fix: 优化虫 机制，新增伊朗支援

crawler/db_merge.py

@@ -26,6 +26,12 @@ MAX_DELTA_PER_MERGE = {
     "civilian_ships": 20, "airport_port": 10,
 }
 
+# 反击情绪 / 华尔街：合理区间，避免爬虫单条提取 0 或 100 导致指标归零或打满
+RETALIATION_SMOOTH_WEIGHT = 0.6  # 当前值权重，1 - 此值为新值权重，使更新平滑
+RETALIATION_HISTORY_MAX_ROWS = 300  # 反击历史条数上限，供前端曲线与回放使用
+WALL_STREET_TREND_MAX_ROWS = 200  # 趋势表保留最近条数，避免无限增长
+VALUE_CLAMP_MIN, VALUE_CLAMP_MAX = 1, 99  # 0/100 视为异常，写入前夹在 [1,99]
+
 
 def _clamp_delta(key: str, value: int) -> int:
     """单次增量上限，避免误提「累计」导致波动"""
@@ -200,38 +206,69 @@ def merge(extracted: Dict[str, Any], db_path: Optional[str] = None) -> bool:
                             updated = True
                     except Exception:
                         pass
-        # retaliation
+        # retaliation：平滑更新，避免单条新闻 0/100 导致指标归零或打满
         if "retaliation" in extracted:
             r = extracted["retaliation"]
-            conn.execute("INSERT OR REPLACE INTO retaliation_current (id, value) VALUES (1, ?)", (r["value"],))
-            conn.execute("INSERT INTO retaliation_history (time, value) VALUES (?, ?)", (r["time"], r["value"]))
+            raw = max(VALUE_CLAMP_MIN, min(VALUE_CLAMP_MAX, int(r.get("value", 50))))
+            row = conn.execute("SELECT value FROM retaliation_current WHERE id = 1").fetchone()
+            current = int(row[0]) if row else 50
+            current = max(VALUE_CLAMP_MIN, min(VALUE_CLAMP_MAX, current))
+            new_val = round(
+                RETALIATION_SMOOTH_WEIGHT * current + (1 - RETALIATION_SMOOTH_WEIGHT) * raw
+            )
+            new_val = max(VALUE_CLAMP_MIN, min(VALUE_CLAMP_MAX, new_val))
+            ts = (r.get("time") or datetime.utcnow().strftime("%Y-%m-%dT%H:%M:%S.000Z"))[:25]
+            conn.execute("INSERT OR REPLACE INTO retaliation_current (id, value) VALUES (1, ?)", (new_val,))
+            conn.execute("INSERT INTO retaliation_history (time, value) VALUES (?, ?)", (ts, new_val))
+            n_ret = conn.execute("SELECT COUNT(*) FROM retaliation_history").fetchone()[0]
+            if n_ret > RETALIATION_HISTORY_MAX_ROWS:
+                conn.execute(
+                    "DELETE FROM retaliation_history WHERE id IN (SELECT id FROM retaliation_history ORDER BY time ASC LIMIT ?)",
+                    (n_ret - RETALIATION_HISTORY_MAX_ROWS,),
+                )
             updated = True
-        # wall_street_trend
+        # wall_street_trend：限幅后写入，并保留最近 N 条避免表无限增长
         if "wall_street" in extracted:
             w = extracted["wall_street"]
-            conn.execute("INSERT INTO wall_street_trend (time, value) VALUES (?, ?)", (w["time"], w["value"]))
+            raw = int(w.get("value", 50))
+            val = max(VALUE_CLAMP_MIN, min(VALUE_CLAMP_MAX, raw))
+            ts = (w.get("time") or datetime.utcnow().strftime("%Y-%m-%dT%H:%M:%S.000Z"))[:25]
+            conn.execute("INSERT INTO wall_street_trend (time, value) VALUES (?, ?)", (ts, val))
+            n = conn.execute("SELECT COUNT(*) FROM wall_street_trend").fetchone()[0]
+            if n > WALL_STREET_TREND_MAX_ROWS:
+                conn.execute(
+                    "DELETE FROM wall_street_trend WHERE id IN (SELECT id FROM wall_street_trend ORDER BY time ASC LIMIT ?)",
+                    (n - WALL_STREET_TREND_MAX_ROWS,),
+                )
             updated = True
-        # key_location：更新双方攻击地点（美军基地被打击 side=us，伊朗设施被打击 side=iran）的 status/damage_level
+        # key_location：更新双方攻击地点（美军基地被打击 side=us，伊朗设施被打击 side=iran）的 status/damage_level/attacked_at
+        event_time = extracted.get("_event_time") or datetime.utcnow().strftime("%Y-%m-%dT%H:%M:%S.000Z")
         if "key_location_updates" in extracted:
             try:
                 for u in extracted["key_location_updates"]:
                     kw_raw = (u.get("name_keywords") or "").strip()
                     if not kw_raw:
                         continue
-                    # 支持 "a|b|c" 或 "a b c" 分隔
                     kw = [k.strip() for k in kw_raw.replace("|", " ").split() if k.strip()]
                     side = u.get("side")
                     status = (u.get("status") or "attacked")[:20]
                     dmg = u.get("damage_level", 2)
                     if not kw or side not in ("us", "iran"):
                         continue
-                    # 简化：name LIKE '%kw%' 对每个关键词 OR 连接，支持中英文
+                    attacked_at = (u.get("attacked_at") or event_time)[:25]
                     conditions = " OR ".join("name LIKE ?" for _ in kw)
-                    params = [status, dmg, side] + [f"%{k}%" for k in kw]
-                    cur = conn.execute(
-                        f"UPDATE key_location SET status=?, damage_level=? WHERE side=? AND ({conditions})",
-                        params,
-                    )
+                    params_with_at = [status, dmg, attacked_at, side] + [f"%{k}%" for k in kw]
+                    try:
+                        cur = conn.execute(
+                            f"UPDATE key_location SET status=?, damage_level=?, attacked_at=? WHERE side=? AND ({conditions})",
+                            params_with_at,
+                        )
+                    except sqlite3.OperationalError:
+                        params_no_at = [status, dmg, side] + [f"%{k}%" for k in kw]
+                        cur = conn.execute(
+                            f"UPDATE key_location SET status=?, damage_level=? WHERE side=? AND ({conditions})",
+                            params_no_at,
+                        )
                     if cur.rowcount > 0:
                         updated = True
             except Exception:

