打通流水线闭环: trends+metrics+optimize_sources → collector

三大数据流精准注入采集器:

1. 热点趋势(trends) -> collector
   _get_trend_context() 读取 trends.json
   热搜注入 LLM 选题 prompt

2. 历史表现(metrics) -> collector
   scheduler 同步指标后按field聚合写入 feedback.json
   collector 读取后高互动领域获优先级提升

3. AI策略(optimize_sources) -> collector
   从 DB SystemConfig 读取 AI 建议注入 prompt

platform/backend/app/core/scheduler.py

@@ -25,6 +25,15 @@ class TaskScheduler:
             logger.warning("Scheduler already started")
             return
         # 使用 CronTrigger 设置每日固定时间点
+        # 顺序: 搜索缓存(01:00)→采集(01:30)→趋势(03:00)→生成(03:30)→审查(04:30)→源优化(05:00)→指标(06:00)
+        self.scheduler.add_job(
+            self._run_refresh_search_cache,
+            CronTrigger(hour=1, minute=0),
+            id='scheduled_refresh_search_cache',
+            replace_existing=True,
+            max_instances=1,
+            coalesce=True
+        )
         self.scheduler.add_job(
             self._run_collect,
             CronTrigger(hour=1, minute=30),
@@ -62,14 +71,6 @@ class TaskScheduler:
             max_instances=1,
             coalesce=True
         )
-        self.scheduler.add_job(
-            self._run_refresh_search_cache,
-            CronTrigger(hour=2, minute=30),
-            id='scheduled_refresh_search_cache',
-            replace_existing=True,
-            max_instances=1,
-            coalesce=True
-        )
         self.scheduler.add_job(
             self._run_metrics_sync,
             CronTrigger(hour=6, minute=0),
@@ -80,7 +81,7 @@ class TaskScheduler:
         )
         self.scheduler.start()
         self._started = True
-        logger.info("Scheduler started: 01:30 collect, 02:30 refresh_search, 03:00 trends, 03:30 generate, 04:30 review, 05:00 optimize_sources, 06:00 metrics_sync")
+        logger.info("Scheduler started: 01:00 search_cache → 01:30 collect → 03:00 trends → 03:30 generate → 04:30 review → 05:00 optimize_sources → 06:00 metrics_sync")
     def shutdown(self):
         if self.scheduler.running:
             self.scheduler.shutdown()
@@ -285,6 +286,36 @@ class TaskScheduler:
             if count:
                 db.commit()
                 logger.info("[Scheduled] Metrics sync completed: synced %d zhihu articles", count)
+                # 生成指标反馈：按 field 聚合表现，写入 metrics_feedback.json 供 collector 读取
+                try:
+                    import json as json_mod
+                    from sqlalchemy import func as sql_func
+                    feedback = db.query(
+                        Topic.field,
+                        sql_func.avg(ContentMetrics.likes).label("avg_likes"),
+                        sql_func.avg(ContentMetrics.views).label("avg_views"),
+                        sql_func.avg(ContentMetrics.comments).label("avg_comments"),
+                        sql_func.count(ContentMetrics.id).label("article_count"),
+                    ).join(ContentMetrics, ContentMetrics.topic_id == Topic.id
+                    ).filter(Topic.field.isnot(None), Topic.field != ""
+                    ).group_by(Topic.field).all()
+                    if feedback:
+                        scored = []
+                        for row in feedback:
+                            score = (row.avg_likes or 0) + (row.avg_views or 0) * 0.01 + (row.avg_comments or 0) * 2
+                            scored.append((row.field, round(score, 1), int(row.article_count)))
+                        scored.sort(key=lambda x: x[1], reverse=True)
+                        feedback_data = {
+                            "updated_at": datetime.now().isoformat(),
+                            "top_domains": [(f, s) for f, s, _ in scored[:5]],
+                            "detail": [{"field": f, "score": s, "articles": c} for f, s, c in scored],
+                        }
+                        feedback_file = Path(__file__).parent.parent.parent.parent / "automation" / "data" / "metrics_feedback.json"
+                        feedback_file.parent.mkdir(parents=True, exist_ok=True)
+                        feedback_file.write_text(json_mod.dumps(feedback_data, ensure_ascii=False, indent=2), encoding='utf-8')
+                        logger.info("[Scheduled] Metrics feedback written: top domain %s (score %.1f)", scored[0][0], scored[0][1])
+                except Exception as e_fb:
+                    logger.warning("[Scheduled] Metrics feedback generation failed: %s", e_fb)
             else:
                 logger.info("[Scheduled] Metrics sync: no zhihu articles to sync")
             db.close()

