Clewd
Clewd 是一种利用网页版 Claude 发出比网页提供的更具灵活性的请求的方式。需要提前获得 Claude 网页版应用的 Cookie。
Info
因为主要作者没有用过这种方式,所以可能不常更新这部分文档。
Tips
省流:酒馆是使用 AI 的场所,小克(Claude)是 AI,Clewd 是拐骗良家小克到酒馆的黄毛,破限是酒。
你又多违反了一条 Anthropic 的服务条款 🙃
这种方式显然违反 Anthropic 的 Claude 消费者服务协议。(3.7 节,禁止以非 Anthropic API 以外的通过机器人、脚本或其他自动或非人工方式访问 Anthropic 服务。) , 当然你越狱和搞颜色其实也违反了 Claude 的可接受使用政策,因此 Anthropic 可能会封你的 Claude 帐号,
开始之前
在开始之前,请确认以下事项:
- 要运行 Clewd 的环境已经安装了 Node.js 和 Git。
- 你拥有一个注册完毕,可以使用的 Claude 帐号。并获得了一个此帐号的 Cookie。
通过 Git 下载 Clewd
克隆 Clewd 的仓库目录:
在 Clewd 的目录里,运行 npm
命令安装依赖。你可以忽略 npm fund
和 npm audit
的警告。(反正你也不会修)
事情并非能够总是一帆风顺x2
因为还是那堵伟大光荣正确的防火墙在给你添堵 🙃
如果你在运行 npm
命令时遇到了任何的网络错误,你可以尝试为 npm 设置替代存储库位置。
例如淘宝的 npmmirror
。
added 26 packages, and audited 27 packages in 8s
3 packages are looking for funding
run `npm fund` for details
2 moderate severity vulnerabilities
To address all issues (including breaking changes), run:
npm audit fix --force
Run `npm audit` for details.
运行 node clewd.js
,首次运行时会生成必要的配置文件 config.js
。
Tips
在 Windows 上,你也可以通过双击 clewd
目录中的 start.bat
来运行。
用你用的顺手的编辑器打开生成的 config.js
文件。
- Windows 推荐使用
Notepad++
或Visual Studio Code
。 - Termux 或其它 Linux 环境的话,推荐使用
nano
, 或者是你用的顺手的其它编辑器。
/*
* https://rentry.org/teralomaniac_clewd
* https://github.com/teralomaniac/clewd
*/
// SET YOUR COOKIE BELOW
module.exports = {
// 如果你只有一个 Cookie,那么把 SET YOUR COOKIE HERE 换成你获得的 Cookie,
// 形如 sessionKey=sk 开头的一串。
"Cookie": "SET YOUR COOKIE HERE",
// 如果你有好几个 Cookie, 那么每行一个的填进 CookieArray 数组里。
"CookieArray": [
"sessionKey=sk-ant-sid01...",
"sessionKey=sk-ant-sid01..."
