置：<\/li>
<\/ul>
default =<\/span> Model(
<\/span><\/span>    name=<\/span>""<\/span>,
<\/span><\/span>    base_provider=<\/span>""<\/span>,
<\/span><\/span>    best_provider=<\/span>RetryProvider([
<\/span><\/span>        Bing,
<\/span><\/span>        ChatgptAi, GptGo, GeekGpt,
<\/span><\/span>        You,
<\/span><\/span>        Chatgpt4Online
<\/span><\/span>    ])
<\/span><\/span>)
<\/span><\/span><\/code><\/pre>各接口实现机制<\/h3>
1. Bing接口实现<\/h4>
实现步骤：<\/strong><\/p>

获取sec_access_token<\/code><\/li>
建立WebSocket连接<\/li>
发送请求并接收响应<\/li>
<\/ol>
关键代码：<\/strong><\/p>
# 获取token和其他参数<\/span>
<\/span><\/span>url1 =<\/span> 'https:\/\/www.bing.com\/turing\/conversation\/create?bundleVersion=1.1467.3'<\/span>
<\/span><\/span>response =<\/span> session.<\/span>get(url1,headers=<\/span>headers)
<\/span><\/span>data =<\/span> json.<\/span>loads(response.<\/span>text)
<\/span><\/span>conversationId =<\/span> data.<\/span>get('conversationId'<\/span>)
<\/span><\/span>clientId =<\/span> data.<\/span>get('clientId'<\/span>)
<\/span><\/span>conversationSignature =<\/span> response.<\/span>headers.<\/span>get('X-Sydney-Encryptedconversationsignature'<\/span>)
<\/span><\/span>
<\/span><\/span># WebSocket连接<\/span>
<\/span><\/span>async<\/span> with<\/span> session.<\/span>ws_connect('wss:\/\/sydney.bing.com\/sydney\/ChatHub'<\/span>, 
<\/span><\/span>                            autoping=<\/span>False<\/span>, 
<\/span><\/span>                            params=<\/span>{'sec_access_token'<\/span>: conversation.<\/span>conversationSignature}, 
<\/span><\/span>                            proxy=<\/span>proxy) as<\/span> wss:
<\/span><\/span>    await<\/span> wss.<\/span>send_str(format_message({'protocol'<\/span>: 'json'<\/span>, 'version'<\/span>: 1<\/span>}))
<\/span><\/span>    await<\/span> wss.<\/span>receive(timeout=<\/span>900<\/span>)
<\/span><\/span>    await<\/span> wss.<\/span>send_str(create_message(conversation, prompt, tone, context, web_search, gpt4_turbo))
<\/span><\/span><\/code><\/pre>2. Chatgpt4Online接口实现<\/h4>
实现步骤：<\/strong><\/p>

获取_wpnonce<\/code>值<\/li>
构造请求发送到API接口<\/li>
<\/ol>
关键代码：<\/strong><\/p>
# 获取_wpnonce<\/span>
<\/span><\/span>response =<\/span> requests.<\/span>get(f<\/span>"<\/span>{<\/span>url}<\/span>\/"<\/span>,headers=<\/span>headers)
<\/span><\/span>result =<\/span> re.<\/span>search(r<\/span>'restNonce&quot;:&quot;(.*?)&quot;'<\/span>, response.<\/span>text)
<\/span><\/span>_wpnonce =<\/span> result.<\/span>group(1<\/span>)
<\/span><\/span>
<\/span><\/span># 发送请求<\/span>
<\/span><\/span>response =<\/span> requests.<\/span>post(
<\/span><\/span>    "https:\/\/chatgpt4online.org\/wp-json\/mwai-ui\/v1\/chats\/submit"<\/span>,
<\/span><\/span>    json=<\/span>data,
<\/span><\/span>    headers=<\/span>{
<\/span><\/span>        "Content-Type"<\/span>:"application\/json"<\/span>,
<\/span><\/span>        "x-wp-nonce"<\/span>: _wpnonce
<\/span><\/span>    }
<\/span><\/span>)
<\/span><\/span><\/code><\/pre>3. GptGo接口实现<\/h4>
实现步骤：<\/strong><\/p>

