提交

backend/app/api/__init__.py

@@ -5,6 +5,7 @@
   market.py    /api/market/*（总览/打板/概念板块/指数 K 线与详情）
   backtest.py  /api/backtest + /api/backtest/event
   screener.py  /api/screener/*（选股/历史/同步/个股预览）
+  signals.py   /api/signals*（买卖点实验室：方案/分析/扫描/评估）
   user.py      /api/preferences + /api/watchlist* + /api/trades*
   _deps.py     共享件：JSON 直返缓存、复权换算、行转 Bar、共享常量与 SQL
 
@@ -18,6 +19,7 @@ from .backtest import router as backtest_router
 from .etfs import router as etfs_router
 from .market import router as market_router
 from .screener import router as screener_router
+from .signals import router as signals_router
 from .stocks import router as stocks_router
 from .user import router as user_router
 
@@ -27,4 +29,5 @@ router.include_router(etfs_router)
 router.include_router(market_router)
 router.include_router(backtest_router)
 router.include_router(screener_router)
+router.include_router(signals_router)
 router.include_router(user_router)

backend/app/backtest/events.py

@@ -135,12 +135,16 @@ def _scan_batch(
     if not candle_rows:
         return trades
     bars = pd.DataFrame(
-        candle_rows, columns=["symbol", "ts", "open", "high", "low", "close"]
+        candle_rows, columns=["symbol", "ts", "open", "high", "low", "close", "volume", "turnover"]
     )
     for symbol, g in bars.groupby("symbol", sort=False):
         if len(g) < 30:
             continue
         g = g.reset_index(drop=True)
+        # pct_chg / amplitude 是派生列（candles 无现成涨跌幅/振幅）：按股内环比补算，
+        # 与选股引擎 _load_bars 同口径——否则「日涨跌幅/当日振幅」条件进事件回测直接 KeyError
+        g["pct_chg"] = g["close"].pct_change() * 100
+        g["amplitude"] = (g["high"] - g["low"]) / g["close"].shift(1) * 100
         ts_code_l = code_by_symbol[symbol]
         cache: dict = {"_families": set()}
         mask = _signal_mask(g, spec, cache)
@@ -243,7 +247,7 @@ async def run_event_backtest(
         code_by_symbol = {sym: code for code, sym in batch}
         candle_rows = (await session.execute(
             select(Candle.symbol, Candle.ts, Candle.open, Candle.high,
-                   Candle.low, Candle.close)
+                   Candle.low, Candle.close, Candle.volume, Candle.turnover)
             .where(and_(Candle.timeframe == "1d",
                         Candle.symbol.in_(symbols),
                         Candle.ts >= buffer_ts, Candle.ts <= end_ts))

backend/app/indicators.py

@@ -56,6 +56,15 @@ def ma(close: pd.Series, period: int) -> pd.Series:
     return close.rolling(period, min_periods=1).mean()
 
 
+def vol_ratio(volume: pd.Series, period: int = 5) -> pd.Series:
+    """量比（通达信口径）：当日成交量 / 前 N 日均量（不含当日，即 REF(MA(VOL,N),1)）。
+
+    >2 约=倍量、<0.7 约=显著缩量；首根无前值、均量为 0（长期停牌）时为 NaN。
+    """
+    base = volume.rolling(period, min_periods=1).mean().shift(1)
+    return volume / base.replace(0, np.nan)
+
+
 def ema2(close: pd.Series, span: int = 10) -> pd.Series:
     """知行短期趋势线：EMA(EMA(C, span), span)。"""
     return ema(ema(close, span), span)

backend/app/models.py

@@ -362,6 +362,44 @@ class ScreenerQuery(Base):
     created_at: Mapped[datetime] = mapped_column(DateTime, default=_utcnow, index=True)
 
