windows-desktop: C端基线（Windows 本地实时分析器 + 采集器 bug 修复）

src/__init__.py

@@ -0,0 +1 @@
+# a_stock_timeline

src/analysis/mine_patterns.py

@@ -0,0 +1,211 @@
+# -*- coding: utf-8 -*-
+"""
+A股时间线实证模式挖掘：从 a_stock.db 挖掘隔夜传导/跳空回补/周内效应/
+波动率机制/龙虎榜统计，输出 markdown 到 E:/Data/skills/a-stock-timeline-patterns/references/
+口径：上证/沪深300 主样本；隔夜传导以纳斯达克为主参照。
+"""
+import sqlite3
+import warnings
+
+import numpy as np
+import pandas as pd
+
+warnings.filterwarnings('ignore')
+
+DB = r'C:' + chr(92) + 'Users' + chr(92) + 'lookt' + chr(92) + 'a_stock_timeline' + chr(92) + 'data' + chr(92) + 'a_stock.db'
+OUT = r'E:' + chr(92) + 'Data' + chr(92) + 'skills' + chr(92) + 'a-stock-timeline-patterns' + chr(92) + 'references'
+
+IDX_NAMES = {'sh000001': '上证指数', 'sz399001': '深证成指', 'sh000300': '沪深300',
+             'sz399006': '创业板指', 'sh000688': '科创50'}
+
+
+def load():
+    conn = sqlite3.connect(DB)
+    daily = pd.read_sql("SELECT ts_code, trade_date, open, high, low, close, volume "
+                        "FROM stock_daily ORDER BY ts_code, trade_date", conn)
+    daily['trade_date'] = pd.to_datetime(daily['trade_date'])
+    for c in ('open', 'high', 'low', 'close'):
+        daily[c] = pd.to_numeric(daily[c], errors='coerce')
+    daily = daily.dropna(subset=['open', 'close'])
+    daily = daily[daily['close'] > 0]
+    g = pd.read_sql("SELECT index_code, trade_date, close, pct_change FROM global_index", conn)
+    g['trade_date'] = pd.to_datetime(g['trade_date'])
+    lhb = pd.read_sql("SELECT * FROM lhb_daily", conn)
+    conn.close()
+    return daily, g, lhb
+
+
+def enrich(daily):
+    """按指数补充：日收益/隔夜跳空/日内波动/已实现波动率"""
+    out = []
+    for code, df in daily.groupby('ts_code'):
+        df = df.sort_values('trade_date').copy()
+        df['prev_close'] = df['close'].shift(1)
+        df['ret'] = df['close'] / df['prev_close'] - 1
+        df['gap'] = df['open'] / df['prev_close'] - 1
+        df['intra'] = df['close'] / df['open'] - 1
+        df['ret20'] = df['ret'].rolling(20).std()
+        df['ret20_prev'] = df['ret20'].shift(20)
+        df['weekday'] = df['trade_date'].dt.dayofweek
+        out.append(df)
+    return pd.concat(out, ignore_index=True)
+
+
+def fmt_pct(v, digits=2):
+    return "{:.{}f}%".format(v * 100, digits)
+
+
+def stat_line(sub, col=None):
+    s = sub.dropna() if hasattr(sub, 'dropna') else sub
+    if len(s) == 0:
+        return "样本不足"
+    up = len(s[s > 0]) / len(s) * 100
+    return "N={}, 均值 {}, 中位数 {}, 胜率 {:.1f}%".format(
+        len(s), fmt_pct(s.mean()), fmt_pct(s.median()), up)
+
+
+def overnight_matrix(daily, g):
+    lines = ["## 隔夜传导矩阵（美股 -> A股）", "",
+             "> 口径：美股T日涨跌幅分箱 -> A股T+1交易日（跳空=开盘/前收-1；日内=收盘/开盘-1）。"
+             "样本 2005-2026（受 A 股指数历史与纳斯达克 2014 起数据限制）。", ""]
+    for us_code, us_name in [('US.NDX', '纳斯达克100'), ('US.SPX', '标普500'), ('US.DJI', '道琼斯')]:
+        gu = g[g['index_code'] == us_code][['trade_date', 'pct_change']].rename(
+            columns={'pct_change': 'us_ret'})
+        if gu.empty:
+            continue
+        gu['us_ret'] = pd.to_numeric(gu['us_ret'], errors='coerce') / 100.0
+        bins = [-np.inf, -0.015, -0.005, 0.005, 0.015, np.inf]
+        labels = ['跌>1.5%', '跌0.5~1.5%', '正负0.5%内', '涨0.5~1.5%', '涨>1.5%']
+        gu['us_bin'] = pd.cut(gu['us_ret'], bins=bins, labels=labels)
+        for idx_code in ['sh000001', 'sh000300']:
+            a = daily[daily['ts_code'] == idx_code][['trade_date', 'gap', 'intra', 'ret']].copy()
+            a = a.sort_values('trade_date')
+            merged = pd.merge_asof(a, gu.sort_values('trade_date'),
+                                   left_on='trade_date', right_on='trade_date',
+                                   direction='backward', allow_exact_matches=False)
+            merged = merged.dropna(subset=['us_bin', 'gap'])
+            name = IDX_NAMES.get(idx_code, idx_code)
+            lines.append("### {}T日 -> {}T+1日".format(us_name, name))
+            lines.append("")
+            for lab in labels:
+                sub = merged[merged['us_bin'] == lab]
+                if len(sub) < 30:
+                    lines.append("- 美股{}: 样本不足（{}）".format(lab, len(sub)))
+                    continue
+                lines.append("- 美股{}（{}天）: A股跳空 {}; 日内 {}; 全天 {}".format(
+                    lab, len(sub), stat_line(sub['gap']), stat_line(sub['intra']),
+                    stat_line(sub['ret'])))
+            lines.append("")
+    return chr(10).join(lines)
+
+
+def gap_intraday(daily):
+    lines = ["## A股指数跳空回补统计（日外->日内的接力规律）", "",
+             "> 口径：跳空=开盘/前收-1；日内=收盘/开盘-1。日内为负即\"高开低走\"（回落/回补）。", ""]
+    bins = [(-np.inf, -0.01, '大幅低开<-1%'), (-0.01, -0.002, '低开0.2~1%'),
+            (-0.002, 0.002, '平开正负0.2%'), (0.002, 0.01, '高开0.2~1%'),
+            (0.01, np.inf, '大幅高开>1%')]
+    for idx_code in ['sh000001', 'sh000300']:
+        a = daily[daily['ts_code'] == idx_code].dropna(subset=['gap', 'intra'])
+        name = IDX_NAMES.get(idx_code, idx_code)
+        years = "{}-{}".format(a['trade_date'].dt.year.min(), a['trade_date'].dt.year.max())
+        lines.append("### {}（{}，{}天）".format(name, years, len(a)))
+        for lo, hi, lab in bins:
+            sub = a[(a['gap'] > lo) & (a['gap'] <= hi)]
+            if len(sub) < 30:
+                continue
+            intra_up = (sub['intra'] > 0).mean() * 100
+            lines.append("- {}（{}天，占比 {:.1f}%）: 日内均值 {}，日内收涨概率 {:.1f}%".format(
+                lab, len(sub), len(sub) / len(a) * 100, fmt_pct(sub['intra'].mean()), intra_up))
+        lines.append("")
+    return chr(10).join(lines)
+
+
+def weekday_effect(daily):
+    lines = ["## A股指数周内效应", "",
+             "> 口径：指数日收益按星期聚合（2003/2004-2026）。", ""]
+    wd_names = ['周一', '周二', '周三', '周四', '周五']
+    for idx_code in ['sh000001', 'sh000300']:
+        a = daily[daily['ts_code'] == idx_code].dropna(subset=['ret'])
+        name = IDX_NAMES.get(idx_code, idx_code)
+        lines.append("### {}（{}天）".format(name, len(a)))
+        for wd in range(5):
+            sub = a[a['weekday'] == wd]
+            lines.append("- {}: {}".format(wd_names[wd], stat_line(sub['ret'])))
+        lines.append("")
+    return chr(10).join(lines)
+
+
+def volatility_regime(daily):
+    lines = ["## 波动率机制切换（20日波动率翻倍/减半 -> 后5日）", "",
+             "> 口径：20日收益标准差 vs 前20日对比；统计切换后 5 日收益分布。", ""]
+    for idx_code in ['sh000001', 'sh000300']:
+        a = daily[daily['ts_code'] == idx_code].copy()
+        a = a.dropna(subset=['ret20', 'ret20_prev'])
+        name = IDX_NAMES.get(idx_code, idx_code)
+        lines.append("### {}".format(name))
+        for lab, cond in [('骤增(>2倍)', a['ret20'] > a['ret20_prev'] * 2),
+                          ('骤降(<0.5倍)', a['ret20'] < a['ret20_prev'] * 0.5)]:
+            sub_idx = a[cond].index
+            if len(sub_idx) < 30:
+                lines.append("- {}: 样本不足（{}）".format(lab, len(sub_idx)))
+                continue
+            fwd = []
+            for idx in sub_idx:
+                pos = a.index.get_loc(idx)
+                window = a.iloc[pos + 1:pos + 6]
+                fwd.append((1 + window['ret']).prod() - 1)
+            fwd = pd.Series(fwd).dropna()
+            lines.append("- {}（{}次）: 后5日 {}".format(lab, len(sub_idx), stat_line(fwd)))
+        lines.append("")
+    return chr(10).join(lines)
+
+
+def lhb_stats(lhb, daily):
+    lines = ["## 龙虎榜统计（近90个交易日）", "",
+             "> 口径：仅覆盖近90日采集窗口；上榜股只有当日快照，暂无次日数据（不提供次日胜率）。", ""]
+    total_up = (lhb['pct_change'] > 0).sum()
+    lines.append("- 上榜记录 {} 条；上榜当日上涨占比 {:.1f}%".format(len(lhb), total_up / len(lhb) * 100))
+    if 'reason' in lhb.columns:
+        lines.append("- 下跌偏离上榜占比 {:.1f}%（跌幅榜多=弱势环境信号）".format(
+            lhb['reason'].str.contains('跌幅').mean() * 100))
+    lines.append("")
+    lines.append("### 上榜原因分布 TOP10")
+    for reason, cnt in lhb['reason'].value_counts().head(10).items():
+        lines.append("- {}次 | {}".format(cnt, reason))
+    lines.append("")
+    daily_sh = daily[daily['ts_code'] == 'sh000001'][['trade_date', 'ret']]
+    merged = lhb.copy()
+    merged['trade_date'] = pd.to_datetime(merged['trade_date'])
+    m = merged.merge(daily_sh, on='trade_date', how='left')
+    m['mkt_down'] = m['ret'] < 0
+    lines.append("### 大盘状态 × 龙虎榜家数")
+    for down, cnt in m.groupby('mkt_down').size().items():
+        label = '指数下跌日' if down else '指数上涨日'
+        sub = m[m['mkt_down'] == down]
+        lines.append("- {}: 平均上榜 {:.1f} 条（{}条/{}天）".format(label, cnt / len(sub), cnt, len(sub)))
+    return chr(10).join(lines)
+
+
+def main():
+    import os
+    daily, g, lhb = load()
+    daily = enrich(daily)
+    refs = {
+        'overnight_transmission.md': overnight_matrix(daily, g),
+        'gap_intraday.md': gap_intraday(daily),
+        'weekday_volatility.md': weekday_effect(daily) + NL2 + volatility_regime(daily),
+        'lhb_stats.md': lhb_stats(lhb, daily),
+    }
+    os.makedirs(OUT, exist_ok=True)
+    for name, content in refs.items():
+        out_path = os.path.normpath(os.path.join(OUT, name))
+        with open(out_path, 'w', encoding='utf-8') as f:
+            f.write(content)
+        print("written:", name, len(content), "chars")
+
+
+NL2 = chr(10) * 2
+
+if __name__ == '__main__':
+    main()

