@weaxs/openclaw-stock-analysis 0.2.5 → 0.2.6

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -0,0 +1,132 @@
1
+ #!/usr/bin/env python3
2
+ """Skill-local tail diagnostics from historical returns or backtest JSON."""
3
+
4
+ import argparse
5
+ import json
6
+ import math
7
+ import sys
8
+ from pathlib import Path
9
+
10
+ import numpy as np
11
+
12
+
13
+ class Parser(argparse.ArgumentParser):
14
+ def error(self, message):
15
+ raise ValueError(message)
16
+
17
+
18
+ def finite_number(value):
19
+ if isinstance(value, bool) or not isinstance(value, (int, float)) or not math.isfinite(value):
20
+ raise ValueError("Input values must be finite JSON numbers")
21
+ return float(value)
22
+
23
+
24
+ def analyze_returns(data, args):
25
+ if not isinstance(data, list) or len(data) < 2:
26
+ raise ValueError("Expected a JSON array of at least two historical returns")
27
+ values = np.array([finite_number(value) for value in data])
28
+ if not 100 <= args.draws <= 1000000 or args.seed < 0 or any(n < 2 or n > 10000 for n in args.n):
29
+ raise ValueError("Require draws in 100..1000000, seed >= 0, and n in 2..10000")
30
+ centered = values - values.mean()
31
+ mad = float(np.abs(centered).mean())
32
+ rng = np.random.default_rng(args.seed)
33
+ rows = []
34
+ for n in args.n:
35
+ total = 0.0
36
+ if mad > 0:
37
+ # Bound temporary arrays when either summands or draws is large.
38
+ batch_size = max(1, 1000000 // n)
39
+ for start in range(0, args.draws, batch_size):
40
+ batch = min(batch_size, args.draws - start)
41
+ sums = rng.choice(centered, size=(batch, n)).sum(axis=1)
42
+ total += float(np.abs(sums).sum())
43
+ summed_mad = total / args.draws
44
+ denominator = math.log(summed_mad / mad) if mad > 0 and summed_mad > 0 else 0
45
+ rows.append(
46
+ {
47
+ "n0": 1,
48
+ "n": n,
49
+ "mean_absolute_deviation_sum": summed_mad,
50
+ "kappa": 2 - math.log(n) / denominator if denominator else None,
51
+ "unavailable_reason": None if denominator else "Zero dispersion or degenerate aggregation",
52
+ }
53
+ )
54
+ event = None
55
+ if args.current is not None:
56
+ current = finite_number(args.current)
57
+ event = {
58
+ "return": current,
59
+ "percentile": float(np.mean(values <= current)),
60
+ "upper_exceedances": int(np.sum(values > current)),
61
+ "lower_exceedances": int(np.sum(values < current)),
62
+ }
63
+ return {
64
+ "sample_count": len(values),
65
+ "draws": args.draws,
66
+ "seed": args.seed,
67
+ "mean_absolute_deviation": mad,
68
+ "kappa": rows,
69
+ "current_event": event,
70
+ "method": "iid resampling of centered empirical returns; not a population tail estimate or trading signal",
71
+ }
72
+
73
+
74
+ def analyze_trades(data):
75
+ if not isinstance(data, dict) or not isinstance(data.get("trades"), list):
76
+ raise ValueError("Expected a backtest JSON object with a trades array")
77
+ if any(not isinstance(trade, dict) or trade.get("type") not in {"buy", "sell"} for trade in data["trades"]):
78
+ raise ValueError("Each trade must have type buy or sell")
79
+ pnl = [finite_number(trade.get("pnl")) for trade in data["trades"] if trade["type"] == "sell"]
80
+ winners = sorted((value for value in pnl if value > 0), reverse=True)
81
+ gross_profit = math.fsum(winners)
82
+ gross_loss = math.fsum(value for value in pnl if value < 0)
83
+ net = math.fsum(pnl)
84
+ return {
85
+ "closed_trades": len(pnl),
86
+ "winning_trades": len(winners),
87
+ "gross_profit": gross_profit,
88
+ "gross_loss": gross_loss,
89
+ "realized_net_pnl": net,
90
+ "worst_trade_pnl": min(pnl) if pnl else None,
91
+ "winner_contribution": [
92
+ {
93
+ "requested_k": k,
94
+ "actual_k": min(k, len(winners)),
95
+ "share_of_gross_profit": math.fsum(winners[:k]) / gross_profit if gross_profit else None,
96
+ "realized_pnl_without_top_k": net - math.fsum(winners[:k]),
97
+ }
98
+ for k in (1, 3)
99
+ ],
100
+ "method": "closed monetary pnl only; deletion sensitivity is not an equity-curve rerun",
101
+ }
102
+
103
+
104
+ def main():
105
+ try:
106
+ parser = Parser(description=__doc__)
107
+ sub = parser.add_subparsers(dest="command", required=True, parser_class=Parser)
108
+ returns = sub.add_parser("returns", help="Analyze a JSON array of historical log returns")
109
+ returns.add_argument("input", type=Path)
110
+ returns.add_argument("--n", nargs="+", type=int, default=[2, 10, 30])
111
+ returns.add_argument("--draws", type=int, default=20000)
112
+ returns.add_argument("--seed", type=int, default=42)
113
+ returns.add_argument("--current", type=float, help="Separate current return, excluded from historical input")
114
+ trades = sub.add_parser("trades", help="Analyze a complete run_backtest JSON result")
115
+ trades.add_argument("input", type=Path)
116
+ args = parser.parse_args()
117
+ data = json.loads(args.input.read_text(encoding="utf-8"))
118
+ result = analyze_returns(data, args) if args.command == "returns" else analyze_trades(data)
119
+ print(json.dumps(result, ensure_ascii=True, allow_nan=False))
120
+ return 0
121
+ except OSError:
122
+ error = "Cannot read input JSON file"
123
+ except (ValueError, OverflowError) as exc:
124
+ error = "Invalid JSON input" if isinstance(exc, json.JSONDecodeError) else str(exc)
125
+ except Exception:
126
+ error = "Tail analysis failed unexpectedly"
127
+ print(json.dumps({"error": error}, ensure_ascii=True))
128
+ return 1
129
+
130
+
131
+ if __name__ == "__main__":
132
+ sys.exit(main())
@@ -22,6 +22,8 @@ gather 工具返回:
22
22
  - `stock_info` — 公司基本信息
23
23
  - `sector_rankings` — 板块排名
24
24
 
