fix: NHK 源配置改 ja(seed 写错了); translate_article 加强 lang_src 兜底

backend/app/scripts/seed_sources.py

@@ -64,7 +64,7 @@ SEEDS = [
         "kind": "rss",
         "url": "https://www3.nhk.or.jp/rss/news/cat0.xml",
         "region": "asia",
-        "language_src": "en",
+        "language_src": "ja",
         "priority": 70,
         "fetch_interval_min": 60,
         "translate_to": "zh",

backend/app/workers/pipeline.py

@@ -155,7 +155,12 @@ async def translate_article(article_id: int) -> None:
             return
         title = art.title
         body_text = (art.body_text or "")[:TRANSLATE_BODY_MAX]
-        lang_src = art.lang_src or "auto"
+        # lang_src 优先级:article.lang_src > source.language_src > "auto"
+        # (article 入库时已经优先用了 feedparser 的 lang,这里再做一次兜底)
+        if not art.lang_src and art.source and art.source.language_src:
+            lang_src = art.source.language_src
+        else:
+            lang_src = art.lang_src or "auto"
         target = "zh"
         article_id_ref = art.id
 

scripts/_check_nhk.py

@@ -0,0 +1,61 @@
+"""查 NHK 源配置 + 已入库文章 lang_src 实际值。"""
+import os, paramiko
+PW = os.environ["REMOTE_PASS"]
+c = paramiko.SSHClient()
+c.set_missing_host_key_policy(paramiko.AutoAddPolicy())
+c.connect("207.57.129.228", port=19717, username="root", password=PW, timeout=15, allow_agent=False, look_for_keys=False)
+def run(cmd, t=15):
+    si, so, se = c.exec_command(cmd, timeout=t)
+    out = so.read().decode("utf-8", "replace")
+    err = se.read().decode("utf-8", "replace")
+    rc = so.channel.recv_exit_status()
+    if out: print(out, end="")
+    return out
+
+# 1) NHK 源配置
+print("--- 1. NHK 源配置 ---")
+run("docker exec news-aggregator-postgres-1 psql -U news -d news -c \"SELECT id, slug, language_src FROM sources WHERE slug = 'nhk-world';\"")
+
+# 2) 实际入库的 NHK 文章 lang_src 分布
+print("\n--- 2. 已入库 NHK 文章 lang_src 分布 ---")
+run("docker exec news-aggregator-postgres-1 psql -U news -d news -c \"SELECT lang_src, count(*) FROM articles WHERE source_id = (SELECT id FROM sources WHERE slug = 'nhk-world') GROUP BY 1;\"")
+
+# 3) 看 NHK RSS feed 实际的 <language> 字段
+print("\n--- 3. NHK RSS 实际 language 字段 ---")
+script = '''
+import feedparser, httpx
+async def main():
+    f = feedparser.parse("https://www3.nhk.or.jp/rss/news/cat0.xml")
+    print("feed.feed.language:", f.feed.get("language"))
+    if f.entries:
+        e = f.entries[0]
+        print("entry.language:", e.get("language"))
+        print("title:", e.title)
+asyncio.run(main())
+'''
+import base64
+b64 = base64.b64encode(script.encode()).decode()
+run("docker exec news-aggregator-worker-1 sh -c 'echo " + b64 + " | base64 -d > /app/_t.py'")
+# 修语法:用 3 撇号
+script = (
+    "import feedparser, httpx, asyncio\n"
+    "async def main():\n"
+    "    f = feedparser.parse('https://www3.nhk.or.jp/rss/news/cat0.xml')\n"
+    "    print('feed.feed.language:', f.feed.get('language'))\n"
+    "    if f.entries:\n"
+    "        e = f.entries[0]\n"
+    "        print('entry.language:', e.get('language'))\n"
+    "        print('title:', e.title)\n"
+    "asyncio.run(main())\n"
+)
+b64 = base64.b64encode(script.encode()).decode()
+run("docker exec news-aggregator-worker-1 sh -c 'echo " + b64 + " | base64 -d > /app/_t.py'")
+run("docker exec -w /app news-aggregator-worker-1 python /app/_t.py 2>&1 | tail -10", t=20)
+
+# 4) 看翻译 service 收到的 source 是什么(我打一行新文章,看 lang_src 传到 service)
+print("\n--- 4. service.translate 实际调用时 source 参数是什么? ---")
+# 看 translate_article 代码
+out = run("docker exec news-aggregator-worker-1 python -c 'import app.workers.pipeline; import inspect; print(inspect.getsource(app.workers.pipeline.translate_article))' 2>&1 | grep -E 'lang_src|translate\\(' | head -10")
+print(out)
+
+c.close()