网页进入后台再进入会自动刷新，切程序什么的都会，用的docker部署。还有就是有时候，聊天过久会直接不回复

Clear and concise description of the problem

希望能跟ChatGPT 一样有深浅主题模式切换，现在的夜间模式白天看不太清楚

Suggested solution

希望能跟ChatGPT 一样有深浅主题模式切换，现在的夜间模式白天看不太清楚

Alternative

No response

Additional context

No response

Validations

• Follow our Code of Conduct
• Read the Contributing Guide.
• Check that there isn't already an issue that reports the same bug to avoid creating a duplicate.

如题

environment:
Linux instance-20211025-1400 5.15.0-1029-oracle #35~20.04.1-Ubuntu SMP Wed Jan 25 10:16:03 UTC 2023 aarch64 aarch64 aarch64 GNU/Linux
npm version v9.6.0
node version v16.19.1

error log :

root@instance:~/chatgpt/chatgpt-demo# npm run dev

> [email protected] dev
> astro dev

file:///root/chatgpt/chatgpt-demo/node_modules/node-fetch-native/dist/shared/node-fetch-native.2b047dc1.mjs:432
                            throw new TypeError(`${context} is not a ReadableStream.`);
                                  ^

TypeError: First parameter has member 'readable' that is not a ReadableStream.
    at assertReadableStream (file:///root/chatgpt/chatgpt-demo/node_modules/node-fetch-native/dist/shared/node-fetch-native.2b047dc1.mjs:432:21)
    at convertReadableWritablePair (file:///root/chatgpt/chatgpt-demo/node_modules/node-fetch-native/dist/shared/node-fetch-native.2b047dc1.mjs:3509:11)
    at ReadableStream.pipeThrough (file:///root/chatgpt/chatgpt-demo/node_modules/node-fetch-native/dist/shared/node-fetch-native.2b047dc1.mjs:3584:33)
    at fetchFinale (/root/chatgpt/chatgpt-demo/node_modules/undici/lib/fetch/index.js:973:52)
    at mainFetch (/root/chatgpt/chatgpt-demo/node_modules/undici/lib/fetch/index.js:773:5)
    at processTicksAndRejections (node:internal/process/task_queues:96:5)

docker镜像呢

Clear and concise description of the problem

Add a copy button to the top right corner of the code block

Suggested solution

Add a copy button to the code block

When used on the official website, the code block has a copy button, hoping to add this function

Alternative

No response

Additional context

No response

Validations

• Follow our Code of Conduct
• Read the Contributing Guide.
• Check that there isn't already an issue that reports the same bug to avoid creating a duplicate.

请问这个报错是什么意思

Hello!

Thank you for creating this demonstration.

Could you please tell me the proper method for deploying it on Vercel? I attempted to do so but it did not work.

在centos上部署很复杂啊。

设置和安装依赖项npm i
制作一个副本.env.example，然后将其重命名为.env

添加到.env OPENAI_API_KEY=sk-xxx...
运行应用 npm run dev

这四步成功了，打开网站，403

不合适小白一键部署，大佬出个完整教程吧，在centos上部署。

请问有人build成功了吗？运行run build，有几个warning,但是看不到dist文件夹

Describe the bug

error！How do I fix it?

Reproduction

error

System Info

Cause {   "errno": -4078,   "code": "ECONNREFUSED",   "syscall": "connect",   "address": "127.0.0.1",   "port": 7890 }

Used Package Manager

npm

Validations

• Follow our Code of Conduct
• Read the Contributing Guide.
• Check that there isn't already an issue that reports the same bug to avoid creating a duplicate.
• Check that this is a concrete bug. For Q&A, please open a GitHub Discussion instead.
• The provided reproduction is a minimal reproducible of the bug.