src/fetcher/__init__.py

@@ -0,0 +1 @@
+# fetcher package

src/fetcher/calendar_fetcher.py

@@ -0,0 +1,27 @@
+"""交易日历采集 - AkShare tool_trade_date_hist_sina"""
+import warnings, logging, sqlite3
+from datetime import datetime
+import pandas as pd
+import akshare as ak
+
+logging.basicConfig(level=logging.INFO, format='%(asctime)s %(levelname)s %(message)s')
+log = logging.getLogger(__name__)
+
+DB = r'C:\Users\lookt\a_stock_timeline\data\a_stock.db'
+
+def fetch_calendar():
+    df = ak.tool_trade_date_hist_sina()
+    log.info(f'获取到 {len(df)} 行，列: {df.columns.tolist()}')
+    # 原始只有 trade_date 一列；按 A 股规则补 is_trade_day
+    # 周六=0, 周日=0, 其他=1
+    df['trade_date'] = pd.to_datetime(df['trade_date'])
+    df['is_trade_day'] = df['trade_date'].dt.dayofweek.apply(lambda x: 0 if x >= 5 else 1)
+    df['trade_date'] = df['trade_date'].dt.strftime('%Y-%m-%d')
+    conn = sqlite3.connect(DB)
+    df.to_sql('trade_calendar', conn, if_exists='replace', index=False)
+    conn.close()
+    log.info(f'写入 trade_calendar: {len(df)} 行')
+    return df
+
+if __name__ == '__main__':
+    fetch_calendar()

src/fetcher/cn_news_fetcher.py

@@ -0,0 +1,42 @@
+"""A股新闻采集 - AkShare stock_news_em"""
+import warnings, logging, sqlite3
+from datetime import datetime
+import pandas as pd
+import akshare as ak
+
+logging.basicConfig(level=logging.INFO, format='%(asctime)s %(levelname)s %(message)s')
+log = logging.getLogger(__name__)
+
+DB = r'C:\Users\lookt\a_stock_timeline\data\a_stock.db'
+
+def fetch_news():
+    df = ak.stock_news_em(symbol='A股')
+    log.info(f'获取到 {len(df)} 行，列: {df.columns.tolist()}')
+    # 列名标准化
+    cn_map = {
+        '关键词': 'keywords',
+        '股票代码': 'ts_code',
+        '新闻标题': 'title',
+        '新闻内容': 'content',
+        '发布时间': 'pub_date',
+        '文章来源': 'source',
+        '新闻链接': 'url',
+    }
+    df = df.rename(columns=cn_map)
+    # 只保留表里有的列
+    db_cols = ['title', 'content', 'pub_date', 'source', 'url', 'keywords']
+    for c in db_cols:
+        if c not in df.columns:
+            df[c] = ''
+    df = df[['id'] + [c for c in db_cols if c in df.columns]] if 'id' in df.columns else df[db_cols]
+    df['inserted_at'] = datetime.now().isoformat()
+    if 'id' not in df.columns:
+        df.insert(0, 'id', range(1, len(df) + 1))
+    conn = sqlite3.connect(DB)
+    df.to_sql('news_cn', conn, if_exists='replace', index=False)
+    conn.close()
+    log.info(f'写入 news_cn: {len(df)} 行')
+    return df
+
+if __name__ == '__main__':
+    fetch_news()

