我的朋友建议我从开源项目开始，还帮助我完成了最初的简单设置步骤。
我几乎在每一步都使用了 ChatGpt，从弄清楚如何设置、安装、将 API 连接到代码的含义，如何重写函数或更改大小。
现在我要详细介绍一下我构建的第一个项目。
第 1 步. 设置环境
摆茶具
在我开始之前，我得到了设置包管理器的建议，比如 tea 来处理开发环境。
? 这些词在这一点上仍然是未知的“包管理器”
sh <(curl https://tea.xyz)
# --- OR ---
# using brew
brew install tea
当我得到它时，茶会帮助安装 node、npm、vercel 和我开发所需的任何其他包。
所以合而为一。
使用 TypeScript 和 Tailwindcss 设置 Next.js
我有一个基本的了解，即我需要前端。
我被教导从生成一个新的 Next.js 项目开始。同样使用 TypeScript 和 Tailwind CSS，所以请遵循以下步骤：
npx create-next-app
# ---
# you'll be asked the following prompts
What is your project named? my-app
Would you like to add TypeScript with this project? Y/N
# select `Y` for typescript
Would you like to use ESLint with this project? Y/N
# select `Y` for ESLint
Would you like to use Tailwind CSS with this project? Y/N
# select `Y` for Tailwind CSS
Would you like to use the `src/ directory` with this project? Y/N
# select `N` for `src/` directory
What import alias would you like configured? `@/*`
# enter `@/*` for import alias
? 第 2 步。查找开源项目并在此基础上构建
我使用了这两个开源项目：
-
Twitter Alghoritm 所以我可以根据 Linkedin 的算法对用户的帖子输入进行评分。
2.twitterbio
为了分别设置和打开每个项目，我将它们以 zip 格式下载到我的计算机上
? Step.3 弄清楚代码
所以我有点被我所看到的震惊了，一开始我什么都想不通。
所以我做了什么，我向 ChatGPT 询问了我的应用程序的基本结构。
我把里面的每一个页面代码都抄下来，问它是做什么的，基本上就是问怎么修改。所以我开始有了一个更好的主意，应用程序前端在哪里，CSS 在哪里。
我仍然没有完全了解所有内容并且遗漏了这些东西，但我认为这绝对是快速学习。
我向 ChatGPT 提出的一些请求非常基本，现在对我来说真的很清楚，那时我问了所有问题，都是愚蠢的问题。
? Step 4 构建项目
因此，在了解了一些基础知识之后，我继续进行自己的更改。开始在这些项目之上构建应用程序。
它有两部分=排名+生成器
Linkedin 帖子生成器算法
排名是根据不同的标准对您的帖子进行排名，从而提高性能。
我使用以下函数为 Linkedin 已知的内容改编了算法：
-
检测多个主题标签的功能
-
检测图像或视频的功能
-
检测帖子中网址的功能
-
支持使用表情符号的帖子的功能
-
宣传负面内容的功能
-
像后期格式化一样优先中断的功能。
-
减少行长度的功能
-
提问功能
与 Twitter 算法相比，Linkedin 并不公开。
// function to detect multiple hashtags
function multipleHashtags({ post }: PostData): Rank {
const regex = /#[\w-]+/g;
const hashtags = post.match(regex);
const lowerCasePost = post.toLowerCase();
if (hashtags && hashtags.length > 3) {
return {
score: 0.5,
message: `Too many hashtags.`,
};
}
if (hashtags && hashtags.length <= 3) {
if (
lowerCasePost.includes("#follow") ||
lowerCasePost.includes("#comment") ||
lowerCasePost.includes("#like")
) {
return {
score: 0.5,
message: `Avoid using hashtags like "follow," "comment," or "like".`,
};
}
return {
score: 1,
message: `Combine general and specific hashtags.`,
};
}
return {
score: 1.0,
};
}
// function to detect image or video
function imageVideoBoost({ postMedia }: PostData): Rank {
const has_media = postMedia;
if (has_media) {
return {
score: 2.0,
// message: `Contains image or video.`,
};
}
return {
score: 1.0,
};
}
// function to detect urls in post
function postHasUrl({ post }: PostData): Rank {
const regex =
/https?:\/\/[\w-]+(\.[\w-]+)+([\w.,@?^=%&:/~+#-]*[\w@?^=%&/~+#-])?/g;
const urls = post.match(regex);
if (urls && urls.length > 0) {
return {
score: 0.5,
message: `Remove the link from post and add in comments.`,
};
}
return {
score: 1.0,
};
}
/**
* Function to favor posts that use emojis
*/
function emojis({ post, sentiment }: PostData): Rank {
const regex = new RegExp("[\uD800-\uDBFF][\uDC00-\uDFFF]", "g");
const emojis = post.match(regex) || [];
const totalMatches = emojis.length;
if (totalMatches > 0) {
return {
score: 1.5,
// message: `Included ${totalMatches} emojis in the post.`,
};
}
return {
score: 1,
message: "No emojis found in the post.",
type: "negative"
};
}
/**
* Promote negative content because it's more likely to go viral.