],
"WastedCookie": [],
"unknownModels": [],
"Cookiecounter": 3,
"CookieIndex": 0,
"ProxyPassword": "",
"Ip": "127.0.0.1",
"Port": 8444,
"localtunnel": false,
"BufferSize": 1,
"SystemInterval": 3,
"rProxy": "",
"api_rProxy": "",
"placeholder_token": "",
"placeholder_byte": "",
"PromptExperimentFirst": "",
"PromptExperimentNext": "",
"PersonalityFormat": "{{char}}'s personality: {{personality}}",
"ScenarioFormat": "Dialogue scenario: {{scenario}}",
"Settings": {
"RenewAlways": true,
"RetryRegenerate": false,
"PromptExperiments": true,
"SystemExperiments": true,
"PreventImperson": true,
"AllSamples": false,
"NoSamples": false,
"StripAssistant": false,
"StripHuman": false,
"PassParams": false,
"ClearFlags": true,
"PreserveChats": false,
"LogMessages": true,
"FullColon": true,
"padtxt": "1000,1000,15000",
"xmlPlot": true,
"SkipRestricted": false,
"Artifacts": false,
// 如果你遇到了 [message]: 'spawn Unknown system error -8' 错误(已知 Termux 中会发生),
// 请把这行的 true 改成 false 以关闭预读取。
"Superfetch": true
}
}
/*
BufferSize
* How many characters will be buffered before the AI types once
* lower = less chance of `PreventImperson` working properly
---
SystemInterval
* How many messages until `SystemExperiments alternates`
---
Other settings
* https://gitgud.io/ahsk/clewd/#defaults
* and
* https://gitgud.io/ahsk/clewd/-/blob/master/CHANGELOG.md
*/
稍后你就可以运行了:
up to date in 418ms
clewd修改版 v4.8(11) by tera
http://127.0.0.1:8144/v1
AllSamples: false
Artifacts: false
ClearFlags: true
FullColon: true
LogMessages: true
NoSamples: false
PassParams: true
PreserveChats: true
PreventImperson: true
PromptExperiments: true
RenewAlways: true
RetryRegenerate: false
SkipRestricted: false
StripAssistant: false
StripHuman: false
Superfetch: true
SystemExperiments: true
padtxt: 1000,1000,15000
xmlPlot: true
superfetch [found] bin/clewd-superfetch-linux-amd64
(index: 1) Logged in {
name: 'xxxxxxxx',
mail: 'xxxxxxxx@gmail.com',
cookieModel: 'claude-3-5-sonnet-20241022',
capabilities: [ 'claude_pro', 'chat', [length]: 2 ]
}
附: Clewd 的设置列表
Info
这就是 config.js
中的 "Settings" 块内的选项。
大多数情况下,如果你看不懂下面这些是什么意思,你就不需要修改这些设置。
SettingName: (DEFAULT)/opt1/opt2...
-
Superfetch
: (true)/false- true will use an alternate method to get past the
We are unable to serve your request
error - false won't use this method and you may get the error again
- true will use an alternate method to get past the
-
PreventImperson
: (true)/false- true trims the bot reply immediately if he says "Human:", "Assistant:", "H:" or "A:"
- making it so it doesn't hallucinate speaking as you (chance of missing some spicy things)
-
PromptExperiments
: (true)/false- true is an alternative way to send your prompt to the AI
- experiment before setting to false
- incompatible with
RenewAlways
set to false
-
RetryRegenerate
: (false)/true- true uses the AI's own retry mechanism when you regenerate on your frontend
- instead of a new conversation
- experiment with it
-
SystemExperiments
: (true)/false- only has any effect when
RenewAlways
is false - true alternates between Main+Jailbreak+User and Jailbreak+User
- false doesn't alternate
- only has any effect when
-
RenewAlways
: (true)/false- true makes a new conversation context each time
- false tries to reutilize the same old conversation, sending only the actual last message each time, taking into consideration
SystemExperiments
(will not work properly unless your Main is the first system prompt and your Jailbreak is the last)
-
StripAssistant
: (false)/true- true strips the "Assistant:" prefix from the last assistant message (if it's the last message)
-
StripHuman
: (false)/true- true strips the "Human:" prefix from the last human message (if it's the last message)
-
AllSamples
: (false)/true- mutually exclusive with
NoSamples
- true converts all real dialogues to "sample dialogues" (except the last Assistant and Human messages)
- you're "H" and the AI is "A"
- whatever the AI replies with is kept (only outgoing)
- see this for more information
- Human->H
- Assistant->A
- mutually exclusive with
-
NoSamples
: (false)/true- mutually exclusive with
AllSamples
- true converts all "sample dialogues" to real dialogues
- you're "Human" and the AI is "Assistant"
- whatever the AI replies with is kept (only outgoing)
- see this for more information
- H->Human
- A->Assistant
- mutually exclusive with
-
LogMessages
: (false)/true- true logs prompt and reply to
log.txt
- true logs prompt and reply to
-
ClearFlags
: (false)/true- possibly snake-oil
- clears your warnings
-
PassParams
: (false)/true- true will send the temperature you set on your frontend
- only values under <=1.0 >= 0.1
- this could get your account banned
- if clewd stops working, set to false
-
PreserveChats
: (false)/true- true prevents the deletion of old chats at any point