crawler/extractor_ai.py

@@ -51,6 +51,7 @@ def _call_ollama_extract(text: str, timeout: int = 15) -> Optional[Dict[str, Any
 - retaliation_sentiment: 0-100，仅当报道涉及伊朗报复/反击情绪时
 - wall_street_value: 0-100，仅当报道涉及美股/市场时
 - key_location_updates: **双方攻击地点**。每项 {{ "name_keywords": "阿萨德|asad|al-asad", "side": "us或iran（被打击方）", "status": "attacked", "damage_level": 1-3 }}。美军基地例：阿萨德|asad、乌代德|udeid、埃尔比勒|erbil、因吉尔利克|incirlik。伊朗例：德黑兰|tehran、布什尔|bushehr、伊斯法罕|isfahan、阿巴斯|abbas、纳坦兹|natanz
+- map_strike_lines（仅当报道为**美/以盟军打击伊朗目标**时）: 数组，每项 {{ "source_id": "israel或lincoln或ford", "target_lng": 经度, "target_lat": 纬度, "target_name": "目标名", "struck_at": "ISO时间" }}。目标坐标例：纳坦兹51.92,33.67；伊斯法罕51.67,32.65；德黑兰51.39,35.69；布什尔50.83,28.97；阿巴斯港56.27,27.18
 - **导弹消耗增量**（仅当报道明确提到「发射/消耗 了 X 枚导弹」时填，用于看板导弹消耗累计）: us_missile_consumed_delta, iran_missile_consumed_delta（本则报道中该方新增消耗枚数，整数）
 