src/fetcher/fund_flow_fetcher.py

@@ -0,0 +1,151 @@
+"""资金流向采集 - 修正版：同花顺个股资金流（逐页校验，跳过异常页），历史保留"""
+import time
+import warnings
+import logging
+from urllib.parse import urlparse
+
+import pandas as pd
+import requests
+
+from src.storage.db import upsert_rows
+
+warnings.filterwarnings('ignore')
+logging.basicConfig(level=logging.INFO, format='%(asctime)s %(levelname)s %(message)s')
+log = logging.getLogger(__name__)
+
+DB = r'C:\Users\lookt\a_stock_timeline\data\a_stock.db'
+
+# 出站请求域名白名单（SSRF 防护：仅 http + 白名单主机，禁重定向跟随）
+_ALLOWED_HOSTS = {'data.10jqka.com.cn'}
+_ALLOWED_PREFIX = 'http://data.10jqka.com.cn/funds/ggzjl/' 
+
+
+def _safe_get(url: str, headers: dict, timeout: int = 15) -> requests.Response:
+    """白名单校验后的 GET：协议/host 白名单 + 禁重定向，防 SSRF"""
+    u = urlparse(url)
+    if u.scheme != 'http' or u.hostname not in _ALLOWED_HOSTS:
+        raise ValueError(f'blocked non-allowlist url: {url}')
+    return requests.get(url, timeout=timeout, allow_redirects=False, headers=headers)
+
+
+def parse_cn_amount(v):
+    """'1.5亿' -> 1.5e8, '-300万' -> -3e6，逗号/空白容错；无效返回 None"""
+    if v is None or (isinstance(v, float) and v != v):
+        return None
+    s = str(v).replace(',', '').replace(' ', '').strip()
+    if not s or s in ('--', '-', 'nan'):
+        return None
+    mult = 1.0
+    if s.endswith('亿'):
+        mult, s = 1e8, s[:-1]
+    elif s.endswith('万'):
+        mult, s = 1e4, s[:-1]
+    try:
+        return float(s) * mult
+    except ValueError:
+        return None
+
+
+def _ths_headers():
+    from akshare.stock_feature.stock_fund_flow import _get_file_content_ths
+    from py_mini_racer import MiniRacer
+    js_code = MiniRacer()
+    js_code.eval(_get_file_content_ths("ths.js"))
+    v_code = js_code.call("v")
+    return {
+        "Accept": "text/html, */*; q=0.01",
+        "hexin-v": v_code,
+        "Host": "data.10jqka.com.cn",
+        "Pragma": "no-cache",
+        "Referer": "http://data.10jqka.com.cn/funds/hyzjl/",
+        "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 "
+                      "(KHTML, like Gecko) Chrome/90.0.4430.85 Safari/537.36",
+        "X-Requested-With": "XMLHttpRequest",
+    }
+
+
+def _page_url(page: int) -> str:
+    """拼分页 URL；page 强制 int，且必须落在白名单前缀内"""
+    page = int(page)
+    url = _ALLOWED_PREFIX + f"field/zdf/order/desc/page/{page}/ajax/1/free/1/"
+    if not url.startswith(_ALLOWED_PREFIX):
+        raise ValueError('url escapes allowlist prefix')
+    return url
+
+
+def fetch_individual_flow(max_pages=None, retries_per_page=2):
+    """
+    同花顺个股资金流排行（修正版）：
+    akshare 原实现对部分页的异常 13 列表格直接崩溃，这里逐页校验：
+    只接受恰好 10 列且表头匹配的表格，异常页跳过并记录。
+    """
+    from bs4 import BeautifulSoup
+    from io import StringIO
+
+    headers = _ths_headers()  # 每次请求刷新 hexin-v
+    expect = ['序号', '股票代码', '股票简称', '最新价', '涨跌幅', '换手率',
+              '流入资金(元)', '流出资金(元)', '净额(元)', '成交额(元)']
+
+    r = _safe_get(_page_url(1), headers)
+    soup = BeautifulSoup(r.text, features="lxml")
+    page_info = soup.find(name="span", attrs={"class": "page_info"})
+    page_num = int(page_info.text.split("/")[1]) if page_info else 1
+    if max_pages:
+        page_num = min(page_num, max_pages)
+
+    frames = []
+    skipped = []
+    for page in range(1, page_num + 1):
+        for attempt in range(retries_per_page):
+            try:
+                r = _safe_get(_page_url(page), headers)
+                dfs = pd.read_html(StringIO(r.text))
+                df = next((d for d in dfs if len(d.columns) == 10
+                           and list(d.columns)[:2] == ['序号', '股票代码']), None)
+                if df is None:
+                    skipped.append(page)
+                    break
+                df.columns = expect
+                frames.append(df)
+                break
+            except Exception:
+                if attempt == retries_per_page - 1:
+                    skipped.append(page)
+                else:
+                    time.sleep(1)
+        time.sleep(0.3)   # 限速，避免同花顺封禁
+    if skipped:
+        log.warning(f'跳过异常页: {skipped}')
+    if not frames:
+        return pd.DataFrame()
+    big = pd.concat(frames, ignore_index=True)
+    big.columns = [c.split('(')[0] for c in big.columns]   # 去掉 (元) 后缀
+    return big
+
+
+def fetch_money_flow():
+    """采集全市场个股当日主力净流入（净额），并入库（历史保留）"""
+    today = time.strftime('%Y-%m-%d')
+    raw = fetch_individual_flow()
+    if raw.empty:
+        log.warning('个股资金流采集为空')
+        return None
+
+    df = pd.DataFrame({
+        'ts_code': raw['股票代码'].astype(str).str.zfill(6),
+        'trade_date': today,
+        'main_net_in': raw['净额'].map(parse_cn_amount),
+        'in_flow': raw['流入资金'].map(parse_cn_amount),
+        'out_flow': raw['流出资金'].map(parse_cn_amount),
+        'amount': raw['成交额'].map(parse_cn_amount),
+        'pct_change': raw['涨跌幅'].str.rstrip('%').map(
+            lambda x: float(x) if x and x not in ('--', '') else None),
+    })
+    df = df[df['main_net_in'].notna()]
+    n = upsert_rows(df, 'money_flow', conflict_cols=['ts_code', 'trade_date'])
+    log.info(f'写入 money_flow: {n} 行（历史保留）')
+    return df
+
+
+if __name__ == '__main__':
+    fetch_money_flow()

src/fetcher/global_index_fetcher.py

@@ -0,0 +1,54 @@
+"""外盘指数采集 - AkShare index_us_stock_sina"""
+import warnings, logging, sqlite3
+from datetime import datetime
+import pandas as pd
+import akshare as ak
+
+logging.basicConfig(level=logging.INFO, format='%(asctime)s %(levelname)s %(message)s')
+log = logging.getLogger(__name__)
+
+DB = r'C:\Users\lookt\a_stock_timeline\data\a_stock.db'
+
+# 要采集的外盘指数
+INDICES = [
+    ('US.DJI', '.DJI', '道琼斯工业'),
+    ('US.NDX', '.NDX', '纳斯达克100'),
+    ('US.SPX', '.INX', '标普500'),
+]
+
+def fetch_global_index():
+    conn = sqlite3.connect(DB)
+    results = []
+    for code, symbol, name in INDICES:
+        try:
+            df = ak.index_us_stock_sina(symbol=symbol)
+            log.info(f'{name}({symbol}): {len(df)} 行，列: {df.columns.tolist()}')
+            # 标准化
+            cn_map = {
+                '时间': 'trade_date',
+                '日期': 'trade_date',
+                'date': 'trade_date',
+                '收盘': 'close',
+                'close': 'close',
+                '涨跌幅': 'pct_change',
+                'pct_change': 'pct_change',
+            }
+            df = df.rename(columns=cn_map)
+            if 'trade_date' not in df.columns:
+                df['trade_date'] = df['date'] if 'date' in df.columns else None
+            df['index_code'] = code
+            df['trade_date'] = pd.to_datetime(df['trade_date']).dt.strftime('%Y-%m-%d')
+            # 计算涨跌幅（基于前一日收盘价）
+            df = df.sort_values('trade_date')
+            df['pct_change'] = df['close'].pct_change() * 100
+            out = df[['index_code', 'trade_date', 'close', 'pct_change']].dropna()
+            out.to_sql('global_index', conn, if_exists='append', index=False)
+            results.append((code, len(out)))
+            log.info(f'  -> 写入 {code}: {len(out)} 行')
+        except Exception as e:
+            log.warning(f'{name}({symbol}) 失败: {e}')
+    conn.close()
+    return results
+
+if __name__ == '__main__':
+    fetch_global_index()