 
+class SignalPlan(Base):
+    """买卖点实验室方案（一股一方案：手动标注/条件扫描的买卖点集合与指标条件）。
+
+    conditions_json 存 IndicatorCondition[].model_dump_json()（结构化条件可直接
+    推送到 /api/backtest/event 做全市场事件回测）；点集在 signal_points 子表。
+    """
+    __tablename__ = "signal_plans"
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True, autoincrement=True)
+    user_id: Mapped[int] = mapped_column(BigInteger, ForeignKey("users.id", ondelete="CASCADE"), index=True)
+    ts_code: Mapped[str] = mapped_column(String(12), index=True)
+    name: Mapped[str] = mapped_column(String(64))
+    note: Mapped[str | None] = mapped_column(Text)
+    conditions_json: Mapped[str | None] = mapped_column(Text)
+    created_at: Mapped[datetime] = mapped_column(DateTime, default=_utcnow)
+    updated_at: Mapped[datetime] = mapped_column(DateTime, default=_utcnow, onupdate=_utcnow)
+
+    __table_args__ = (
+        Index("ix_signal_plans_user_code", "user_id", "ts_code"),
+    )
+
+
+class SignalPoint(Base):
+    """方案内的买卖点（manual=K线图手动点击标注；scan=条件扫描命中后勾选保留）。"""
+    __tablename__ = "signal_points"
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True, autoincrement=True)
+    plan_id: Mapped[int] = mapped_column(Integer, ForeignKey("signal_plans.id", ondelete="CASCADE"), index=True)
+    kind: Mapped[str] = mapped_column(String(4))        # buy | sell
+    trade_date: Mapped[date] = mapped_column(Date)
+    source: Mapped[str] = mapped_column(String(6))      # manual | scan
+    created_at: Mapped[datetime] = mapped_column(DateTime, default=_utcnow)
+
+    __table_args__ = (
+        UniqueConstraint("plan_id", "kind", "trade_date", name="uq_signal_point_plan_kind_date"),
+    )
+
+
 class TradeCalendar(Base):
     """交易日历缓存（trade_cal 拉取一次宽范围后本地维护，低积分 token 限频 1 次/小时）。"""
     __tablename__ = "trade_calendar"

backend/app/schemas.py

@@ -710,3 +710,150 @@ class IndexWeightsResponse(BaseModel):
     total: int                       # 成分股总数（返回 items 为按权重降序的子集）
     items: list[IndexWeightItemOut] = []
 
+
+# ---------- 买卖点实验室（/signals：手动标注共同点分析 + 条件扫描） ----------
+class SignalPlanCreate(BaseModel):
+    ts_code: str = Field(min_length=6, max_length=12)
+    name: str = Field(default="", max_length=64)   # 空则后端默认「{code} 买点方案」
+
+
+class SignalPlanUpdate(BaseModel):
+    name: str | None = Field(default=None, min_length=1, max_length=64)
+    note: str | None = None
+    conditions: list[IndicatorCondition] | None = None   # None=不改；[] = 清空
+
+
+class SignalPointIn(BaseModel):
+    kind: Literal["buy", "sell"]
+    trade_date: date
+    source: Literal["manual", "scan"] = "manual"
+
+
+class SignalPointOut(BaseModel):
+    kind: str                        # buy | sell
+    trade_date: date
+    source: str                      # manual | scan
+
+    model_config = {"from_attributes": True}
+
+
+class SignalPlanOut(BaseModel):
+    id: int
+    ts_code: str
+    name: str
+    note: str | None = None
+    conditions: list[IndicatorCondition] = Field(default_factory=list)
+    points: list[SignalPointOut] = Field(default_factory=list)
+    created_at: datetime
+    updated_at: datetime
+
+
+class SignalPlanListResponse(BaseModel):
+    items: list[SignalPlanOut]
+
+
+class SignalsAnalyzeRequest(BaseModel):
+    ts_code: str = Field(min_length=6, max_length=12)
+    buy_dates: list[date] = Field(min_length=1, max_length=200)
+
+
+class FeatureFindingOut(BaseModel):
+    """共同特征发现：特征在买点日的命中率 vs 全期基线比例（纯算法统计）。"""
+    key: str
+    label: str
+    samples_hit: int                 # 命中特征的买点数
+    samples: int                     # 有效买点数
+    hit_rate: float                  # 买点命中率 %（0-100）
+    base_rate: float                 # 全期基线比例 %（0-100）
+    lift: float                      # 命中率 / 基线
+    mappable: bool                   # 是否可转成 IndicatorCondition（进扫描/回测）
+    condition: IndicatorCondition | None = None
+
+
+class ValueSummaryOut(BaseModel):
+    """连续值摘要：买点日中位数 vs 全期中位数（直觉对照用）。"""
+    key: str
+    label: str
+    at_points: float | None
+    baseline: float | None
+
+
+class SignalsAnalyzeResponse(BaseModel):
+    ts_code: str
+    samples: int                     # 有效买点数（落在该股交易日上）
+    total_bars: int
+    findings: list[FeatureFindingOut]
+    summaries: list[ValueSummaryOut]
+
+
+class WindowStatOut(BaseModel):
+    window: int                      # 持有 N 交易日
+    samples: int
+    mean: float                      # 平均收益 %
+    median: float
+    win_rate: float                  # 收益>0 占比 %
+    p10: float
+    p90: float
+
+
+class EvalPointOut(BaseModel):
+    """单个买点的评估：未来 N 交易日收益矩阵（复权校正）。"""
+    date: date                       # 买点（信号）日
+    entry_date: date | None = None   # 实际入场日（默认信号次日）
+    entry_price: float | None = None
+    rets: dict[str, float | None] = Field(default_factory=dict)      # "5" -> 收益%（越界/缺数据 None）
+    max_gain: dict[str, float | None] = Field(default_factory=dict)  # 窗口内最大涨幅 %
+    max_dd: dict[str, float | None] = Field(default_factory=dict)    # 窗口内最大回撤 %
+
+
+class EvalPairOut(BaseModel):
+    """手动配对（买点 -> 其后最近卖点）的一笔评估。"""
+    buy_date: date
+    sell_date: date
+    entry_date: date | None = None
+    entry_price: float | None = None
+    exit_price: float | None = None
+    ret_pct: float | None = None     # 复权校正收益 %
+
+
+class PairStatsOut(BaseModel):
+    samples: int
+    mean: float
+    median: float
+    win_rate: float
+    max: float
+    min: float
+
+
+class SignalsEvaluateResult(BaseModel):
+    samples: int
+    points: list[EvalPointOut] = Field(default_factory=list)
+    stats: list[WindowStatOut] = Field(default_factory=list)
+    pairs: list[EvalPairOut] = Field(default_factory=list)
+    pair_stats: PairStatsOut | None = None
+
+
+class SignalsScanRequest(BaseModel):
+    ts_code: str = Field(min_length=6, max_length=12)
+    conditions: list[IndicatorCondition] = Field(min_length=1, max_length=10)
+    windows: list[int] = Field(default_factory=lambda: [1, 3, 5, 10, 20, 60])
+    start: date | None = None
+    end: date | None = None
+
+
+class SignalsScanResponse(BaseModel):
+    ts_code: str
+    dates: list[date]                # 命中信号日（升序，已丢弃指标预热区）
+    total: int
+    evaluate: SignalsEvaluateResult
+
+
+class SignalsEvaluateRequest(BaseModel):
+    ts_code: str = Field(min_length=6, max_length=12)
+    buy_dates: list[date] = Field(min_length=1, max_length=500)
+    sell_dates: list[date] = Field(default_factory=list, max_length=500)
+    windows: list[int] = Field(default_factory=lambda: [1, 3, 5, 10, 20, 60])
+    # 进出场时机默认与 /api/backtest/event 对齐：信号次日入场、到期收盘卖出
+    entry_timing: Literal["next_open", "next_close"] = "next_open"
+    exit_timing: Literal["close", "open"] = "close"
+