RT

Describe the bug

机器 Mac mini M1

❯ node -v
v16.13.2

❯ npm -v
8.3.0

按照 readme 操作之后，运行不起来

❯ npm run dev

[email protected] dev
astro dev

(node:55714) ExperimentalWarning: stream/web is an experimental feature. This feature could change at any time
(Use node --trace-warnings ... to show where the warning was created)
🚀 astro v2.0.17 started in 40ms

┃ Local http://localhost:3000/
┃ Network use --host to expose

file:///Users/ming/code/chatgpt-demo/node_modules/node-fetch-native/dist/shared/node-fetch-native.2b047dc1.mjs:432
throw new TypeError(${context} is not a ReadableStream.);
^

TypeError: First parameter has member 'readable' that is not a ReadableStream.
at assertReadableStream (file:///Users/ming/code/chatgpt-demo/node_modules/node-fetch-native/dist/shared/node-fetch-native.2b047dc1.mjs:432:21)
at convertReadableWritablePair (file:///Users/ming/code/chatgpt-demo/node_modules/node-fetch-native/dist/shared/node-fetch-native.2b047dc1.mjs:3509:11)
at ReadableStream.pipeThrough (file:///Users/ming/code/chatgpt-demo/node_modules/node-fetch-native/dist/shared/node-fetch-native.2b047dc1.mjs:3584:33)
at fetchFinale (/Users/ming/code/chatgpt-demo/node_modules/undici/lib/fetch/index.js:973:52)
at mainFetch (/Users/ming/code/chatgpt-demo/node_modules/undici/lib/fetch/index.js:773:5)
at processTicksAndRejections (node:internal/process/task_queues:96:5)

Reproduction

1

System Info

1

Used Package Manager

npm

Validations

• Follow our Code of Conduct
• Read the Contributing Guide.
• Check that there isn't already an issue that reports the same bug to avoid creating a duplicate.
• Check that this is a concrete bug. For Q&A, please open a GitHub Discussion instead.
• The provided reproduction is a minimal reproducible of the bug.

the code is not working GHSA-5ff8-jcf9-fw62

大佬 我可以分担服务器费用 跟梯子费用 能加个好友吗 discord vx qq 都可以

root@ # ./dist/server/entry.mjs
-bash: ./dist/server/entry.mjs: No such file or directory

It can turn out to be a public service rather than self use.

还要等他全部生成完毕后手动滑倒底部看完整回答，能优化这个步骤吗

Hello, may we please add a new feature that allows the export of conversations using markdown format and a click-to-copy function when clicking a message/conversation?

And can we use the official OpenAI package and typing? https://platform.openai.com/docs/api-reference/chat/create

Thank you.

超级快啊

Clear and concise description of the problem

The front end will always show that the AI is Thinking instead of throwing an exception that exceeds the length limit. This interaction can easily confuse users.

Suggested solution

Determine whether the message length exceeds the limit, or handle the error message returned by openai.

Alternative

No response

Additional context

No response

Validations

• Follow our Code of Conduct
• Read the Contributing Guide.
• Check that there isn't already an issue that reports the same bug to avoid creating a duplicate.