 原文：
@@ -133,6 +134,31 @@ def extract_from_news(text: str, timestamp: Optional[str] = None) -> Dict[str, A
                 })
         if valid:
             out["key_location_updates"] = valid
+    # map_strike_lines：盟军打击伊朗目标
+    if "map_strike_lines" in parsed and isinstance(parsed["map_strike_lines"], list):
+        valid_lines = []
+        for line in parsed["map_strike_lines"]:
+            if not isinstance(line, dict):
+                continue
+            sid = str(line.get("source_id") or "").strip().lower()
+            if sid not in ("israel", "lincoln", "ford"):
+                continue
+            try:
+                lng = float(line.get("target_lng", 0))
+                lat = float(line.get("target_lat", 0))
+            except (TypeError, ValueError):
+                continue
+            name = str(line.get("target_name") or "")[:200]
+            struck_at = str(line.get("struck_at") or ts)[:25]
+            valid_lines.append({
+                "source_id": sid,
+                "target_lng": lng,
+                "target_lat": lat,
+                "target_name": name or None,
+                "struck_at": struck_at,
+            })
+        if valid_lines:
+            out["map_strike_lines"] = valid_lines
     # force_summary 增量：导弹消耗（看板「导弹消耗」由 force_summary.missile_consumed 提供）
     fs_delta = {}
     for side_key, side_val in [("us_missile_consumed_delta", "us"), ("iran_missile_consumed_delta", "iran")]:

crawler/extractor_dashscope.py

@@ -42,6 +42,7 @@ def _call_dashscope_extract(text: str, timeout: int = 15) -> Optional[Dict[str,
 - retaliation_sentiment: 0-100（仅当报道涉及伊朗报复情绪时）
 - wall_street_value: 0-100（仅当报道涉及美股/市场时）
 - key_location_updates: **双方攻击地点**。每项 {{"name_keywords":"阿萨德|asad","side":"us或iran（被打击方）","status":"attacked","damage_level":1-3}}。美军基地：阿萨德|asad、乌代德|udeid、埃尔比勒|erbil、因吉尔利克|incirlik。伊朗：德黑兰|tehran、布什尔|bushehr、伊斯法罕|isfahan、阿巴斯|abbas、纳坦兹|natanz
+- **map_strike_lines**（仅当报道明确为**美/以盟军打击伊朗或伊朗目标**时）: 数组，每项 {{"source_id":"israel或lincoln或ford","target_lng":经度,"target_lat":纬度,"target_name":"目标名如纳坦兹","struck_at":"ISO时间"}}。以色列打击→source_id=israel；林肯号→lincoln；福特号→ford。目标坐标：纳坦兹51.92,33.67；伊斯法罕51.67,32.65；德黑兰51.39,35.69；布什尔50.83,28.97；阿巴斯港56.27,27.18
 - **导弹消耗增量**（仅当报道明确提到「发射/消耗 了 X 枚导弹」时填）: us_missile_consumed_delta, iran_missile_consumed_delta（本则该方新增消耗枚数，整数）
 
 原文：
@@ -133,4 +134,29 @@ def extract_from_news(text: str, timestamp: Optional[str] = None) -> Dict[str, A
         if valid:
             out["key_location_updates"] = valid
 