src/fetcher/lhb_fetcher.py

@@ -0,0 +1,83 @@
+"""
+龙虎榜采集器
+主方案：AkShare stock_lhb_detail_daily_sina（已实测可用，含上榜原因）
+备选：东财 datacenter 接口（报表名不确定，暂时注释备用）
+"""
+
+import sys
+from pathlib import Path
+sys.path.insert(0, str(Path(__file__).parent.parent.parent))
+
+import akshare as ak
+import pandas as pd
+from datetime import datetime, timedelta
+from src.storage.db import upsert_df, upsert_log, get_last_fetch
+
+
+def fetch_lhb_date(trade_date: str) -> pd.DataFrame:
+    """
+    获取单日龙虎榜
+    trade_date: "YYYY-MM-DD" 或 "YYYYMMDD"
+    """
+    # 统一为 YYYYMMDD 格式
+    ds = trade_date.replace("-", "")
+    try:
+        df = ak.stock_lhb_detail_daily_sina(date=ds)
+    except Exception as e:
+        print(f"  [WARN] LHB Sina failed for {trade_date}: {e}")
+        return pd.DataFrame()
+
+    if df is None or df.empty:
+        return pd.DataFrame()
+
+    # Sina 返回列（实测）：序号, 股票代码, 股票名称, 收盘价, 涨跌幅, 成交额, 成交额, 上榜原因
+    # 编码问题导致中文列名显示乱码，用位置索引映射
+    # 0=序号, 1=代码, 2=名称, 3=收盘, 4=涨跌幅, 5=成交额, 6=??? 7=上榜原因
+    df.columns = ["no", "ts_code", "name", "close", "pct_change", "amount", "amount2", "reason"]
+
+    df["trade_date"] = trade_date.replace("-", "")
+    # 格式化为 YYYY-MM-DD
+    df["trade_date"] = pd.to_datetime(df["trade_date"]).dt.strftime("%Y-%m-%d")
+
+    # 保留有用字段
+    keep = ["trade_date", "ts_code", "name", "close", "pct_change", "amount", "reason"]
+    df = df[[c for c in keep if c in df.columns]]
+    return df
+
+
+def fetch_lhb_range(start_date: str, end_date: str = None) -> pd.DataFrame:
+    """抓取日期区间的龙虎榜（逐日）"""
+    end_date = end_date or datetime.today().strftime("%Y-%m-%d")
+    all_days = []
+    cur  = datetime.strptime(start_date.replace("-", ""), "%Y%m%d")
+    end  = datetime.strptime(end_date.replace("-", ""),   "%Y%m%d")
+    while cur <= end:
+        ds = cur.strftime("%Y-%m-%d")
+        df = fetch_lhb_date(ds)
+        if df is not None and not df.empty:
+            all_days.append(df)
+        cur += timedelta(days=1)
+    if all_days:
+        return pd.concat(all_days, ignore_index=True)
+    return pd.DataFrame()
+
+
+def incremental_lhb(source: str = "lhb_daily") -> pd.DataFrame:
+    """增量采集：只抓上次之后的新交易日"""
+    last = get_last_fetch(source)
+    start = last.get("last_date") or (datetime.today() - timedelta(days=7)).strftime("%Y-%m-%d")
+    end   = datetime.today().strftime("%Y-%m-%d")
+    df = fetch_lhb_range(start, end)
+    if df is not None and not df.empty:
+        upsert_df(df, "lhb_daily")
+        upsert_log(source, param="", last_date=end)
+    return df
+
+
+if __name__ == "__main__":
+    import fire
+    fire.Fire({
+        "date":  lambda d: upsert_df(fetch_lhb_date(d), "lhb_daily"),
+        "range": lambda s, e=None: upsert_df(fetch_lhb_range(s, e), "lhb_daily"),
+        "incr":  lambda: incremental_lhb(),
+    })

src/fetcher/macro_event_fetcher.py

@@ -0,0 +1,133 @@
+"""
+每日大事提醒 / 财经日历采集器
+主方案：AkShare news_economic_baidu（财经日历事件，日期/时间/重要性完整）
+备选：东财 datacenter 大事提醒接口
+"""
+
+import sys
+from pathlib import Path
+sys.path.insert(0, str(Path(__file__).parent.parent.parent))
+
+import requests
+import pandas as pd
+import akshare as ak
+from datetime import datetime, timedelta
+from src.storage.db import upsert_df, upsert_rows, upsert_log, get_last_fetch
+
+EM_BASE = "https://datacenter-web.eastmoney.com/api/data/v1/get"
+
+
+def em_headers() -> dict:
+    return {
+        "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36",
+        "Referer": "https://data.eastmoney.com/",
+        "Accept": "application/json",
+    }
+
+
+# ── 主方案：AkShare 财经日历 ─────────────────────────────────────────────────
+
+def fetch_macro_calendar() -> pd.DataFrame:
+    """
+    AkShare news_economic_baidu 返回财经日历事件
+    列: 标题, 时间, 重要性, 前值, 预测值, 公布值, 影响
+    """
+    try:
+        df = ak.news_economic_baidu()
+    except Exception as e:
+        print(f"  [WARN] news_economic_baidu failed: {e}")
+        return pd.DataFrame()
+
+    if df is None or df.empty:
+        return pd.DataFrame()
+
+    # 实测列（akshare 1.18.92）: 日期, 时间, 地区, 事件, 公布, 预期, 前值, 重要性
+    # 按列【名称】映射，避免位置错位（严禁再按位置覆盖 df.columns）
+    col_map = {'日期': 'event_date', '时间': 'event_time', '地区': 'source',
+               '事件': 'title', '公布': 'actual', '预期': 'forecast',
+               '前值': 'prev', '重要性': 'event_type'}
+    df = df.rename(columns=col_map)
+
+    # 防御：akshare 偶发返回缺列，补默认值
+    for c, default in [('event_time', ''), ('actual', None), ('forecast', None),
+                       ('prev', None), ('event_type', 0), ('source', '')]:
+        if c not in df.columns:
+            df[c] = default
+    df["title"] = df["title"].fillna("").astype(str)
+    df["event_date"] = pd.to_datetime(df["event_date"], errors="coerce").dt.strftime("%Y-%m-%d")
+    df["event_time"] = df["event_time"].fillna("").astype(str)
+    df["event_type"] = pd.to_numeric(df["event_type"], errors="coerce").fillna(0).astype(int)
+    df["source"] = df["source"].fillna("").astype(str)
+
+    keep = ["event_date", "event_time", "title", "event_type", "source",
+            "actual", "forecast", "prev"]
+    df = df[[c for c in keep if c in df.columns]]
+    return df
+
+
+# ── 备选：东财大事提醒接口 ───────────────────────────────────────────────────
+
+def fetch_em_events(
+    start_date: str = None,
+    end_date: str = None,
+    page: int = 1,
+    page_size: int = 50
+) -> pd.DataFrame:
+    """
+    东财大事提醒接口（需确认报表名）
+    已知可用报表: RPT_MAJOR_NOTICE（公告）、RPT_LHB_ALL_STOCKS（龙虎榜）
+    大事提醒报表名待抓包确认，暂时返回空
+    """
+    if start_date is None:
+        start_date = (datetime.today() - timedelta(days=3)).strftime("%Y-%m-%d")
+    if end_date is None:
+        end_date = (datetime.today() + timedelta(days=60)).strftime("%Y-%m-%d")
+
+    # 尝试 RPT_MAJOR_NOTICES（公告大事）
+    report_names = ["RPT_MAJOR_NOTICES", "RPT_IMPORTANT_NEWS"]
+    for report in report_names:
+        params = {
+            "reportName": report,
+            "columns": "ALL",
+            "filter": f'(EVENT_DATE>=\'{start_date}\') and (EVENT_DATE<=\'{end_date}\')',
+            "pageNumber": page,
+            "pageSize": page_size,
+            "sortTypes": "1",
+            "sortColumns": "EVENT_DATE",
+            "source": "WEB",
+            "client": "WEB",
+        }
+        try:
+            r = requests.get(EM_BASE, params=params, headers=em_headers(), timeout=15)
+            r.raise_for_status()
+            resp = r.json()
+            result = resp.get("result", {}) or {}
+            data = result.get("data", []) or []
+            if data:
+                df = pd.DataFrame(data)
+                df.columns = [c.lower() for c in df.columns]
+                return df
+        except Exception:
+            continue
+    return pd.DataFrame()
+
+
+def incremental_events(source: str = "macro_event") -> pd.DataFrame:
+    """增量采集宏观事件（按 event_date+event_time+title 去重，历史保留）"""
+    df = fetch_macro_calendar()
+    if df is not None and not df.empty:
+        need = [c for c in ['event_date', 'event_time', 'title', 'event_type', 'source',
+                            'actual', 'forecast', 'prev', 'inserted_at'] if c in df.columns]
+        upsert_rows(df[need], source,
+                    conflict_cols=['event_date', 'event_time', 'title'])
+        upsert_log(source, "", last_date=datetime.today().strftime("%Y-%m-%d"))
+    return df
+
+
+if __name__ == "__main__":
+    import fire
+    fire.Fire({
+        "ak":   lambda: upsert_df(fetch_macro_calendar(), "macro_event"),
+        "em":   lambda s=None, e=None: upsert_df(fetch_em_events(s, e), "macro_event"),
+        "incr": lambda: incremental_events(),
+    })