25
+ A 股事件原文先调用 `get_announcements`:业绩预告用 `forecast`、快报用 `express`、减持用 `reduction`、增发用 `issuance`、问询及回复用 `inquiry`、分红用 `dividend`,指定 `start_date/end_date`。返回 PDF 链接而非正文;读取原文后再量化,`truncated=true` 时缩小日期范围。
26
+
25
27
  新闻摘要只用于发现事件。金额、期限、合同对手、业绩变化或政策原文必须由公告/原文确认;只有摘要时不做精确量化。
26
28
 
27
29
  制定短线计划时读取[短线交易决策框架](../stock-analysis/references/short-term-trading.md),用状态机、相对强弱和风险单位约束事件交易。
@@ -22,6 +22,8 @@ gather 工具返回:
22
22
  - `stock_info` — 公司基本信息(行业、板块)
23
23
  - `sector_rankings` — 板块排名
24
24
 
25
+ 需要财务趋势时调用 `get_financials(symbol, annual=true, periods=10)`。港股配置 `LIXINGER_TOKEN` 后优先理杏仁非金融公司财报,按真实 `report_type` 判断年报,保留实际财年末、披露日期、币种;历史 ROE/毛利率/净利率为百分数,顶层 `roe/profit_margin` 延续小数比例。读取 `available_periods/warnings`,缺期不补造。有效 Yahoo 顶层快照与历史行分开读:`currency` 为快照财报币种、`market_cap_currency` 为市值币种,历史币种在 `periods` 每行,`snapshot_source` 标记快照来源;null 不用历史值补成跨期或跨币种数字;顶层 `report_date` 未知时读取 `periods` 的历史日期,不当作当前快照报告期。
26
+
25
27
  `financials` 通常只是单期摘要。多期趋势、经营现金流、应收账款或同业估值未返回时,把对应维度标为“数据不足”,不推断趋势、不计分。
26
28
 
27
29
  ## 分析框架(4 步法)
@@ -5,6 +5,10 @@ description: 综合股票分析:联合技术、基本面、资金、消息和
5
5
 
6
6
  # 综合股票分析
7
7
 
8
+ 美股机构研究调用 `get_us_holdings(cik, mode=auto)`:有当前或已检索历史 13F 时返回机构组合,否则按发行人 CIK 反查机构;也可显式选 `manager/issuer`,如 Berkshire `0001067983` 查组合、Apple `0000320193` 用 `issuer` 查机构。发行人映射仅使用核对 CIK 的官方 13D/G CUSIP,不做名称猜测;无映射直接报缺失。首次下载约 100 MB,后续缓存,Hermes 查询预算 300 秒。读取 `report_date/filing_window_end/cusip_evidence`,区分持仓日期与披露窗口;修订重述替换、新增补充,缺基础版或未知修订时总量 null,并计入 `unresolved_manager_count`。`mapping_complete=false` 表示不能保证涵盖所有类别或变更,`complete_for_mapped_cusips` 只表示已映射证券在公开数据窗口的申报链完整性,不能称实时全部股东。期权/本金在 `positions` 单列,不能把跨机构持股简单合计成去重比例。机构组合查询仍返回最新已知报告期的原始 13F 与修订链接,缺少对应原始版时明确报错;`history_truncated=true` 时不能将旧记录当作已完整检索。存在修订先读原文,不能自动推断变化。
9
+
10
+ 美股内部人研究调用 `get_insider_trades(symbol, days=90, top=10)`,需设置含真实联系邮箱的 `SEC_USER_AGENT`。`days` 按申报日期过滤,`top` 限制文件数;只将代码 P/S 视作买卖信号,授予、行权、赠与等保留原始代码和脚注,衍生品单列。Form 4/A 与原申报勿重复计算,检查截断和失败提示;`screen_risk` 仍仅支持 A 股。
11
+
8
12
  ## 意图路由表
9
13
 
