feat(news-harvester): panel 重新采集 (refetch by URL)

- platform.sh nh_refetch <url>: run scripts/refetch.js in the skill dir
  with NH_LOCK_DIR wired; refetch resolves the owning source itself.
- web /refetch page (GET form + POST background thread, 5s auto-refresh
  while running, last-run output shown); nav link added.

bin/platform.sh

@@ -445,6 +445,18 @@ nh_quit_instance() {  # <source> 回收该源所属站点的共享浏览器实
   return 0
 }
 
+nh_refetch() {  # <url>  单篇重新采集:按 URL 匹配源→站点实例抓取→覆盖原文→单篇重推
+  local url="$1" tools_dir
+  [ -n "$url" ] || { echo "用法: nh_refetch <文章URL>"; return 2; }
+  tools_dir="$(nh_skill_dir)"
+  if [ ! -f "$tools_dir/scripts/refetch.js" ]; then
+    echo "❌ 技能缺少 scripts/refetch.js(请更新 news-hub-skills 技能仓)" >&2
+    return 1
+  fi
+  # NH_LOCK_DIR 供脚本判断站点采集锁;PATH 补齐 launchd 环境(node 在 homebrew)
+  ( cd "$tools_dir" && NH_LOCK_DIR="$WORK_DIR/runlocks" node scripts/refetch.js "$url" )
+}
+
 nh_login_preflight() {  # <source>
   local source="$1" creds tools_dir
   creds="$(nh_login_creds_file "$source" || true)"
@@ -454,8 +466,7 @@ nh_login_preflight() {  # <source>
     echo "⚠️  已配置 $source 的登录凭据,但技能缺少 login-tools.js,跳过预检" >&2
     return 0
   fi
-  nh_ensure_instance "$source"
-  export NH_LOGIN_USER="$(jq -r '.user // empty' "$creds" 2>/dev/null)"
+  nh_ensure_instance "$source"  export NH_LOGIN_USER="$(jq -r '.user // empty' "$creds" 2>/dev/null)"
   export NH_LOGIN_PASS="$(jq -r '.password // empty' "$creds" 2>/dev/null)"
   if (cd "$tools_dir" && node scripts/login-tools.js check "$source") >/dev/null 2>&1; then
     echo "ℹ️  登录预检:$source 已登录(端口 ${CDP_PORT_OVERRIDE:-默认})"

web/app.py

@@ -1059,6 +1059,46 @@ def toggle_source(name):
     return redirect(url_for("index"))
 
 
+# ── 单篇重新采集(按 URL:覆盖原文 + 单篇重推) ──────────────────
+
+_refetch_lock = threading.Lock()
+_refetch_state = {}  # {"running": bool, "done": bool, "rc": int, "url": str, "output": str, "start": str}
+
+
+@app.route("/refetch", methods=["GET", "POST"])
+@require_auth
+def refetch_page():
+    if request.method == "POST":
+        url = request.form.get("url", "").strip()
+        if not re.match(r"^https?://\S+$", url):
+            flash("请输入合法的文章链接(http:// 或 https:// 开头)", "error")
+            return redirect(url_for("refetch_page"))
+        with _refetch_lock:
+            if _refetch_state.get("running"):
+                flash("已有一次重新采集在进行中,请等它结束(本页自动刷新进度)", "error")
+                return redirect(url_for("refetch_page"))
+            _refetch_state.clear()
+            _refetch_state.update({
+                "running": True, "done": False, "url": url,
+                "start": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
+            })
+
+        def _run():
+            rc, out, err = pl.call("nh_refetch", url, timeout=600)
+            with _refetch_lock:
+                _refetch_state.update({
+                    "running": False, "done": True, "rc": rc,
+                    "output": ((out or "") + (("\n" + err) if err else ""))[-4000:],
+                })
+
+        threading.Thread(target=_run, daemon=True).start()
+        flash("重新采集已启动(约 0.5-2 分钟,页面自动刷新显示结果)", "success")
+        return redirect(url_for("refetch_page"))
+    with _refetch_lock:
+        state = dict(_refetch_state)
+    return render_template("refetch.html", state=state)
+
+
 # ── 站点登录凭据管理(站点 = 源名第一段,同站多频道共享一份凭据/配方) ──
 
 _login_test_lock = threading.Lock()

web/templates/base.html

@@ -224,6 +224,7 @@
     <a href="{{ url_for('guide') }}">接入指南</a>
     <a href="{{ url_for('provider_list') }}">Provider</a>
     <a href="{{ url_for('login_list') }}">站点登录</a>
+    <a href="{{ url_for('refetch_page') }}">重新采集</a>
     <a href="{{ url_for('latest_content') }}">最近抓取</a>
   </nav>
   <div class="container">

web/templates/refetch.html

+{% extends "base.html" %}
+{% block title %}重新采集 — News Harvester{% endblock %}
+{% block extra_head %}
+{% if state and state.running %}
+<meta http-equiv="refresh" content="5">
+{% endif %}
+{% endblock %}
+{% block content %}
+
+<h1>重新采集(按链接)</h1>
+<p style="color:#666;font-size:13px">
+  粘贴一篇文章的原文链接,系统会<strong>重新抓取该篇正文与配图</strong>:已归档过的文章覆盖原
+  <code>原文.md</code>(保留中文摘要)、更新 registry/index,并<strong>单篇重推 News Hub</strong>
+  (后端按 sourceId+url 幂等更新,版本 +1);未归档过的链接作为新文章写入今天目录
+  (待补中文摘要后随日目录推送)。浏览器源走该站共享实例,站点正在采集时会拒绝执行,稍等重试即可。
+</p>
+
+{% with messages = get_flashed_messages() %}
+  {% if messages %}{% for m in messages %}<div class="flash">{{ m }}</div>{% endfor %}{% endif %}
+{% endwith %}
+
+<div class="card" style="max-width:860px">
+  <form method="POST" action="{{ url_for('refetch_page') }}" style="display:flex;gap:8px;flex-wrap:wrap">
+    <input type="text" name="url" placeholder="https://www.reuters.com/technology/…" required
+           style="flex:1;min-width:340px" value="{{ state.url or '' }}">
+    <button type="submit">🔁 重新采集</button>
+  </form>
+</div>
+
+{% if state %}
+<div class="card" style="max-width:860px">
+  <h2 style="margin-top:0">上次执行</h2>
+  <p>
+    <strong>链接:</strong> {{ state.url }}<br>
+    <strong>开始:</strong> {{ state.start or '—' }}
+    {% if state.running %} · <span style="color:#e6a23c">⏳ 进行中…(每 5 秒自动刷新)</span>
+    {% elif state.done %}
+      {% if state.rc == 0 %} · <span style="color:#67c23a">✅ 成功</span>
+      {% else %} · <span style="color:#f56c6c">✗ 失败(退出码 {{ state.rc }})</span>{% endif %}
+    {% endif %}
+  </p>
+  {% if state.output %}
+  <pre style="white-space:pre-wrap;background:#1e1e1e;color:#d4d4d4;padding:12px;border-radius:6px;font-size:12px;max-height:420px;overflow:auto">{{ state.output }}</pre>
+  {% endif %}
+</div>
+{% endif %}
+
+{% endblock %}