@cablate/banini-tracker 2.0.7 → 2.0.9

package/dist/db.js

@@ -19,6 +19,21 @@ export function getDb() {
 }
 function migrate(db) {
     db.exec(`
+    CREATE TABLE IF NOT EXISTS posts (
+      id TEXT PRIMARY KEY,
+      source TEXT NOT NULL DEFAULT 'facebook',
+      text TEXT NOT NULL DEFAULT '',
+      ocr_text TEXT NOT NULL DEFAULT '',
+      transcript_text TEXT NOT NULL DEFAULT '',
+      media_type TEXT NOT NULL DEFAULT 'text',
+      media_url TEXT NOT NULL DEFAULT '',
+      url TEXT NOT NULL DEFAULT '',
+      like_count INTEGER NOT NULL DEFAULT 0,
+      comment_count INTEGER NOT NULL DEFAULT 0,
+      post_timestamp TEXT NOT NULL,
+      fetched_at TEXT NOT NULL
+    );
+
     CREATE TABLE IF NOT EXISTS predictions (
       id INTEGER PRIMARY KEY AUTOINCREMENT,
       post_id TEXT NOT NULL,

package/dist/facebook.d.ts

@@ -3,6 +3,7 @@ export interface FacebookPost {
     source: 'facebook';
     text: string;
     ocrText: string;
+    captionText: string;
     timestamp: string;
     likeCount: number;
     commentCount: number;

package/dist/facebook.js

@@ -10,6 +10,7 @@ export async function fetchFacebookPosts(pageUrl, token, maxPosts = 3) {
         body: JSON.stringify({
             startUrls: [{ url: pageUrl }],
             resultsLimit: maxPosts,
+            captionText: true,
         }),
         signal: AbortSignal.timeout(180_000),
     });
@@ -23,11 +24,17 @@ export async function fetchFacebookPosts(pageUrl, token, maxPosts = 3) {
         const ocrTexts = (item.media ?? [])
             .map((m) => m.ocrText ?? '')
             .filter((t) => t.length > 0);
+        // captionText 可能在 media item 或頂層
+        const captionTexts = (item.media ?? [])
+            .map((m) => m.captionText ?? '')
+            .filter((t) => t.length > 0);
+        const captionText = captionTexts.join('\n') || item.captionText || '';
         return {
             id: `fb_${item.postId ?? item.id ?? ''}`,
             source: 'facebook',
             text: item.text ?? item.message ?? '',
             ocrText: ocrTexts.join('\n'),
+            captionText,
             timestamp: item.time ?? new Date().toISOString(),
             likeCount: item.likes ?? 0,
             commentCount: item.comments ?? 0,

package/dist/index.js

@@ -18,6 +18,7 @@ import { filterNewPosts as filterNew, markPostsSeen } from './seen.js';
 import { withRetry } from './retry.js';
 import { createTranscriber, transcribeVideoPosts } from './transcribe.js';
 import { recordPredictions, updateTracking } from './tracker.js';
+import { getDb } from './db.js';
 // ── Config ──────────────────────────────────────────────────
 const FB_PAGE_URL = 'https://www.facebook.com/DieWithoutBang/';
 const DATA_DIR = process.env.DATA_DIR || join(process.cwd(), 'data');
@@ -34,7 +35,7 @@ function fromFacebook(p) {
         source: 'facebook',
         text: p.text,
         ocrText: p.ocrText,
-        transcriptText: '',
+        transcriptText: p.captionText || '',
         timestamp: p.timestamp,
         likeCount: p.likeCount,
         replyCount: p.commentCount,
@@ -81,17 +82,56 @@ async function runInner(opts) {
         console.log('沒有新貼文，結束');
         return;
     }
-    // 2.5. 影片轉錄
+    // 2.5. 影片轉錄（captionText 有值則跳過 Groq）
     const transcriberType = (process.env.TRANSCRIBER ?? 'noop');
     const transcriber = createTranscriber(transcriberType);
     if (transcriber.name !== 'noop') {
-        const transcripts = await transcribeVideoPosts(newPosts, transcriber);
-        for (const p of newPosts) {
-            const result = transcripts.get(p.id);
-            if (result)
-                p.transcriptText = result.text;
+        const needsTranscribe = newPosts.filter((p) => !p.transcriptText);
+        if (needsTranscribe.length > 0) {
+            const transcripts = await transcribeVideoPosts(needsTranscribe, transcriber);
+            for (const p of needsTranscribe) {
+                const result = transcripts.get(p.id);
+                if (result)
+                    p.transcriptText = result.text;
+            }
         }
     }
+    // 2.6. 貼文入庫
+    try {
+        const db = getDb();
+        const upsertPost = db.prepare(`
+      INSERT INTO posts (id, source, text, ocr_text, transcript_text, media_type, media_url, url, like_count, comment_count, post_timestamp, fetched_at)
+      VALUES (@id, @source, @text, @ocr_text, @transcript_text, @media_type, @media_url, @url, @like_count, @comment_count, @post_timestamp, @fetched_at)
+      ON CONFLICT(id) DO UPDATE SET
+        transcript_text = CASE WHEN excluded.transcript_text != '' THEN excluded.transcript_text ELSE posts.transcript_text END,
+        like_count = excluded.like_count,
+        comment_count = excluded.comment_count
+    `);
+        const now = new Date().toISOString();
+        const insertMany = db.transaction(() => {
+            for (const p of newPosts) {
+                upsertPost.run({
+                    id: p.id,
+                    source: p.source,
+                    text: p.text,
+                    ocr_text: p.ocrText,
+                    transcript_text: p.transcriptText,
+                    media_type: p.mediaType,
+                    media_url: p.mediaUrl,
+                    url: p.url,
+                    like_count: p.likeCount,
+                    comment_count: p.replyCount,
+                    post_timestamp: p.timestamp,
+                    fetched_at: now,
+                });
+            }
+        });
+        insertMany();
+        console.log(`[DB] 已存入 ${newPosts.length} 篇貼文`);
+    }
+    catch (err) {
+        console.error(`[DB] 貼文入庫失敗: ${err instanceof Error ? err.message : err}`);
+    }
     // 按時間從新到舊排序
     newPosts.sort((a, b) => new Date(b.timestamp).getTime() - new Date(a.timestamp).getTime());
     // 標記當天貼文

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@cablate/banini-tracker",
-  "version": "2.0.7",
+  "version": "2.0.9",
   "description": "巴逆逆反指標追蹤器 — 常駐排程 + CLI 雙模式",
   "type": "module",
   "bin": {