feat: add service to fetch and generate podcast content

services/fetch_and_generate.ts

@@ -0,0 +1,137 @@
+import Parser from "rss-parser";
+import { openAI_ClassifyFeed, openAI_GeneratePodcastContent } from "./llm";
+import { generateTTS } from "./tts";
+import { saveEpisode, markAsProcessed } from "./database";
+import { updatePodcastRSS } from "./podcast";
+import crypto from "crypto";
+
+interface FeedItem {
+  id: string;
+  title: string;
+  link: string;
+  pubDate: string;
+  contentSnippet?: string;
+}
+
+import fs from "fs/promises";
+import path from "path";
+import { fileURLToPath } from "url";
+
+const __filename = fileURLToPath(import.meta.url);
+const __dirname = path.dirname(__filename);
+
+export async function batchProcess() {
+  const feedUrlsFile = import.meta.env["FEED_URLS_FILE"] ?? "feed_urls.txt";
+  const feedUrlsPath = path.resolve(__dirname, "..", feedUrlsFile);
+  let feedUrls: string[];
+  try {
+    const data = await fs.readFile(feedUrlsPath, "utf-8");
+    feedUrls = data
+      .split("\n")
+      .map((url) => url.trim())
+      .filter((url) => url.length > 0);
+  } catch (err) {
+    console.warn(`フィードURLファイルの読み込みに失敗: ${feedUrlsFile}`);
+    feedUrls = [];
+  }
+
+  // フィードごとに処理
+  for (const url of feedUrls) {
+    try {
+      await processFeedUrl(url);
+    } finally {
+      await updatePodcastRSS();
+    }
+  }
+
+  console.log("処理完了:", new Date().toISOString());
+}
+
+const processFeedUrl = async (url: string) => {
+  const parser = new Parser<FeedItem>();
+  const feed = await parser.parseURL(url);
+
+  // フィードのカテゴリ分類
+  const feedTitle = feed.title || url;
+  const category = await openAI_ClassifyFeed(feedTitle);
+  console.log(`フィード分類完了: ${feedTitle} - ${category}`);
+
+  // 昨日の記事のみフィルタリング
+  const yesterday = new Date();
+  yesterday.setDate(yesterday.getDate() - 1);
+
+  const yesterdayItems = feed.items.filter((item) => {
+    const pub = new Date(item.pubDate || "");
+    return (
+      pub.getFullYear() === yesterday.getFullYear() &&
+      pub.getMonth() === yesterday.getMonth() &&
+      pub.getDate() === yesterday.getDate()
+    );
+  });
+
+  if (yesterdayItems.length === 0) {
+    console.log(`昨日の記事が見つかりません: ${feedTitle}`);
+    return;
+  }
+
+  // ポッドキャスト原稿生成
+  console.log(`ポッドキャスト原稿生成開始: ${feedTitle}`);
+  const validItems = yesterdayItems.filter((item): item is FeedItem => {
+    return !!item.title && !!item.link;
+  });
+  const podcastContent = await openAI_GeneratePodcastContent(
+    feedTitle,
+    validItems,
+  );
+
+  // トピックごとの統合音声生成
+  const feedUrlHash = crypto.createHash("md5").update(url).digest("hex");
+  const categoryHash = crypto.createHash("md5").update(category).digest("hex");
+  const uniqueId = `${feedUrlHash}-${categoryHash}`;
+
+  const audioFilePath = await generateTTS(uniqueId, podcastContent);
+  console.log(`音声ファイル生成完了: ${audioFilePath}`);
+
+  // エピソードとして保存（各フィードにつき1つの統合エピソード）
+  const firstItem = yesterdayItems[0];
+  if (!firstItem) {
+    console.warn("アイテムが空です");
+    return;
+  }
+  const pub = new Date(firstItem.pubDate || "");
+
+  await saveEpisode({
+    id: uniqueId,
+    title: `${category}: ${feedTitle}`,
+    pubDate: pub.toISOString(),
+    audioPath: audioFilePath,
+    sourceLink: url,
+  });
+
+  console.log(`エピソード保存完了: ${category} - ${feedTitle}`);
+
+  // 個別記事の処理記録
+  for (const item of yesterdayItems) {
+    const itemId = item["id"] as string | undefined;
+    const fallbackId = item.link || item.title || JSON.stringify(item);
+    const finalItemId =
+      itemId && typeof itemId === "string" && itemId.trim() !== ""
+        ? itemId
+        : `fallback-${Buffer.from(fallbackId).toString("base64")}`;
+
+    if (!finalItemId || finalItemId.trim() === "") {
+      console.warn(`フィードアイテムのIDを生成できませんでした`, {
+        feedUrl: url,
+        itemTitle: item.title,
+        itemLink: item.link,
+      });
+      continue;
+    }
+
+    const already = await markAsProcessed(url, finalItemId);
+    if (already) {
+      console.log(`既に処理済み: ${finalItemId}`);
+      continue;
+    }
+  }
+};