backend/app/screener/engine.py

@@ -56,6 +56,10 @@ def _ma(df: pd.DataFrame, p: dict) -> dict[str, pd.Series]:
     return {"ma": ind.ma(df["close"], period=int(p["period"]))}
 
 
+def _vol_ratio(df: pd.DataFrame, p: dict) -> dict[str, pd.Series]:
+    return {"vol_ratio": ind.vol_ratio(df["volume"], period=int(p["period"]))}
+
+
 def _boll(df: pd.DataFrame, p: dict) -> dict[str, pd.Series]:
     out = ind.bollinger(df["close"], period=int(p["period"]), std=float(p["std"]))
     return {"boll_upper": out["upper"], "boll_mid": out["mid"], "boll_lower": out["lower"]}
@@ -75,6 +79,16 @@ def _pct_chg(df: pd.DataFrame, p: dict) -> dict[str, pd.Series]:
     return {"pct_chg": df["pct_chg"]}
 
 
+def _amplitude(df: pd.DataFrame, p: dict) -> dict[str, pd.Series]:
+    # 当日振幅 %（通达信口径：(最高-最低)/昨收×100），依赖派生列 amplitude
+    return {"amplitude": df["amplitude"]}
+
+
+def _turnover_rate(df: pd.DataFrame, p: dict) -> dict[str, pd.Series]:
+    # 逐日换手率 %（candles.turnover 直读；ETF/缺数据日为 NaN，条件按 False 处理）
+    return {"turnover_rate": df["turnover"]}
+
+
 @dataclass(frozen=True)
 class FamilyDef:
     label: str                     # 族中文标签（条件回显/表头）
@@ -89,20 +103,24 @@ FAMILIES: dict[str, FamilyDef] = {
     "macd": FamilyDef("MACD", ("fast", "slow", "signal"), {"fast": 12, "slow": 26, "signal": 9}, 60, _macd),
     "rsi": FamilyDef("RSI", ("period",), {"period": 14}, 25, _rsi),
     "ma": FamilyDef("MA", ("period",), {"period": 20}, 25, _ma),
+    "vol_ratio": FamilyDef("量比", ("period",), {"period": 5}, 10, _vol_ratio),
     "boll": FamilyDef("BOLL", ("period", "std"), {"period": 20, "std": 2}, 25, _boll),
     "zhixing": FamilyDef("知行", ("m1", "m2", "m3", "m4"),
                          {"m1": 14, "m2": 28, "m3": 57, "m4": 114}, 114, _zhixing),
     "close": FamilyDef("收盘价", (), {}, 1, _close),
     "pct_chg": FamilyDef("日涨跌幅", (), {}, 1, _pct_chg),
+    "amplitude": FamilyDef("当日振幅", (), {}, 2, _amplitude),
+    "turnover_rate": FamilyDef("换手率", (), {}, 1, _turnover_rate),
 }
 
