我用 Cloudflare 免费搭了一套 AI 内容流水线

复制# wrangler.toml
name = "ai-content-pipeline"
main = "src/index.js"
compatibility_date = "2026-04-30"

[triggers]
crons = ["0 0 * * *"]

复制export default {
  // 这是 Cron 定时触发的入口
  async scheduled(event, env, ctx) {
    // 1. 去源头抓数据 (这里拿 Hacker News 举例)
    const hnResponse = await fetch("https://hacker-news.firebaseio.com/v0/topstories.json");
    const storyIds = await hnResponse.json();

    // 取前 5 篇
    const top5Ids = storyIds.slice(0, 5);
    let articles = [];

    for (let id of top5Ids) {
      const itemRes = await fetch("https://hacker-news.firebaseio.com/v0/item/" + id + ".json");
      const item = await itemRes.json();
      articles.push(item);
    }

    // 把文章传给下一步处理 (见下文)
    await processArticles(articles, env);
  }
};

复制[ai]
binding = "AI"

复制async function processArticles(articles, env) {
  for (let article of articles) {
    if (!article.url) continue;

    const prompt =
      "请将以下新闻标题和链接整理成严格 JSON，不要输出 Markdown。\n" +
      "字段包括：\n" +
      "- title_zh：中文标题\n" +
      "- summary_zh：2-3 句话中文摘要，只根据标题和链接做保守判断\n" +
      "- why_it_matters：1-2 句话说明它为什么值得创作者或开发者关注\n" +
      "- tags：2-3 个分类标签数组\n\n" +
      "原文标题：" + article.title + "\n" +
      "原文链接：" + article.url;

    // 调免费的 Llama 3 8B 模型干活
    const aiResponse = await env.AI.run("@cf/meta/llama-3.1-8b-instruct-fast", {
      messages: [{ role: "user", content: prompt }]
    });

    const aiResult = aiResponse.response;
    console.log("AI 处理结果:", aiResult);

    // 接下来把它存起来
    await saveToDatabase(article, aiResult, env);
  }
}

复制npx wrangler r2 bucket create content-store

复制[[r2_buckets]]
binding = "BUCKET"
bucket_name = "content-store"

复制async function saveToStorage(article, env) {
  const fileKey = "raw/" + article.id + ".json";

  // 直接丢进 R2
  await env.BUCKET.put(fileKey, JSON.stringify(article));
  return fileKey;
}

复制DROP TABLE IF EXISTS articles;
CREATE TABLE articles (
  id INTEGER PRIMARY KEY,
  title TEXT,
  url TEXT,
  summary_zh TEXT,
  why_it_matters TEXT,
  tags TEXT,
  created_at DATETIME DEFAULT CURRENT_TIMESTAMP
);

复制npx wrangler d1 create content-db
npx wrangler d1 execute content-db --local --file=schema.sql
npx wrangler d1 execute content-db --remote --file=schema.sql

复制[[d1_databases]]
binding = "DB"
database_name = "content-db"
database_id = "终端里吐出来的那个ID"

复制async function saveToDatabase(article, aiResult, env) {
  // 假设我们解析了 AI 的结果
  const zhTitle = "AI 生成的中文标题";
  const summary = "2-3 句话中文摘要";
  const whyItMatters = "这条内容为什么值得看";
  const tags = "AI, 自动化, 资讯";

  // 用标准的 SQL 写入 D1
  await env.DB.prepare(
    "INSERT INTO articles (id, title, url, summary_zh, why_it_matters, tags) " +
    "VALUES (?, ?, ?, ?, ?, ?) " +
    "ON CONFLICT(id) DO NOTHING"
  ).bind(article.id, zhTitle, article.url, summary, whyItMatters, tags).run();
}

复制// 当访问 /api/articles 时触发
// 文件：functions/api/articles.js
export async function onRequest(context) {
  // 从绑定的 D1 里查数据
  const { results } = await context.env.DB.prepare(
    "SELECT * FROM articles ORDER BY created_at DESC LIMIT 20"
  ).all();

  return Response.json(results);
}

复制<!DOCTYPE html>
<html>
<head>
  <title>我的 AI 科技周刊</title>
</head>
<body>
  <h1>每日 AI 科技精选</h1>
  <div id="content"></div>

  <script>
    fetch("/api/articles")
      .then(res => res.json())
      .then(data => {
        const html = data.map(a =>
          "<div>" +
          "<h3><a href=\"" + a.url + "\">" + a.title + "</a></h3>" +
          "<p>摘要：" + a.summary_zh + "</p>" +
          "<p>为什么值得看：" + a.why_it_matters + "</p>" +
          "<p>标签：" + a.tags + "</p>" +
          "</div>"
        ).join("");
        document.getElementById("content").innerHTML = html;
      });
  </script>
</body>
</html>