JamesFlare1212

As GraphGPT works like make a request to GPT 3 ``` Given a prompt, extrapolate as many relationships as possible from it and provide a list of updates. If an...

https://github.com/acheong08/ChatGPT/commit/64f5f3db66cd3e6cb773379f900801a4ad8bc8ad

Listen as an OpenAI API?

1

I checked the document, I am thinking is that possible to make the GPTCache listen as an OpenAI like API? So we can just connect it to other service via...

CHATGPT_API_PREFIX with CORS Error

2

我试图用 `CHATGPT_API_PREFIX` 变量指向我用 go-chatgpt-api 搭建的 ChatGPT 逆向代理。但是遇到了 CORS 错误，即便我手动在搭建了的网页中人为插入 header，`Access-Control-Allow-Origin: *` 依旧无法解决问题。保险起见我还安装了一个 Chrome 插件来移除 CORS 限制（原理上似乎就是插入一个 `Access-Control-Allow-Origin: *` 头）依旧不起作用。 我看到有 2-3 个 Issue 都有此问题，照理说如果他们使用 CloudFlare Workers 反代现在的 fakeopen 的 ChatGPT...

[Feature] 优化长文本对话框的UI性能 | Optimizing the UI performance of long text dialog boxes

2

随着我测试更多Large Token Window的模型，比如Claude v1.3 100k。当文本长度达到30k Token以上后可以感到明显的卡顿，90k Token的时候基本上难以使用。我认为这是一个效率和优化的问题，以前可能不明显，但是现在越来越明显了。 ---- As I test more models with Large Token Window, such as Claude v1.3 100k, I can feel obvious lag when the text...

对于rate limit的处理

1

今天又测试了一下，发现缺少对Claude API rate_limit_error 的处理，没记错OpenAI 报错会发一个message之类的信息，里面就可以写上rate limit，或者我记得官方就有rate limit error来着。 我的想法是，当探测到返回字段是： ``` {"error":{"type":"rate_limit_error","message":"Number of concurrent connections to Claude exceeds your rate limit. Please try again, or contact [email protected] to discuss your options...

I catched a sample ``` https://www.xiaohongshu.com/explore/63cb99e3000000001b024b12?app_platform=ios&app_version=7.72.2&share_from_user_hidden=true&type=normal&xhsshare=WeixinSession&appuid=5f783fc3000000000101fb79&apptime=1674319435 ``` And ``` app_platform=ios app_version=7.72.2 share_from_user_hidden=true type=normal xhsshare=WeixinSession appuid=5f783fc3000000000101fb79 apptime=1674319435 ``` are all useless value, maybe trackers. I removed them and checked, they will...