scripts/collector.py

@@ -377,6 +377,61 @@ class SustainabilityCollector:
             logger.warning(f"web_search失败 {source.name}: {e}")
             return []
 
+    def _get_trend_context(self) -> str:
+        """读取热点趋势数据和指标反馈, 返回markdown上下文"""
+        parts = []
+        try:
+            from trends import load_trends
+            trends = load_trends()
+            if trends:
+                lines = ["## 当前热点趋势", ""]
+                for t in trends[:5]:
+                    src = {"weibo": "🔥", "zhihu": "📖", "baidu": "🔍", "llm": "🤖"}.get(t.get("source", ""), "")
+                    lines.append(f"- {src} **{t['topic']}**（{t.get('platform','')}）：{t.get('reason','')[:80]}")
+                    kw = t.get("hot_keywords", [])
+                    if kw:
+                        lines.append(f"  搜索热词：{' '.join(kw[:3])}")
+                parts.append("\n".join(lines))
+        except Exception:
+            pass
+
+        metrics_file = DATA_DIR / "metrics_feedback.json"
+        if metrics_file.exists():
+            try:
+                feedback = json.loads(metrics_file.read_text(encoding='utf-8'))
+                top_domains = feedback.get("top_domains", [])
+                if top_domains:
+                    lines = ["## 历史表现反馈（高互动领域优先", ""]
+                    for d, s in top_domains[:3]:
+                        lines.append(f"- {d}：平均分 {s}")
+                    parts.append("\n".join(lines))
+            except Exception:
+                pass
+
+        try:
+            from app.database import SessionLocal
+            from app.models import SystemConfig
+            db = SessionLocal()
+            try:
+                sc = db.query(SystemConfig).filter(SystemConfig.key == "collector_ai_advice").first()
+                if sc and sc.value:
+                    advice = json.loads(sc.value)
+                    summary = advice.get("summary", "")
+                    new_cats = advice.get("suggested_new_categories", [])
+                    if summary:
+                        parts.append(f"## AI策略建议\n{summary}")
+                    if new_cats:
+                        suggested = [f"- {c['name']}（{c.get('reason','')[:50]}）" for c in new_cats[:2]]
+                        parts.append("建议关注的新方向：\n" + "\n".join(suggested))
+            except Exception:
+                pass
+            finally:
+                db.close()
+        except Exception:
+            pass
+
+        return "\n\n".join(parts)
+
     def _generate_topics_with_llm(self, cases: List[SustainabilityCase] = None, search_results: List[Dict] = None) -> List[SustainabilityTopic]:
         """用LLM基于采集数据生成选题（数据充分时精确生成，无数据时凭知识生成）"""
         try:
@@ -402,11 +457,15 @@ class SustainabilityCollector:
             case_lines = [f"- {c.title[:40]}({c.category})" for c in cases[:5]]
             data_section += "\n采集案例：\n" + "\n".join(case_lines) + "\n"
 
+        trend_context = self._get_trend_context()
+
         prompt = f"""你是一个内容策略师。基于以下信息，为「{target_category}」类别生成一个高质量选题。
 
 {data_section if data_section else "（当前无实时采集数据，请基于你对中文互联网趋势的了解直接生成）"}
 {existing_hint}
 
+{trend_context}
+
 输出一个选题，格式JSON：
 {{{{
   "title": "标题（20字内，含核心关键词）",