建议api key 从环境变量中取

反馈

1. 输出内容很长的时候，滚动条不会自往下滚
2. 光标位置有问题，不能继续输入得点一下对话框

功能需求

1. 能返回图片就好了
2. 可否增加一个自定义API的功能

输入能放到下面固定么
建议实现多行输入

https://stats.ddiu.io/app.js

!
function() {
    "use strict"; !
    function(t) {
        var e = t.screen,
        n = e.width,
        r = e.height,
        a = t.navigator.language,
        i = t.location,
        o = t.localStorage,
        c = t.document,
        u = t.history,
        s = i.hostname,
        f = i.pathname,
        l = i.search,
        d = c.currentScript;
        if (d) {
            var v, p = function(t, e) {
                return Object.keys(e).forEach((function(n) {
                    void 0 !== e[n] && (t[n] = e[n])
                })),
                t
            },
            h = function(t, e, n) {
                var r = t[e];
                return function() {
                    for (var e = [], a = arguments.length; a--;) e[a] = arguments[a];
                    return n.apply(null, e),
                    r.apply(t, e)
                }
            },
            m = function() {
                return o && o.getItem("umami.disabled") || E &&
                function() {
                    var e = t.doNotTrack,
                    n = t.navigator,
                    r = t.external,
                    a = "msTrackingProtectionEnabled",
                    i = e || n.doNotTrack || n.msDoNotTrack || r && a in r && r[a]();
                    return "1" == i || "yes" === i
                } () || A && !N.includes(s)
            },
            g = "data-",
            b = "false",
            y = d.getAttribute.bind(d),
            w = y(g + "website-id"),
            S = y(g + "host-url"),
            k = y(g + "auto-track") !== b,
            E = y(g + "do-not-track"),
            T = y(g + "css-events") !== b,
            A = y(g + "domains") || "",
            N = A.split(",").map((function(t) {
                return t.trim()
            })),
            j = (S ? S.replace(/\/$/, "") : d.src.split("/").slice(0, -1).join("/")) + "/api/collect",
            x = n + "x" + r,
            O = /^umami--([a-z]+)--([\w]+[\w-]*)$/,
            K = "[class*='umami--']",
            L = {},
            D = "" + f + l,
            P = c.referrer,
            $ = function() {
                return {
                    website: w,
                    hostname: s,
                    screen: x,
                    language: a,
                    url: D
                }
            },
            _ = function(t, e) {
                var n;
                if (!m()) return fetch(j, {
                    method: "POST",
                    body: JSON.stringify({
                        type: t,
                        payload: e
                    }),
                    headers: p({
                        "Content-Type": "application/json"
                    },
                    (n = {},
                    n["x-umami-cache"] = v, n))
                }).then((function(t) {
                    return t.text()
                })).then((function(t) {
                    return v = t
                }))
            },
            q = function(t, e, n) {
                return void 0 === t && (t = D),
                void 0 === e && (e = P),
                void 0 === n && (n = w),
                _("pageview", p($(), {
                    website: n,
                    url: t,
                    referrer: e
                }))
            },
            z = function(t, e, n, r) {
                return void 0 === n && (n = D),
                void 0 === r && (r = w),
                _("event", p($(), {
                    website: r,
                    url: n,
                    event_name: t,
                    event_data: e
                }))
            },
            C = function(t) {
                var e = t.querySelectorAll(K);
                Array.prototype.forEach.call(e, I)
            },
            I = function(t) {
                var e = t.getAttribute.bind(t); (e("class") || "").split(" ").forEach((function(n) {
                    if (O.test(n)) {
                        var r = n.split("--"),
                        a = r[1],
                        o = r[2],
                        c = L[n] ? L[n] : L[n] = function(n) {
                            "click" !== a || "A" !== t.tagName || n.ctrlKey || n.shiftKey || n.metaKey || n.button && 1 === n.button || e("target") ? z(o) : (n.preventDefault(), z(o).then((function() {
                                var t = e("href");
                                t && (i.href = t)
                            })))
                        };
                        t.addEventListener(a, c, !0)
                    }
                }))
            },
            J = function(t, e, n) {
                if (n) {
                    P = D;
                    var r = n.toString(); (D = "http" === r.substring(0, 4) ? "/" + r.split("/").splice(3).join("/") : r) !== P && q()
                }
            };
            if (!t.umami) {
                var M = function(t) {
                    return z(t)
                };
                M.trackView = q,
                M.trackEvent = z,
                t.umami = M
            }
            if (k && !m()) {
                u.pushState = h(u, "pushState", J),
                u.replaceState = h(u, "replaceState", J);
                var V = function() {
                    "complete" === c.readyState && (q(), T && (C(c), new MutationObserver((function(t) {
                        t.forEach((function(t) {
                            var e = t.target;
                            I(e),
                            C(e)
                        }))
                    })).observe(c, {
                        childList: !0,
                        subtree: !0
                    })))
                };
                c.addEventListener("readystatechange", V, !0),
                V()
            }
        }
    } (window)
} ();

这个js会向作者api发送域名，客户端分辨率等不算特别敏感的信息。虽然说不是很敏感，也明白可能是为了统计，但是这样做也是很不妥的吧。

Thanks for the great work.

According to https://platform.openai.com/docs/guides/chat/introduction

openai.ChatCompletion.create(
  model="gpt-3.5-turbo",
  messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Who won the world series in 2020?"},
        {"role": "assistant", "content": "The Los Angeles Dodgers won the World Series in 2020."},
        {"role": "user", "content": "Where was it played?"}
    ]
)

Seen like we can specify a "system role" to apply on the assistant role?
If that's so, maybe we can add a box in the beginning of the chat to let user write a role for the assistant?