src/fetcher/news_flash_fetcher.py

@@ -0,0 +1,141 @@
+"""
+7x24 快讯流采集器
+主方案：东财 np-anotice-stock 快讯接口（秒级时间戳）
+备选方案：AkShare news_economic_baidu（财经日历事件）
+"""
+
+import sys
+from pathlib import Path
+sys.path.insert(0, str(Path(__file__).parent.parent.parent))
+
+import requests
+import pandas as pd
+import akshare as ak
+from datetime import datetime
+from src.storage.db import upsert_df, upsert_log, get_last_fetch
+
+
+# ── 主方案：东财快讯接口 ────────────────────────────────────────────────────
+
+EM_FLASH_API = "https://np-anotice-stock.eastmoney.com/api/security/ann"
+
+def fetch_em_flash(category: str = "全部", page: int = 1, page_size: int = 50) -> dict | None:
+    headers = {
+        "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64)",
+        "Referer": "https://www.eastmoney.com/",
+        "Accept": "application/json",
+    }
+    params = {
+        "sr": -1,
+        "page": page,
+        "pageSize": page_size,
+        "category": category,
+        "type": "",
+    }
+    try:
+        r = requests.get(EM_FLASH_API, params=params, headers=headers, timeout=15)
+        r.raise_for_status()
+        return r.json()
+    except Exception as e:
+        print(f"  [WARN] EastMoney flash request failed: {e}")
+        return None
+
+
+def parse_em_flash(resp: dict) -> list[dict]:
+    """解析东财快讯响应"""
+    try:
+        items = resp.get("data", []) or resp.get("list", []) or []
+        return items
+    except Exception:
+        return []
+
+
+def fetch_flash_pages(category: str = "全部", pages: int = 3, page_size: int = 30) -> pd.DataFrame:
+    """抓取多页快讯"""
+    all_items = []
+    for page in range(1, pages + 1):
+        resp = fetch_em_flash(category, page=page, page_size=page_size)
+        if resp is None:
+            break
+        items = parse_em_flash(resp)
+        if not items:
+            break
+        all_items.extend(items)
+
+    if not all_items:
+        return pd.DataFrame()
+
+    df = pd.DataFrame(all_items)
+    df.columns = [c.lower() for c in df.columns]
+
+    # 字段映射（东财字段 → 标准名）
+    col_map = {
+        "title":      "content",
+        "showtime":   "event_time",
+        "time":       "event_time",
+        "notice_date":"event_time",
+        "source":     "source",
+        "media":      "source",
+        "site":       "source",
+        "url":        "url",
+    }
+    df = df.rename(columns={k: v for k, v in col_map.items() if k in df.columns})
+
+    # 时间标准化
+    if "event_time" in df.columns:
+        df["event_time"] = pd.to_datetime(df["event_time"], errors="coerce").dt.strftime("%Y-%m-%d %H:%M:%S")
+
+    keep = ["content", "event_time", "source"]
+    df = df[[c for c in keep if c in df.columns]]
+    return df
+
+
+# ── 备选方案：AkShare 新闻 ───────────────────────────────────────────────────
+
+def fetch_akshare_news() -> pd.DataFrame:
+    """AkShare 东财财经新闻（带发布时间戳）
+    stock_news_em 返回列: [关键词, 股票代码, 新闻标题, 发布时间, 新闻来源, 新闻内容]
+    """
+    try:
+        df = ak.stock_news_em(symbol="A股")
+    except Exception as e:
+        print(f"  [WARN] AkShare news failed: {e}")
+        return pd.DataFrame()
+
+    if df is None or df.empty:
+        return pd.DataFrame()
+
+    # AkShare 返回中文列名，用位置映射更稳定
+    # 列0=关键词, 1=股票代码, 2=新闻标题(content), 3=发布时间, 4=新闻来源, 5=新闻内容
+    df.columns = ["keyword", "ts_code", "content", "event_time", "source", "body"]
+    if "event_time" in df.columns:
+        df["event_time"] = pd.to_datetime(df["event_time"], errors="coerce").dt.strftime("%Y-%m-%d %H:%M:%S")
+
+    keep = ["content", "event_time", "source"]
+    return df[keep]
+
+
+# ── 入口函数 ─────────────────────────────────────────────────────────────────
+
+def incremental_flash(source: str = "news_flash") -> pd.DataFrame:
+    """
+    增量采集：优先东财快讯，降级到 AkShare 新闻
+    """
+    # 尝试东财快讯（category="全部" 获取最广）
+    df = fetch_flash_pages(category="全部", pages=3)
+    if df is None or df.empty:
+        df = fetch_akshare_news()
+
+    if df is not None and not df.empty:
+        upsert_df(df, source, csv备份=False)
+        upsert_log(source, "", last_ts=datetime.now().strftime("%Y-%m-%d %H:%M:%S"))
+    return df
+
+
+if __name__ == "__main__":
+    import fire
+    fire.Fire({
+        "em":    lambda c="全部", p=3: upsert_df(fetch_flash_pages(c, pages=p), "news_flash"),
+        "ak":    lambda: upsert_df(fetch_akshare_news(), "news_flash"),
+        "incr":  lambda: incremental_flash(),
+    })

src/fetcher/stock_api.py

@@ -0,0 +1,179 @@
+"""
+日K线采集器 — AkShare（免费，无需 Token）
+支持前复权日线、分钟线（东财）
+"""
+
+import sys
+from pathlib import Path
+sys.path.insert(0, str(Path(__file__).parent.parent.parent))
+
+import akshare as ak
+import pandas as pd
+from datetime import datetime, timedelta
+from src.storage.db import upsert_df, upsert_rows, upsert_log, get_last_fetch
+
+
+def fetch_daily(
+    ts_code: str,
+    start_date: str = None,
+    end_date: str = None,
+    adjust: str = "qfq"
+) -> pd.DataFrame | None:
+    """
+    获取单只A股日K线（前复权）
+    ts_code: 6位代码，如 "000001"
+    start_date / end_date: "YYYYMMDD"
+    adjust: "qfq" 前复权 / "hfq" 后复权 / "" 不复权
+    """
+    try:
+        df = ak.stock_zh_a_hist(
+            symbol=ts_code,
+            period="daily",
+            start_date=start_date or (datetime.today() - timedelta(days=365)).strftime("%Y%m%d"),
+            end_date=end_date or datetime.today().strftime("%Y%m%d"),
+            adjust=adjust
+        )
+    except Exception as e:
+        print(f"  [WARN] stock_zh_a_hist failed for {ts_code}: {e}")
+        return None
+
+    if df is None or df.empty:
+        return None
+
+    # AkShare 列名统一转小写
+    df.columns = [c.lower() for c in df.columns]
+
+    # 统一列名映射（兼容多种 AkShare 版本列名）
+    col_map = {}
+    for old in df.columns:
+        ol = old.lower()
+        if ol in ("日期", "date"):
+            col_map[old] = "trade_date"
+        elif ol in ("股票代码", "代码", "symbol", "code"):
+            col_map[old] = "ts_code"
+        elif ol in ("开盘", "open"):
+            col_map[old] = "open"
+        elif ol in ("最高", "high"):
+            col_map[old] = "high"
+        elif ol in ("最低", "low"):
+            col_map[old] = "low"
+        elif ol in ("收盘", "close"):
+            col_map[old] = "close"
+        elif ol in ("成交量", "volume", "vol"):
+            col_map[old] = "volume"
+        elif ol in ("成交额", "amount", "amt"):
+            col_map[old] = "amount"
+    df = df.rename(columns=col_map)
+
+    # 确保 ts_code 列存在
+    if "ts_code" not in df.columns:
+        df["ts_code"] = ts_code
+
+    # 只保留核心列
+    keep = ["ts_code", "trade_date", "open", "high", "low", "close", "volume", "amount"]
+    df = df[[c for c in keep if c in df.columns]]
+
+    # 日期格式统一为 YYYY-MM-DD
+    if "trade_date" in df.columns:
+        df["trade_date"] = pd.to_datetime(df["trade_date"]).dt.strftime("%Y-%m-%d")
+
+    return df
+
+
+def fetch_index_daily(
+    symbol: str = "sh000001",   # 注意：AkShare 需 sh/zs 前缀，如 sh000001
+    start_date: str = None,
+    end_date: str = None
+) -> pd.DataFrame | None:
+    """获取指数日K线（用于大盘背景板）"""
+    try:
+        df = ak.stock_zh_index_daily(symbol=symbol)
+    except Exception as e:
+        print(f"  [WARN] stock_zh_index_daily failed for {symbol}: {e}")
+        return None
+
+    if df is None or df.empty:
+        return None
+    df.columns = [c.lower() for c in df.columns]
+
+    # AkShare 1.18+ 返回列名: date, open, high, low, close, volume（无 amount）
+    col_map = {"date": "trade_date", "open": "open", "high": "high",
+               "low": "low", "close": "close", "volume": "volume"}
+    df = df.rename(columns={k: v for k, v in col_map.items() if k in df.columns})
+    # amount 列不存在于指数，设为 NaN 保持 schema 一致
+    if "amount" not in df.columns:
+        df["amount"] = None
+
+    if "trade_date" in df.columns:
+        df["trade_date"] = pd.to_datetime(df["trade_date"]).dt.strftime("%Y-%m-%d")
+
+    # 按日期过滤
+    if start_date:
+        df = df[df["trade_date"] >= start_date]
+    if end_date:
+        df = df[df["trade_date"] <= end_date]
+
+    if "ts_code" not in df.columns:
+        df["ts_code"] = symbol
+    return df
+
+
+def batch_fetch_indices(
+    symbols: list[str] = None,
+    days: int = 90
+) -> pd.DataFrame:
+    """批量获取主要指数（沪深300/上证/创业板等）"""
+    # AkShare 前缀：sh=上证, sz=深圳
+    symbols = symbols or [
+        "sh000001",  # 上证指数
+        "sz399001",  # 深证成指
+        "sz399006",  # 创业板指
+        "sh000300",  # 沪深300
+        "sh000016",  # 上证50
+    ]
+    end   = datetime.today().strftime("%Y-%m-%d")
+    start = (datetime.today() - timedelta(days=days)).strftime("%Y-%m-%d")
+
+    results = []
+    for sym in symbols:
+        df = fetch_index_daily(sym)
+        if df is not None and not df.empty:
+            # 日期过滤在内存做
+            if "trade_date" in df.columns:
+                df = df[(df["trade_date"] >= start) & (df["trade_date"] <= end)]
+            if not df.empty:
+                results.append(df)
+
+    if results:
+        return pd.concat(results, ignore_index=True)
+    return pd.DataFrame()
+
+
+def run(ts_codes: list[str] = None, days: int = 90):
+    """
+    采集入口
+    ts_codes: 指定代码列表，None 时采集主要指数
+    """
+    end   = datetime.today().strftime("%Y%m%d")
+    start = (datetime.today() - timedelta(days=days)).strftime("%Y%m%d")
+
+    if ts_codes:
+        for code in ts_codes:
+            df = fetch_daily(code, start, end)
+            if df is not None and not df.empty:
+                upsert_rows(df, "stock_daily", conflict_cols=["ts_code", "trade_date"])
+                upsert_log("akshare_daily", code, last_date=end)
+    else:
+        # 默认采集主要指数
+        df = batch_fetch_indices(days=days)
+        if df is not None and not df.empty:
+            upsert_rows(df, "stock_daily", conflict_cols=["ts_code", "trade_date"])
+
+
+if __name__ == "__main__":
+    import fire
+    fire.Fire({
+        "daily": lambda ts_code: upsert_df(fetch_daily(ts_code), "stock_daily"),
+        "indices": lambda: run(),
+        "batch": lambda ts_codes: run(ts_codes=ts_codes),
+    })