* Hide anything positive or uplifting.
*/
function sentiment({ post, sentiment }: PostData): Rank {
if (sentiment.comparative >= 0.5) {
if (sentiment.comparative > 1.5) {
return {
score: 1.5,
// message: `Exceptionally positive.`,
};
} else {
return {
score: 1.1,
// message: `Positive sentiment.`,
};
}
} else if (sentiment.comparative <= -0.5) {
if (sentiment.comparative < -1.5) {
return {
score: 0.5,
// message: `Exceptionally negative.`,
};
} else {
return {
score: 0.9,
// message: `Negative sentiment.`,
};
}
} else {
return {
score: 1,
};
}
}
/**
* Prioritize break like post formatting.
*/
function lineBreaks({ post, sentiment }: PostData): Rank {
const breaks = post.split(/\n\s*\n/);
const totalBreaks = breaks.length - 1;
if (totalBreaks >= 1) {
return {
score: 1.5,
// message: `Used ${totalBreaks} line breaks.`,
};
} else {
return {
score: 1,
message: `Add line breaks between the lines.`,
type: "negative"
};
}
}
/**
* Reduce line length
*/
function lineLength({ post }: PostData): Rank {
const lines = post.split('\n');
let score = 1.0;
for (let i = 0; i < lines.length; i++) {
if (lines[i].length > 200) {
return {
score: 0.9,
message: `Reduce line length to improve readability (200 characters).`,
};
}
}
return {
score: 1,
// message: `Good, keep line length 200 characters or less.`,
type: "positive"
};
}
/**
* Function to ask questions
*/
function questions({ post, sentiment }: PostData): Rank {
if (post.includes("?")) {
return {
score: 1.5,
// message: "Great! Questions can help to activate discussion"
};
} else {
return {
score: 1,
message: "Add questions to activate discussion",
type: "negative"
};
}
}
算法的用户界面
它检测上面代码中的所有功能，并针对其中的一些功能展示如何改进帖子。我没有为所有功能调整它。
return (
<>
<div>
<div className="slider bg-gray-300 h-4 rounded-full relative overflow-hidden">
<div
className={classNames(
"absolute top-0 transition-width duration-250 ease-linear h-20",
sliderColor
)}
style={{ width: percentage }}
/>
</div>
{/* <p className="explanation text-gray-600 italic text-sm mt-2">
Positive rankings result in greater reach
</p> */}
<ul className="mt-5 p-0">
{positive.map((item, index) => (
<li
className="positive text-green-600 flex items-center space-x-2 list-style-none my-5 text-sm"
key={`positive-${index}`}
>
<span>?</span>
<span>{item.message.replace(/\(\s*[+-]?\d+\s*\)/, '')}</span>
</li>
))}
{negative.map((item, index) => (
<li
className="negative text-red-600 flex items-center space-x-2 list-style-none my-1 text-sm"
key={`negative-${index}`}
>
<span>?</span>
<span>{item.message.replace(/\(\s*[+-]?\d+\s*\)/, '')}</span>
</li>
))}
</ul>
</div>
<style jsx>{`
.slider:after {
content: " ";
display: block;
width: 2px;
height: 20px;
position: absolute;
top: 0;
left: calc(25% - 1px);
background: #000;
}
`}</style>
</>
);
};
打开 AI Api 和提示生成器
我使用 handle Prompt 来生成帖子。还有类型过滤器，所以我有 5 个基于类型的不同提示。 我刚刚为此连接了我的 OpenAI API。
const handlePrompt = () => {
let prompt;
switch (vibe) {
下面提示
prompt = `Generate post using this prompt, based on ${post}. You are a LinkedinGPT, a large language model that generates viral posts for Linkedin. You are given a prompt of a post and must generate a post that is more likely to be liked and reposted than the original post.