+    if "map_strike_lines" in parsed and isinstance(parsed["map_strike_lines"], list):
+        valid_lines = []
+        for line in parsed["map_strike_lines"]:
+            if not isinstance(line, dict):
+                continue
+            sid = str(line.get("source_id") or "").strip().lower()
+            if sid not in ("israel", "lincoln", "ford"):
+                continue
+            try:
+                lng = float(line.get("target_lng", 0))
+                lat = float(line.get("target_lat", 0))
+            except (TypeError, ValueError):
+                continue
+            name = str(line.get("target_name") or "")[:200]
+            struck_at = str(line.get("struck_at") or ts)[:25]
+            valid_lines.append({
+                "source_id": sid,
+                "target_lng": lng,
+                "target_lat": lat,
+                "target_name": name or None,
+                "struck_at": struck_at,
+            })
+        if valid_lines:
+            out["map_strike_lines"] = valid_lines
+
     return out

crawler/extractor_rules.py

@@ -1,11 +1,24 @@
 # -*- coding: utf-8 -*-
 """
 基于规则的新闻数据提取（无需 Ollama）
-从新闻文本中提取战损、报复情绪等数值，供 db_merge 写入
+从新闻文本中提取战损、报复情绪、攻击地点与盟军打击线，供 db_merge 写入
 """
 import re
 from datetime import datetime, timezone
-from typing import Any, Dict, Optional
+from typing import Any, Dict, List, Optional, Tuple
+
+# 伊朗境内常见打击目标： (显示名, 经度, 纬度, 匹配关键词)
+IRAN_STRIKE_TARGETS: List[Tuple[str, float, float, str]] = [
+    ("纳坦兹", 51.916, 33.666, "natanz|纳坦兹"),
+    ("伊斯法罕", 51.67, 32.65, "isfahan|esfahan|伊斯法罕"),
+    ("德黑兰", 51.389, 35.689, "tehran|德黑兰"),
+    ("布什尔", 50.83, 28.97, "bushehr|布什尔"),
+    ("阿巴斯港", 56.27, 27.18, "bandar abbas|abbas|阿巴斯|霍尔木兹"),
+    ("克尔曼沙赫", 47.06, 34.31, "kermanshah|克尔曼沙赫"),
+    ("大不里士", 46.29, 38.08, "tabriz|大不里士"),
+    ("卡拉季", 50.99, 35.83, "karaj|卡拉季"),
+    ("米纳布", 57.08, 27.13, "minab|米纳布"),
+]
 
 
 def _first_int(text: str, pattern: str) -> Optional[int]:
@@ -251,4 +264,30 @@ def extract_from_news(text: str, timestamp: Optional[str] = None) -> Dict[str, A
         if updates:
             out["key_location_updates"] = updates
 
+    # map_strike_lines：盟军（以色列/林肯/福特）打击伊朗目标，供地图攻击动画更新
+    strike_verbs = ("strike" in t or "struck" in t or "strikes" in t or "hit" in t or "attack" in t
+                    or "打击" in (text or "") or "空袭" in (text or "") or "袭击" in (text or ""))
+    if strike_verbs and ("iran" in t or "伊朗" in (text or "") or any(
+        any(p in t for p in kw.split("|")) for _n, _lng, _lat, kw in IRAN_STRIKE_TARGETS
+    )):
+        source_id = "israel"
+        if "lincoln" in t or "林肯" in (text or ""):
+            source_id = "lincoln"
+        elif "ford" in t or "福特" in (text or ""):
+            source_id = "ford"
+        elif ("israel" in t or "idf" in t or "以色列" in (text or "")) and ("us " in t or "american" in t or "pentagon" in t):
+            source_id = "israel"  # 多国时优先以色列
+        lines = []
+        for name, lng, lat, kw in IRAN_STRIKE_TARGETS:
+            if any(p in t for p in kw.split("|")):
+                lines.append({
+                    "source_id": source_id,
+                    "target_lng": lng,
+                    "target_lat": lat,
+                    "target_name": name,
+                    "struck_at": ts,
+                })
+        if lines:
+            out["map_strike_lines"] = lines
+
     return out

crawler/pipeline.py

@@ -67,6 +67,8 @@ def _extract_and_merge(items: list, db_path: str) -> bool:
                 except Exception:
                     pass
             extracted = extract_from_news(text, timestamp=ts)
+            if ts:
+                extracted["_event_time"] = ts
             if extracted and merge(extracted, db_path=db_path):
                 merged_any = True
         return merged_any