feat(stock): news_issues (감성 기반 악재 flag)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

stock/app/holdings_intel.py

@@ -172,6 +172,43 @@ def market_events(ticker: str, ticker_prices: "pd.DataFrame",
     return events
 
 
+# ---- Task 3.2: news_issues ----
+
+NEG_SENTIMENT = -0.3   # 이하면 악재 후보
+
+
+def _news_sentiment_map(date: str) -> dict:
+    """date 기준 news_sentiment 테이블에서 ticker → {score_raw, news_count} 맵 반환."""
+    with db._conn() as conn:
+        try:
+            rows = conn.execute(
+                "SELECT ticker, score_raw, news_count FROM news_sentiment WHERE date=?",
+                (date,),
+            ).fetchall()
+        except Exception:
+            return {}
+    return {r["ticker"]: {"score_raw": r["score_raw"], "news_count": r["news_count"]}
+            for r in rows}
+
+
+def news_issues(tickers: list[str], date: str, use_llm: bool = True) -> dict[str, list]:
+    """news_sentiment 음수 → 악재 flag. (LLM 요약은 best-effort; 단위 테스트는 use_llm=False로.)"""
+    senti = _news_sentiment_map(date)
+    out: dict[str, list] = {}
+    for t in tickers:
+        s = senti.get(t)
+        if not s or s["score_raw"] is None:
+            continue
+        if s["score_raw"] <= NEG_SENTIMENT:
+            sev = "high" if s["score_raw"] <= NEG_SENTIMENT * 2 else "med"
+            out.setdefault(t, []).append({
+                "type": "news",
+                "severity": sev,
+                "summary": f"부정 뉴스 감성({s['score_raw']:+.2f}, {s.get('news_count', 0)}건)",
+            })
+    return out
+
+
 def decide_action(tech_score: float, exit_flags: dict, pnl: float | None,
                   add_score: float = ADD_SCORE) -> tuple[str, str]:
     """액션 결정 매트릭스: sell > trim > add > hold (우선순위 순).