The Linkedin algorithm contains boosts and demotions based on what you are writing. Positive boosts are:
- in each post add emoji
- 200 characters in sentence maximum
- Start each sentecnce from new line and ad numbers in first 2 lines
- add 3 hashtags which 2 are generic and one very specific (at the end) Tags relate to post theme
- add a question at the end of the post to start a discussion. Before the hashtags
- first two lines should be catchy
- Dont add links - links are not good.
- If post copied in the field contain some numbers keep them the same.
Add idea about which image or visual can be added at the end of the post (this text is not counted as part of post)
${post}
---
Generated post length must be more than 800-1200 characters
---
Between each line must be a space
---
Keep all mentions of people in there
---
Start the firs line from smth like: I did smth, In year, I do, Tired of, Sometimes it is just, A path toward, Because this is not,I've been struggling, (change the begginign depends on the context )
---
Add emoji if it fits
---
It should be a story`;
生成器接口
这是我的索引文件。来自帖子生成器。
<main>
<nav className="bg-blue-900 text-white ">
<div className="px-5">
<div className="max-w-5xl mx-auto">
<div className="flex justify-between items-center h-16 ">
<div className="flex items-center text-base ">
<a target="_blank"
href="https://www.linkedin.com/in/iuliia-shnai/"
rel="noreferrer"
className="text-white flex max-w-fit items-center justify-center space-x-2 text-xl"
>
<p>?•?</p>
</a>
</div>
</div>
</div>
</div>
</nav>
<section className="py-10 lg:py-20 ">
{/* bg-[url('/image1.svg')] */}
<div className="px-4">
<div className="max-w-5xl mx-auto">
<div className="w-full mx-auto">
<h1 className="text-6xl text-center font-bold pb-1 text-slate-900">
Linkedin Post Generator ?
</h1>
<p className="mt-3 mb-10 text-center">
See how your post performs and generate a better one with AI. Time to go viral. <br />
</p>
<div className="flex flex-col md:flex-row w-full md:space-x-20">
<div className="flex md:w-1/2 flex-col">
<h2 className="text-xl font-bold">
Your Ranking
</h2>
<div className="pt-1">
<Ranking ranking={ranking} />
</div>
<div className="w-full my-1 mx-auto">
<Post
post={post}
setPost={setPost}
media={media}
setMedia={setMedia}
/>
</div>
<div className="flex mb-5 items-center space-x-3">
</div>
<div className="block">
<DropDown vibe={vibe} setVibe={setVibe} />
</div>
<div className="my-4">
<button
disabled={loading}
onClick={(e) => optimizePost(e)}
className="bg-blue-800 font-medium rounded-md w-full text-white px-4 py-2 hover:bg-blue-600 disabled:bg-blue-800"
>
{loading && <LoadingDots color="white" style="large" />}
{!loading && `Generate new post `}
</button>
</div>
</div>
<div className="flex md:w-1/2 md:flex-col">
<Toaster
position="top-right"
reverseOrder={false}
toastOptions={{ duration: 2000 }}
/>
{optimizedPost && (
<div className="my-1">
<div className="flex justify-between items-center pb-2 border-b border-gray-300">
<h2 className="text-xl font-bold">
Your Generated Post
</h2>
</div>
<div className="max-w-2xl my-4 mx-auto">
<div
className="bg-white rounded-xl shadow-md p-4 hover:bg-gray-100 transition cursor-copy border"
onClick={() => {
navigator.clipboard.write([
new ClipboardItem({
"text/html": new Blob([optimizedPost], { type: "text/html" }),
}),
]);
toast("Post copied to clipboard", {
icon: "?",
});
}}
key={optimizedPost}
>
<p className="text-black-700" dangerouslySetInnerHTML={{ __html: optimizedPost }} />
</div>
</div>
</div>
)}
</div>
</div>
</div>
</div>
</div>
</section>
<div className="max-w-5xl mx-auto">
<Footer />
</div>
</main>
</>
);
}
? Step 5 推送项目
最后一步，我准备好推动了。
我在 GitHub 上创建存储库
$ git remote add origin ..
git branch -M main
git push -u origin main
并进一步在 Vercel 上创建帐户，用 Vercel 推送它并检查错误。
我推送的每一次更新
git add .
git commit -m “fix type”
git push
为了检查我使用的错误，所以我不会推送所有错误。ChatGPT 帮助修复了很多错误，而我什至不知道如何找到它们。
npm run build