建议: 添加客户端端速率限制,以防止 API 信用耗尽

顿! 喜欢这个项目。

公共AI应用的一大问题是bots可以轻松地将聊天端点相吐出,在几分钟后耗尽了所有者的OpenAI/Anthropic的API信用.

[[1](https://www.google.com/url?sa=E&q=https%3A%2F%2Fvertexaisearch.cloud.google.com%2Fgreening-api-redict%2FAUZIYQHEK1N23y2cScK4rzsG8lWTVlVlUCPNlWH3h53V4iBpEw=PqSOMi6x5FPWQZWLJK8 cTHDi0qr98Xv-Xe6r D2zIJw9JctZDHWT6KlHCQGCWTUTLBTVYTUDUAMLFS2D]]. 传统的费率限制(基于IP)往往过于复杂,无法设置起动套件(需要Redis/KV).

[1](https://www.google.com/url?w=%3A%2F%2Fvertexaisearch.cloud.google.com%2Fgreening-api-redict%2FAUZIYQHEK1N23y2cS8WTVlUCimjpNlWh3h53V4BpEw=PqSOMQSOMSOMMM6X0XX0X0X0X0X0X0X0X0X0X0F0X0X0X0X0X0X0BlHCQCHQWGWVUTVUTVUTUTUK0X0X0X0X0X0F0X0F0X0X0F0X0F0X0X0F0X0F0X0X0X0F0X0X0XF0X2XF0XF0X0X0X0X0XF0X0X0XF0X 解决方案:

我建了一个名为 [IronWall SDK] (https://www.google.com/url?sa=E&q=https%3A%2F%2Fgithub.com%2Fclein154%2Fronwall-sdk)的微小开放源代码库. 它使用客户端-Side Prof-work强制浏览器在发送消息前解决密码谜题.

1])][[4](https://www.googl.com/Chh3NNNH53VH5X0XA&K0Z0Z0XqqQQQQ%F%F%FQ%F%F%F%F%F%FQK%F [1](https://www.google.com/url?sa=E&q=https%3A%2F%2Fvertexaisearch.cloud.google.com%2Fgreening-api-redict%2FAUZIYQHEk1N23y2cScScK4rzsG8lWTVLUpNl?Wh3h53V4iBpEw=PqSOMI6x5FPWWqzWLJGK8 cThDi0qr98Xv-Xe6r D2zIJw9JctZdHwT6klHCQGCWUTLBtEvyTuDUAMLFS2D3D)]* 省钱:保护OPENAI API KE配额,没有设置后端. [1](https://www.google.com/url?sa=E&q=https%3A%2F%2Fvertexaisearch.cloud.google.com%2Fgreening-api-redict%2FAUZIYQHEK1N23y2cScK4rzsG8lWTVlUcCimjpNlWh3h53V4iBpEwQPqSOMIX5FPWQZWLJCK8 cTHDDi0qr98Xv-Xe6r D2zIJwTHW6KlHCQGCWTUTLBTUVYTUDUAMLFS2D%3D)] * Tin:3KB (vs Google reCTCHA).

建议:

我可以提交一个PR,在聊天输入器上添加一个可选的瓶子保护包装. 将为用户使用"Zero Config"(仅仅是一个API密钥)并保护钱包出框. 文档 : . . . . . . .