 INDICATOR_FAMILY: dict[str, str] = {
     "kdj_k": "kdj", "kdj_d": "kdj", "kdj_j": "kdj",
     "macd_dif": "macd", "macd_dea": "macd", "macd_hist": "macd",
-    "rsi": "rsi", "ma": "ma",
+    "rsi": "rsi", "ma": "ma", "vol_ratio": "vol_ratio",
     "boll_upper": "boll", "boll_mid": "boll", "boll_lower": "boll",
     "zhixing_dkx": "zhixing", "zhixing_trend": "zhixing",
     "close": "close", "pct_chg": "pct_chg",
+    "amplitude": "amplitude", "turnover_rate": "turnover_rate",
 }
 
 _IND_SUFFIX = {"kdj_k": "K", "kdj_d": "D", "kdj_j": "J",
@@ -317,23 +335,27 @@ async def _load_bars(session: AsyncSession, ts_codes: list[str],
     """
     symbols = [plain_code(t) for t in ts_codes]
     stmt = select(
-        Candle.symbol, Candle.ts, Candle.open, Candle.high, Candle.low, Candle.close,
+        Candle.symbol, Candle.ts, Candle.open, Candle.high, Candle.low, Candle.close, Candle.volume,
+        Candle.turnover,
     ).where(Candle.timeframe == "1d", Candle.ts >= min_date, Candle.ts <= target_date)
     if len(symbols) <= 2000:
         stmt = stmt.where(Candle.symbol.in_(set(symbols)))
     rows = (await session.execute(stmt)).all()
-    df = pd.DataFrame(rows, columns=["symbol", "trade_date", "open", "high", "low", "close"])
+    df = pd.DataFrame(rows, columns=["symbol", "trade_date", "open", "high", "low", "close", "volume", "turnover"])
     if not df.empty and len(symbols) > 2000:
         df = df[df["symbol"].isin(set(symbols))]
     df = df.sort_values(["symbol", "trade_date"]).reset_index(drop=True)
     if df.empty:
         df["ts_code"] = pd.Series(dtype=object)
         df["pct_chg"] = pd.Series(dtype=object)
+        df["amplitude"] = pd.Series(dtype=object)
     else:
+        prev_close = df.groupby("symbol")["close"].shift(1)
         df["pct_chg"] = df.groupby("symbol")["close"].pct_change() * 100
+        df["amplitude"] = (df["high"] - df["low"]) / prev_close * 100
         ts_map = {plain_code(t): t for t in ts_codes}
         df["ts_code"] = df["symbol"].map(ts_map)
-    return df[["ts_code", "trade_date", "open", "high", "low", "close", "pct_chg"]]
+    return df[["ts_code", "trade_date", "open", "high", "low", "close", "volume", "turnover", "pct_chg", "amplitude"]]
 
 
 def _f(v) -> float | None:

backend/app/screener/llm.py

@@ -22,8 +22,8 @@ SYSTEM_PROMPT = """你是 A 股选股条件解析器。把用户的自然语言
 （indicator 与 snapshot 至少一个非空；用户没有提到的条件不要编造）
 