src/pipeline/collect_all.py

@@ -0,0 +1,48 @@
+"""
+P0 全量采集脚本 — 依次执行所有 P0 数据源
+用法: python -m src.pipeline.collect_all
+"""
+import sys
+from pathlib import Path
+sys.path.insert(0, str(Path(__file__).parent.parent.parent))
+
+from datetime import datetime, timedelta
+
+def run():
+    print(f"\n{'='*50}")
+    print(f"  A股时间线数据采集 — {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}")
+    print(f"{'='*50}\n")
+
+    # 1. 日K线（主要指数）
+    print("[1/4] 日K线（主要指数 90天）...")
+    from src.fetcher import stock_api
+    stock_api.run()
+
+    # 2. 龙虎榜（近7日）
+    print("\n[2/4] 龙虎榜（近7日）...")
+    from src.fetcher import lhb_fetcher
+    start = (datetime.today() - timedelta(days=7)).strftime("%Y-%m-%d")
+    end   = datetime.today().strftime("%Y-%m-%d")
+    df = lhb_fetcher.fetch_lhb_range(start, end)
+    from src.storage.db import upsert_df, upsert_log
+    if df is not None and not df.empty:
+        upsert_df(df, "lhb_daily")
+        upsert_log("lhb_daily", "", last_date=end)
+
+    # 3. 7x24 快讯（最新3页）
+    print("\n[3/4] 7x24 快讯流...")
+    from src.fetcher import news_flash_fetcher
+    news_flash_fetcher.incremental_flash()
+
+    # 4. 大事提醒
+    print("\n[4/4] 每日大事提醒...")
+    from src.fetcher import macro_event_fetcher
+    macro_event_fetcher.incremental_events()
+
+    print(f"\n{'='*50}")
+    print(f"  采集完成")
+    print(f"{'='*50}\n")
+
+
+if __name__ == "__main__":
+    run()