获取token<\/li>
使用token查询结果<\/li>
<\/ol>
关键代码：<\/strong><\/p>
# 获取token<\/span>
<\/span><\/span>response =<\/span> session.<\/span>post(
<\/span><\/span>    "https:\/\/gptgo.ai\/get_token.php"<\/span>,
<\/span><\/span>    data=<\/span>{"ask"<\/span>: message},
<\/span><\/span>    headers=<\/span>headers
<\/span><\/span>)
<\/span><\/span>token =<\/span> base64.<\/span>b64decode(response.<\/span>text[10<\/span>:-<\/span>20<\/span>]).<\/span>decode()
<\/span><\/span>
<\/span><\/span># 查询结果<\/span>
<\/span><\/span>response =<\/span> session.<\/span>get(
<\/span><\/span>    "https:\/\/api.gptgo.ai\/web.php"<\/span>,
<\/span><\/span>    params=<\/span>{"array_chat"<\/span>: token},
<\/span><\/span>    headers=<\/span>headers
<\/span><\/span>)
<\/span><\/span><\/code><\/pre>4. GeekGpt接口实现<\/h4>
实现特点：<\/strong><\/p>

每个用户有一定初始额度<\/li>
使用特定token访问<\/li>
<\/ul>
关键代码：<\/strong><\/p>
headers =<\/span> {
<\/span><\/span>    'authorization'<\/span>: 'Bearer pk-this-is-a-real-free-pool-token-for-everyone'<\/span>,
<\/span><\/span>    'content-type'<\/span>: 'application\/json'<\/span>,
<\/span><\/span>}
<\/span><\/span>
<\/span><\/span>response =<\/span> requests.<\/span>post("https:\/\/ai.fakeopen.com\/v1\/chat\/completions"<\/span>,
<\/span><\/span>                         headers=<\/span>headers, data=<\/span>data, stream=<\/span>True<\/span>)
<\/span><\/span><\/code><\/pre>5. You接口实现<\/h4>
实现特点：<\/strong><\/p>

使用curl_cffi<\/code>库绕过Akamai防护<\/li>
访问\/api\/streamingSearch<\/code>接口<\/li>
<\/ul>
关键代码：<\/strong><\/p>
# 使用curl_cffi绕过检测<\/span>
<\/span><\/span>from<\/span> curl_cffi import<\/span> requests
<\/span><\/span>
<\/span><\/span>response =<\/span> requests.<\/span>get('https:\/\/tls.browserleaks.com\/json'<\/span>, 
<\/span><\/span>                        headers=<\/span>headers,
<\/span><\/span>                        impersonate=<\/span>"chrome110"<\/span>)
<\/span><\/span><\/code><\/pre>反爬虫绕过技术<\/h2>
Akamai防护绕过<\/h3>
Akamai提供的内容传输网络(CDN)和网络安全解决方案会检测客户端特征，项目使用curl_cffi<\/code>库生成浏览器指纹信息来绕过检测。<\/p>
对比测试：<\/strong><\/p>

浏览器访问：akamai_hash<\/code>和akamai_text<\/code>字段有值<\/li>
普通requests访问：这两个字段为空<\/li>
使用curl_cffi访问：能生成正确的指纹信息<\/li>
<\/ol>
from<\/span> curl_cffi import<\/span> requests
<\/span><\/span>
<\/span><\/span>response =<\/span> requests.<\/span>get('https:\/\/tls.browserleaks.com\/json'<\/span>,
<\/span><\/span>                       headers=<\/span>headers,
<\/span><\/span>                       impersonate=<\/span>"chrome110"<\/span>)
<\/span><\/span><\/code><\/pre>参考资源<\/h2>

GPT4Free GitHub仓库<\/a><\/li>
curl_cffi使用指南<\/a><\/li>
绕过Cloudflare验证的方法<\/a><\/li>
<\/ol>
注意事项<\/h2>

部分接口需要特定网络环境才能访问<\/li>
接口稳定性不一，可能会被服务提供商检测和封锁<\/li>
使用这些接口可能违反服务条款，请谨慎使用<\/li>
国内IP访问某些接口可能会失败<\/li>
部分接口有使用频率限制或初始额度限制<\/li>
<\/ol>
通过分析GPT4Free项目的实现，我们可以学习到如何逆向分析各类GPT服务的API接口，以及如何绕过常见的反爬虫机制。这些技术不仅适用于GPT服务，也可以应用于其他类似场景的API分析和调用。<\/p>