 【indicator 数组】技术指标条件，元素字段：
-- "indicator": 指标名，白名单：kdj_k / kdj_d / kdj_j（KDJ 的 K/D/J 值）、rsi、macd_dif / macd_dea / macd_hist（MACD 的 DIF/DEA/柱）、ma（收盘价均线）、boll_upper / boll_mid / boll_lower（布林轨道）、zhixing_dkx（知行多空线，四条收盘价均线的均值）、zhixing_trend（知行短期趋势线）、close（收盘价）、pct_chg（日涨跌幅%）
-- "params": 指标参数（可选），默认：KDJ {"n":9,"m1":3,"m2":3}；RSI {"period":14}；MACD {"fast":12,"slow":26,"signal":9}；MA {"period":20}；BOLL {"period":20,"std":2}；知行多空线 {"m1":14,"m2":28,"m3":57,"m4":114}；知行趋势线无参数（固定算法），不要给它填 params
+- "indicator": 指标名，白名单：kdj_k / kdj_d / kdj_j（KDJ 的 K/D/J 值）、rsi、macd_dif / macd_dea / macd_hist（MACD 的 DIF/DEA/柱）、ma（收盘价均线）、vol_ratio（量比，当日成交量/前5日均量）、boll_upper / boll_mid / boll_lower（布林轨道）、zhixing_dkx（知行多空线，四条收盘价均线的均值）、zhixing_trend（知行短期趋势线）、close（收盘价）、pct_chg（日涨跌幅%）、amplitude（当日振幅%，(最高-最低)/昨收）、turnover_rate（换手率%，逐日历史值）
+- "params": 指标参数（可选），默认：KDJ {"n":9,"m1":3,"m2":3}；RSI {"period":14}；MACD {"fast":12,"slow":26,"signal":9}；MA {"period":20}；量比 {"period":5}；BOLL {"period":20,"std":2}；知行多空线 {"m1":14,"m2":28,"m3":57,"m4":114}；知行趋势线/振幅/换手率无参数，不要给它们填 params
 - "op": "gt" | "ge" | "lt" | "le" | "between"
 - "value": 比较数值（between 时为下界），"value2": between 上界
 - "value_indicator": 可选。指标与指标比较时填另一指标名（同白名单），如 "DIF大于DEA" -> indicator=macd_dif, op=gt, value_indicator=macd_dea, value=0；"股价在布林带下轨之下" -> indicator=close, op=lt, value_indicator=boll_lower, value=0
@@ -171,8 +171,8 @@ EVENT_SYSTEM_PROMPT = """你是 A 股事件回测参数解析器。用户描述
 {"entry": {"indicator": [...], "snapshot": [], "exclude_st": true, "exclude_delisted": true, "exclude_bj": true}, "entry_timing": "next_open", "holding_days": 3, "exit_timing": "close"}
 
 【entry.indicator 数组】入场信号条件（必填，至少 1 条），元素字段与白名单：
-- "indicator": kdj_k / kdj_d / kdj_j（KDJ 的 K/D/J 值）、rsi、macd_dif / macd_dea / macd_hist（MACD 的 DIF/DEA/柱）、ma（收盘价均线）、boll_upper / boll_mid / boll_lower（布林轨道）、zhixing_dkx（知行多空线）、zhixing_trend（知行短期趋势线）、close（收盘价）、pct_chg（日涨跌幅%）
-- "params": 指标参数（可选），默认：KDJ {"n":9,"m1":3,"m2":3}；RSI {"period":14}；MACD {"fast":12,"slow":26,"signal":9}；MA {"period":20}；BOLL {"period":20,"std":2}；知行多空线 {"m1":14,"m2":28,"m3":57,"m4":114}；知行趋势线无参数（固定算法），不要给它填 params
+- "indicator": kdj_k / kdj_d / kdj_j（KDJ 的 K/D/J 值）、rsi、macd_dif / macd_dea / macd_hist（MACD 的 DIF/DEA/柱）、ma（收盘价均线）、vol_ratio（量比，当日成交量/前5日均量）、boll_upper / boll_mid / boll_lower（布林轨道）、zhixing_dkx（知行多空线）、zhixing_trend（知行短期趋势线）、close（收盘价）、pct_chg（日涨跌幅%）、amplitude（当日振幅%，(最高-最低)/昨收）、turnover_rate（换手率%，逐日历史值）
+- "params": 指标参数（可选），默认：KDJ {"n":9,"m1":3,"m2":3}；RSI {"period":14}；MACD {"fast":12,"slow":26,"signal":9}；MA {"period":20}；量比 {"period":5}；BOLL {"period":20,"std":2}；知行多空线 {"m1":14,"m2":28,"m3":57,"m4":114}；知行趋势线/振幅/换手率无参数，不要给它们填 params
 - "op": "gt" | "ge" | "lt" | "le" | "between"；"value"（between 时为下界）、"value2"（上界）
 - "value_indicator": 指标与指标比较时填另一指标名（同白名单），如 "DIF 大于 DEA" -> indicator=macd_dif, op=gt, value_indicator=macd_dea, value=0
 - "value_params": 比较对象指标参数不同时指定，如 "MA5 上穿 MA20" -> indicator=ma, params={"period":5}, op=gt, value_indicator=ma, value_params={"period":20}, value=0