src/realtime/realtime_analyzer.py

@@ -0,0 +1,276 @@
+# -*- coding: utf-8 -*-
+"""
+实时市场时间线分析器（常驻运行，不自动退出）
+==========================================
+功能（每分钟一轮，按市场状态自动分派）：
+  1. 盘前（09:15 前）：美股隔夜收盘 → 依据 a-stock-timeline-patterns 实证先验
+     输出今日 A 股跳空/日内路径的历史概率预估
+  2. 盘中（09:15-15:05）：上证指数实时点位相对昨收位置 + 快讯关键词告警
+  3. 盘后（15:10）：当日龙虎榜入库 + 日报（指数涨跌/跳空/日内分解）
+输出：控制台 + 追加写入 ~/.a_stock_timeline/realtime_feed.jsonl 与 signal 日志
+停止：Ctrl+C（优雅退出）
+"""
+import io
+import json
+import os
+import sqlite3
+import sys
+import time
+import traceback
+from datetime import datetime, timedelta
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+import warnings
+warnings.filterwarnings('ignore')
+
+import akshare as ak
+import requests
+
+ROOT = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+DB = os.path.join(ROOT, 'data', 'a_stock.db')
+FEED = os.path.join(os.path.expanduser('~'), '.a_stock_timeline', 'realtime_feed.jsonl')
+SKILL_REF = r'E:\Data\skills\a-stock-timeline-patterns\references\overnight_transmission.md'
+
+KEYWORDS = ['降息', '降准', '加息', '关税', '制裁', '收购', '重组', '国债', '证监会', 'PMI', 'CPI']
+
+HIS = None   # 历史先验缓存 {('NDX'|'SPX'|'DJI', bin): {...}}
+
+
+def load_priors():
+    """解析实证先验表（overnight_transmission.md），供盘前预估"""
+    global HIS
+    import re
+    HIS = {}
+    path = SKILL_REF
+    if not os.path.exists(path):
+        return
+    cur_bin, cur_us, cur_idx = None, None, None
+    with io.open(path, encoding='utf-8') as f:
+        for line in f:
+            m = re.match(r'### (\S+)T日 -> (\S+)T\+1日', line)
+            if m:
+                cur_us, cur_idx = m.group(1), m.group(2)
+                continue
+            m = re.match(r'- 美股(\S+?)（(\d+)天）: A股跳空 N=\d+, 均值 (-?[\d.]+)%, 中位数 (-?[\d.]+)%, 胜率 ([\d.]+)%; '
+                         r'日内 N=\d+, 均值 (-?[\d.]+)%, 中位数 (-?[\d.]+)%, 胜率 ([\d.]+)%', line)
+            if m and cur_us and cur_idx:
+                cur_bin = m.group(1)
+                HIS[(cur_us, cur_idx, cur_bin)] = {
+                    'n': int(m.group(2)),
+                    'gap_mean': float(m.group(3)), 'gap_median': float(m.group(4)),
+                    'gap_win': float(m.group(5)),
+                    'intra_mean': float(m.group(6)), 'intra_median': float(m.group(7)),
+                    'intra_win': float(m.group(8)),
+                }
+    print('[init] 实证先验加载: {} 条'.format(len(HIS)), flush=True)
+
+
+def db_conn():
+    conn = sqlite3.connect(DB, check_same_thread=False)
+    return conn
+
+
+def us_overnight(conn):
+    """最近一个已收盘的美股交易日涨跌（三大指数）"""
+    out = {}
+    for code in ('US.NDX', 'US.SPX', 'US.DJI'):
+        row = conn.execute(
+            "SELECT trade_date, pct_change FROM global_index WHERE index_code=? "
+            "ORDER BY trade_date DESC LIMIT 1", (code,)).fetchone()
+        if row:
+            out[code] = {'date': row[0], 'pct': row[1] if row[1] is not None else 0.0}
+    return out
+
+
+def forecast_from_priors(us):
+    """按先验表预估：取纳斯达克分箱对应的上证跳空/日内历史分布"""
+    if not HIS or 'US.NDX' not in us:
+        return None
+    pct = us['US.NDX']['pct'] / 100.0
+    if pct <= -0.015:
+        b = '跌>1.5%'
+    elif pct <= -0.005:
+        b = '跌0.5~1.5%'
+    elif pct < 0.005:
+        b = '正负0.5%内'
+    elif pct < 0.015:
+        b = '涨0.5~1.5%'
+    else:
+        b = '涨>1.5%'
+    k = ('US.NDX', 'sh000001', b)
+    if k not in HIS:
+        return None
+    p = HIS[k]
+    return ('隔夜预估[美股纳指{:+.2f}% -> 分箱"{}"]: 历史上上证次日跳空均值 {:+.2f}%（低开概率 {:.0f}%），'
+            '日内均值 {:+.2f}%（日内收涨概率 {:.0f}%），全天均值 {:+.2f}%。'
+            '（样本{}天，美股99-24先验，仅参考）').format(
+        us['US.NDX']['pct'], b, p['gap_mean'], 100 - p['gap_win'],
+        p['intra_mean'], p['intra_win'], p['gap_mean'] + p['intra_mean'], p['n'])
+
+
+def index_spot_sh():
+    """上证指数实时点位（新浪，免token）"""
+    try:
+        df = ak.stock_zh_index_spot_em(symbol='上证系列指数')
+        row = df[df['名称'] == '上证指数']
+        if not row.empty:
+            r = row.iloc[0]
+            return {'price': float(r['最新价']), 'pct': float(r['涨跌幅'])}
+    except Exception:
+        pass
+    try:
+        df = ak.stock_zh_index_spot_sina()
+        row = df[df['代码'] == 'sh000001']
+        if not row.empty:
+            r = row.iloc[0]
+            return {'price': float(r['最新价']), 'pct': float(r['涨跌幅'])}
+    except Exception:
+        pass
+    return None
+
+
+def today_lhb_count(conn):
+    today = datetime.now().strftime('%Y-%m-%d')
+    n = conn.execute("SELECT COUNT(*) FROM lhb_daily WHERE trade_date=?", (today,)).fetchone()[0]
+    return n
+
+
+def news_alerts(conn):
+    """扫描快讯/新闻关键词命中（近2小时新增）"""
+    hits = []
+    cutoff = (datetime.now() - timedelta(minutes=30)).strftime('%Y-%m-%d %H:%M:%S')
+    for table, tcol, ccol in (('news_flash', 'event_time', 'content'),
+                              ('news_cn', 'pub_date', 'title')):
+        try:
+            rows = conn.execute(
+                f"SELECT {tcol}, {ccol} FROM {table} WHERE {tcol} >= ? ORDER BY {tcol} DESC LIMIT 50",
+                (cutoff,)).fetchall()
+        except sqlite3.OperationalError:
+            continue
+        for ts, text in rows:
+            for kw in KEYWORDS:
+                if kw in (text or ''):
+                    hits.append({'time': ts, 'kw': kw, 'text': (text or '')[:80]})
+                    break
+    return hits
+
+
+def emit(kind, payload):
+    rec = {'ts': datetime.now().strftime('%Y-%m-%d %H:%M:%S'), 'kind': kind, 'data': payload}
+    line = json.dumps(rec, ensure_ascii=False)
+    os.makedirs(os.path.dirname(FEED), exist_ok=True)
+    with io.open(FEED, 'a', encoding='utf-8') as f:
+        f.write(line + '\n')
+    print('[{}] {} {}'.format(rec['ts'], kind, json.dumps(payload, ensure_ascii=False)), flush=True)
+
+
+def save_lhb_today(conn):
+    """收盘后采集当日龙虎榜"""
+    try:
+        from src.fetcher.lhb_fetcher import fetch_lhb   # 项目内已有实现
+        df = fetch_lhb()
+        if df is not None and not df.empty:
+            from src.storage.db import upsert_rows
+            upsert_rows(df, 'lhb_daily', conflict_cols=['trade_date', 'ts_code', 'reason'])
+            return len(df)
+    except Exception as e:
+        print('[lhb] 采集失败:', e, flush=True)
+    return 0
+
+
+
+
+def fetch_em_flash(conn):
+    """东财 7x24 快讯轮询：增量入库 news_flash（按 content+event_time 去重）"""
+    try:
+        r = requests.get(
+            'https://np-weblist.eastmoney.com/comm/web/getFastNewsList',
+            params={'client': 'web', 'biz': 'web_724', 'fastColumn': '102',
+                    'sortEnd': '', 'pageSize': '20', 'req_trace': '1'},
+            headers={'User-Agent': 'Mozilla/5.0'}, timeout=10)
+        data = r.json().get('data', {}) or {}
+        rows = []
+        for n in data.get('fastNewsList', []) or []:
+            ts = n.get('showTime', '')
+            summary = (n.get('summary') or n.get('title') or '').strip()
+            if ts and summary:
+                dup = conn.execute(
+                    "SELECT 1 FROM news_flash WHERE event_time=? AND content=? LIMIT 1",
+                    (ts, summary)).fetchone()
+                if not dup:
+                    conn.execute(
+                        "INSERT INTO news_flash(content, event_time, source, inserted_at) "
+                        "VALUES (?,?,?,datetime('now','localtime'))",
+                        (summary, ts, '东财7x24'))
+                    rows.append((ts, summary[:60]))
+        conn.commit()
+        return rows
+    except Exception as e:
+        print('[flash] 拉取失败:', e, flush=True)
+        return []
+
+def one_cycle(conn, state):
+    now = datetime.now()
+    t = now.time()
+    hm = t.hour * 100 + t.minute
+
+    # 1) 盘前（07:00-09:25）：隔夜预估
+    if 700 <= hm < 925 and not state.get('premarket_done'):
+        us = us_overnight(conn)
+        fc = forecast_from_priors(us)
+        emit('盘前隔夜预估', {'us': us, 'forecast': fc})
+        state['premarket_done'] = True
+
+    # 2) 盘中（09:25-15:05）：实时点位 + 快讯告警
+    if 925 <= hm < 1505:
+        for ts, txt in fetch_em_flash(conn):
+            emit('新快讯', {'time': ts, 'text': txt})
+        spot = index_spot_sh()
+        if spot and spot.get('price', 0) > 0:
+            if 'session_open_price' not in state or hm < 940 and state.get('last_price') is None:
+                pass
+            emit('盘中点位', {'上证': spot['price'], '涨跌幅%': spot['pct']})
+        hits = news_alerts(conn)
+        for h in hits:
+            emit('快讯关键词告警', h)
+
+    # 3) 盘后（15:10 后）：龙虎榜 + 日报
+    if hm >= 1510 and not state.get('postmarket_done'):
+        n = save_lhb_today(conn)
+        sh = conn.execute("SELECT trade_date, close FROM stock_daily WHERE ts_code='sh000001' "
+                          "ORDER BY trade_date DESC LIMIT 1").fetchone()
+        emit('收盘日报', {'龙虎榜新增': n, '上证最新收盘': sh})
+        state['postmarket_done'] = True
+
+    # 4) 日切重置
+    if state.get('date') != now.strftime('%Y-%m-%d'):
+        state.clear()
+        state['date'] = now.strftime('%Y-%m-%d')
+        load_priors()
+        emit('日切', {'date': state['date']})
+
+
+def main():
+    print('=== JQuant 市场时间线实时分析器 ===')
+    print('数据库:', DB)
+    print('输出流:', FEED)
+    print('Ctrl+C 停止', flush=True)
+    conn = db_conn()
+    load_priors()
+    state = {'date': datetime.now().strftime('%Y-%m-%d')}
+    interval = 60
+    while True:
+        try:
+            one_cycle(conn, state)
+        except KeyboardInterrupt:
+            print('收到停止指令，退出', flush=True)
+            break
+        except Exception:
+            traceback.print_exc()
+            time.sleep(10)
+        time.sleep(interval)
+
+
+if __name__ == '__main__':
+    main()