牛

As the title

希望能实现类似openai官网的功能：保存和管理历史会话、继续历史会话
通过询问chatgpt，应该是可以实现的，具体包括：

1. 在开始一个新会话时，使用UUID模块指定一个新的会话ID，并保存在本地数据库中
2. 在数据库中保存每个会话ID对应的对话内容
3. 如果想继续历史会话，在调用API接口时指定历史会话ID即可

Clear and concise description of the problem

增加类似官网的new chat功能，有个侧边栏可以保留历史聊天记录。

Suggested solution

增加类似官网的new chat功能，有个侧边栏可以保留历史聊天记录。

Alternative

No response

Additional context

No response

Validations

• Follow our Code of Conduct
• Read the Contributing Guide.
• Check that there isn't already an issue that reports the same bug to avoid creating a duplicate.

作者的工作真的做的很棒！

访问作者贴出的demo链接，发现回答的速度特别快以及回答的内容都是完整的（chatgpt网页版总是会限制每次回答的字数）

想请问下作者哪些因素和技术决定了回答的速度和回答的内容字数呢？

还有作者的demo链接用的是您自己的api key吗，这样大家都可以访问的话，18美元的免费额度是不是很快就会用完呀😂

宝塔node项目新建里部署，端口填写错误 502。 这个项目正确端口应该是多少？

Hello, it has come to my attention that Katext poses serious security vulnerabilities which have yet to be addressed. As a precaution, I advise using alternative, related packages that are consistently updated and maintained at a high level.

加了key以后，一直connect error的错误。求救。。。

输入完文字回车以后ai 回答完可以自动focus 在输入框中，然后就是

输入框上的建议也可以屏蔽掉。

TypeError: fetch failed
at Object.fetch (node:internal/deps/undici/undici:11413:11)
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async Module.post (D:\new_phpstudy_pro\WWW\chatgpt-demo-main\src\pages\api\generate.ts:14:22)
at async call (file:///D:/new_phpstudy_pro/WWW/chatgpt-demo-main/node_modules/astro/dist/core/endpoint/index.js:71:20)
at async call (file:///D:/new_phpstudy_pro/WWW/chatgpt-demo-main/node_modules/astro/dist/core/endpoint/dev/index.js:15:10)
at async handleRoute (file:///D:/new_phpstudy_pro/WWW/chatgpt-demo-main/node_modules/astro/dist/vite-plugin-astro-server/route.js:117:20)
at async run (file:///D:/new_phpstudy_pro/WWW/chatgpt-demo-main/node_modules/astro/dist/vite-plugin-astro-server/request.js:46:14)
at async runWithErrorHandling (file:///D:/new_phpstudy_pro/WWW/chatgpt-demo-main/node_modules/astro/dist/vite-plugin-astro-server/controller.js:65:5)
at async handleRequest (file:///D:/new_phpstudy_pro/WWW/chatgpt-demo-main/node_modules/astro/dist/vite-plugin-astro-server/request.js:40:3)

Cause:
ConnectTimeoutError: Connect Timeout Error
at onConnectTimeout (D:\new_phpstudy_pro\WWW\chatgpt-demo-main\node_modules\undici\lib\core\connect.js:182:24)
at D:\new_phpstudy_pro\WWW\chatgpt-demo-main\node_modules\undici\lib\core\connect.js:129:46
at Immediate._onImmediate (D:\new_phpstudy_pro\WWW\chatgpt-demo-main\node_modules\undici\lib\core\connect.js:168:33)
at process.processImmediate (node:internal/timers:476:21)

现在开放了3.5的api，不进行登录权限的控制，很容易把额度刷完，只需要加一个简单的验证，然后是否可以把部署方式搞到railway里面

frist： 非常感谢 developer 奉献了这么一个好的项目， 也为了节省大家的时间，提供一下部署的步骤：

# ---- Dependencies ----
FROM node:19.7.0 AS dependencies
WORKDIR /app
COPY package.json /app
RUN npm install

# ---- RUN ----
FROM dependencies AS RUN
WORKDIR /app
COPY . /app
COPY .env.example /app/.env
ENV PORT=3000
VOLUME ["./env"]
CMD [ "npm", "run", "dev", "--", "--port", "3000", "--host", "0.0.0.0"]

如果还是不会部署的话 这边帮忙部署环境；只需要提供域名即可；Q:1194191181