src/storage/__init__.py

@@ -0,0 +1 @@
+# storage package

src/storage/db.py

@@ -0,0 +1,250 @@
+"""
+SQLite + CSV 双写存储层
+表结构按"接口族"隔离，方便按需降级
+"""
+
+import sqlite3
+from pathlib import Path
+import pandas as pd
+from datetime import datetime
+from typing import Optional, Literal
+
+DB_PATH = Path(__file__).parent.parent.parent / "data" / "a_stock.db"
+CSV_DIR  = Path(__file__).parent.parent.parent / "data" / "raw"
+
+# ─────────────────────────────────────────────────────────────────────────────
+# 连接管理器
+# ─────────────────────────────────────────────────────────────────────────────
+
+def get_conn() -> sqlite3.Connection:
+    DB_PATH.parent.mkdir(parents=True, exist_ok=True)
+    conn = sqlite3.connect(DB_PATH, check_same_thread=False)
+    conn.execute("PRAGMA journal_mode=WAL")
+    conn.execute("PRAGMA foreign_keys=ON")
+    return conn
+
+
+def to_csv(df: pd.DataFrame, table: str, ts: Optional[str] = None):
+    """入库后同步写一份 CSV 备份（方便溯源）"""
+    if df is None or df.empty:
+        return
+    ts = ts or datetime.now().strftime("%Y%m%d_%H%M%S")
+    CSV_DIR.mkdir(parents=True, exist_ok=True)
+    path = CSV_DIR / f"{table}_{ts}.csv"
+    df.to_csv(path, index=False, encoding="utf-8-sig")
+    return path
+
+
+# ─────────────────────────────────────────────────────────────────────────────
+# Schema 创建
+# ─────────────────────────────────────────────────────────────────────────────
+
+SCHEMA_SQL = """
+-- 交易日历
+CREATE TABLE IF NOT EXISTS trade_calendar (
+    trade_date  TEXT PRIMARY KEY,
+    is_trade_day INTEGER NOT NULL
+);
+
+-- A股日K线（前复权）
+CREATE TABLE IF NOT EXISTS stock_daily (
+    ts_code      TEXT    NOT NULL,
+    trade_date   TEXT    NOT NULL,
+    open         REAL,
+    high         REAL,
+    low          REAL,
+    close        REAL,
+    volume       REAL,
+    amount       REAL,
+    PRIMARY KEY (ts_code, trade_date)
+);
+
+-- 龙虎榜
+CREATE TABLE IF NOT EXISTS lhb_daily (
+    trade_date       TEXT    NOT NULL,
+    ts_code          TEXT    NOT NULL,
+    name             TEXT,
+    close            REAL,
+    pct_change       REAL,
+    amount           REAL,
+    reason           TEXT,
+    buy_seats        INTEGER,
+    sell_seats       INTEGER,
+    net_amount       REAL,
+    PRIMARY KEY (trade_date, ts_code, reason)
+);
+
+-- 7x24 快讯流（精确到秒）
+CREATE TABLE IF NOT EXISTS news_flash (
+    id           INTEGER PRIMARY KEY AUTOINCREMENT,
+    content      TEXT    NOT NULL,
+    event_time   TEXT    NOT NULL,
+    source       TEXT,
+    inserted_at  TEXT    NOT NULL
+);
+
+-- 每日大事提醒（政策级）
+CREATE TABLE IF NOT EXISTS macro_event (
+    id           INTEGER PRIMARY KEY AUTOINCREMENT,
+    event_date   TEXT    NOT NULL,
+    event_time   TEXT,
+    title        TEXT    NOT NULL,
+    event_type   INTEGER,
+    source       TEXT,
+    actual       TEXT,
+    forecast     TEXT,
+    prev         TEXT,
+    inserted_at  TEXT    NOT NULL DEFAULT (datetime('now','localtime'))
+);
+
+-- 全网财经新闻
+CREATE TABLE IF NOT EXISTS news_cn (
+    id           INTEGER PRIMARY KEY AUTOINCREMENT,
+    title        TEXT    NOT NULL,
+    content      TEXT,
+    pub_date     TEXT    NOT NULL,
+    author       TEXT,
+    source       TEXT,
+    url          TEXT,
+    keywords     TEXT,
+    inserted_at  TEXT    NOT NULL
+);
+
+-- 北向/主力资金流（日频）
+CREATE TABLE IF NOT EXISTS money_flow (
+    ts_code      TEXT    NOT NULL,
+    trade_date   TEXT    NOT NULL,
+    main_net_in  REAL,
+    in_flow      REAL,
+    out_flow     REAL,
+    amount       REAL,
+    pct_change   REAL,
+    PRIMARY KEY (ts_code, trade_date)
+);
+
+-- 全球指数（日频，外盘收盘后入库）
+CREATE TABLE IF NOT EXISTS global_index (
+    index_code   TEXT    NOT NULL,
+    trade_date   TEXT    NOT NULL,
+    close        REAL,
+    pct_change   REAL,
+    PRIMARY KEY (index_code, trade_date)
+);
+
+-- 采集记录（断点续采用）
+CREATE TABLE IF NOT EXISTS fetch_log (
+    source      TEXT    NOT NULL,
+    param       TEXT,
+    last_ts     TEXT,
+    last_date   TEXT,
+    updated_at  TEXT    NOT NULL,
+    PRIMARY KEY (source, param)
+);
+"""
+
+def init_db():
+    with get_conn() as conn:
+        conn.executescript(SCHEMA_SQL)
+    print(f"[OK] Database initialized: {DB_PATH}")
+
+
+# ─────────────────────────────────────────────────────────────────────────────
+# 通用 upsert
+# ─────────────────────────────────────────────────────────────────────────────
+
+def upsert_df(
+    df: pd.DataFrame,
+    table: str,
+    pk_cols: list[str] = None,   # 暂未使用，to_sql(replace) 已处理主键冲突
+    csv备份: bool = True
+) -> int:
+    """主键冲突时 replace，等效于 upsert"""
+    if df is None or df.empty:
+        return 0
+    # 自动追加 inserted_at / updated_at
+    now = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+    if "inserted_at" in df.columns:
+        df["inserted_at"] = now
+    elif table in ("news_flash", "macro_event", "news_cn"):
+        df["inserted_at"] = now
+
+    with get_conn() as conn:
+        # sqlite 不支持 df 直接 execute，需要逐行或用 executemany
+        df.to_sql(table, conn, if_exists="replace", index=False)
+        row_count = len(df)
+
+    if csv备份:
+        to_csv(df, table)
+    print(f"  [OK] {table}: {row_count} rows upserted")
+    return row_count
+
+
+def upsert_log(source: str, param: str, last_ts: str = None, last_date: str = None):
+    now = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+    with get_conn() as conn:
+        conn.execute("""
+            INSERT INTO fetch_log (source, param, last_ts, last_date, updated_at)
+            VALUES (?, ?, ?, ?, ?)
+            ON CONFLICT(source, param) DO UPDATE SET
+                last_ts    = COALESCE(excluded.last_ts,    last_ts),
+                last_date  = COALESCE(excluded.last_date,  last_date),
+                updated_at = excluded.updated_at
+        """, (source, param, last_ts, last_date, now))
+
+
+def get_last_fetch(source: str, param: str = "") -> dict:
+    with get_conn() as conn:
+        row = conn.execute(
+            "SELECT last_ts, last_date, updated_at FROM fetch_log WHERE source=? AND param=?",
+            (source, param)
+        ).fetchone()
+        if row:
+            return {"last_ts": row[0], "last_date": row[1], "updated_at": row[2]}
+    return {}
+
+
+if __name__ == "__main__":
+    init_db()
+
+# ---------------------------------------------------------------------------
+# 保历史的 upsert（INSERT OR REPLACE，不清空表）
+# ---------------------------------------------------------------------------
+
+_IDENT_RE = __import__('re').compile(r'^[A-Za-z0-9_]+$')
+
+
+def _safe_ident(name: str) -> str:
+    """SQL 标识符白名单校验（列名/表名不可参数化，只允许字母数字下划线）"""
+    if not _IDENT_RE.match(str(name)):
+        raise ValueError("illegal identifier: %r" % (name,))
+    return str(name)
+
+
+def ensure_index(table: str, cols: list) -> None:
+    """按冲突列建唯一索引（幂等）"""
+    tab = _safe_ident(table)
+    col_ids = [_safe_ident(c) for c in cols]
+    idx = "ux_{}_{}".format(tab, "_".join(col_ids))
+    with get_conn() as conn:
+        conn.execute(
+            "CREATE UNIQUE INDEX IF NOT EXISTS {} ON {}({})".format(
+                idx, tab, ",".join(col_ids)))
+
+
+def upsert_rows(df, table: str, conflict_cols: list) -> int:
+    """INSERT OR REPLACE 按 conflict_cols 去重，保留既有历史。值全部走 ? 占位符。"""
+    if df is None or df.empty:
+        return 0
+    now = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+    if "inserted_at" in df.columns:
+        df["inserted_at"] = now
+    tab = _safe_ident(table)
+    col_ids = [_safe_ident(c) for c in df.columns]
+    sql = "INSERT OR REPLACE INTO {}({}) VALUES ({})".format(
+        tab, ",".join(col_ids), ",".join("?" * len(col_ids)))
+    rows = [tuple(None if (isinstance(v, float) and v != v) else v
+                  for v in tup) for tup in df.itertuples(index=False, name=None)]
+    with get_conn() as conn:
+        conn.executemany(sql, rows)
+    print("  [OK] {}: {} rows upserted (conflict on {})".format(tab, len(rows), conflict_cols))
+    return len(rows)