10
14
  快速问答先判断意图,按下表走最短调用链;完整综合研判走下方执行流程(gather 采集管线),不受此表限制。单工具可答的问题按各工具描述中的适用场景选择,本表不再重复。
@@ -22,16 +26,18 @@ description: 综合股票分析:联合技术、基本面、资金、消息和
22
26
 
23
27
  ### 第一步:采集数据
24
28
 
25
- 运行数据采集脚本(并行获取 quote/kline/technical/financials/capital_flow/news/risk/regime 共 8 项数据):
29
+ 运行数据采集脚本(并行采集行情、K 线、财务、资金、筹码、新闻、舆情、风险和大盘状态;技术指标复用本次 K 线计算):
26
30
 
27
31
  ```bash
28
32
  python3 scripts/gather.py <symbol>
29
33
  ```
30
34
 
31
- 解析返回的 JSON 数据。如果某个字段为 `null`,表示该数据获取失败,基于可用数据继续分析。
35
+ 解析返回的 JSON 数据。如果某个字段为 `null` 或包含 `error`,表示该数据不可用,基于可用数据继续分析。大盘状态按股票所属市场查询;日/韩/台市场暂不支持,标明缺失,不能套用 A 股状态。
32
36
 
33
37
  解读实时行情前先调 `get_trading_phase` 确认该市场当前盘中阶段(盘前/早盘/午间休市/午盘/交易中/盘后/休市),避免把休市时的静态快照当作盘中信号解读。
34
38
 
39
+ 港股 `get_quote` 自动按 yfinance → 腾讯 → finnhub → longbridge 容灾,腾讯无需 API key,使用 `r_hk` 实时报价通道(`hk` 通道延迟约 15 分钟)。腾讯港股成交量为股,成交额及市值为港元;PB、换手率、量比为 null 时按缺失处理。该备用源只覆盖报价,K 线仍走原链路。
40
+
35
41
  ### 第二步:综合研判
36
42
 
37
43
  根据获取的数据进行多维度分析:
@@ -60,7 +66,7 @@ python3 scripts/gather.py <symbol>
60
66
 
61
67
  **风险筛查**:
62
68
  - 7维度风险标记(估值极端/技术预警/解禁到期/内部人减持/业绩预警/监管处罚/行业政策)
63
- - 风险评级(low/medium/high)和一票否决(veto_buy)
69
+ - 风险评级(low/medium/high/unknown)和一票否决(veto_buy);查看 data_complete、warnings,unknown 或 null 表示缺少数据,不能据此判定低风险或已通过检查
64
70
  - 如果 veto_buy=true,必须在报告中醒目提示
65
71
 
66
72
  ### 第三步:渲染研报
@@ -5,6 +5,8 @@ description: 全市场股票筛选:按市场、估值、成长、趋势或自
5
5
 
6
6
  # 全市场股票筛选
7
7
 
8
+ 显式筛选条件缺少数据时工具返回 `error`,不能把该条件当作已通过。`filters_applied` 只包含实际执行的条件,基于返回的覆盖情况说明候选范围。
9
+
8
10
  ## 执行流程
9
11
 
10
12
  ### 第一步:确定筛选条件
@@ -7,6 +7,8 @@ description: 策略回测:定义或读取 AlphaEvo YAML 策略,回测收益
7
7
 
8
8
  开始前读取[短线交易决策框架](../stock-analysis/references/short-term-trading.md)的风险单位和回测约束。回测用于反证策略,不把样本内收益包装成预测。
9
9
 
10
+ 解读回测时,`period` 是实际取得的数据期间,`requested_period` 保留用户请求,`warnings` 提示两者差异。年化收益按实际期间计算;覆盖不完整时明确说明,不能声称完成了完整请求期间的验证。买入预算包含佣金,`position.size` 为 0..1 的资金比例。
11
+
10
12
  ## 工作流程
11
13
 
12
14
  ### 第一步:策略定义
@@ -77,6 +79,8 @@ python3 tools/backtest.py run <path.yaml> <symbol> --start <date> --end <date>
77
79
  - 使用用户指定的互斥日期区间分别运行样本内和样本外回测。
78
80
  - 成本假设引用结果中的 `costs` 字段(实际生效的佣金/滑点/印花税费率与实付费用合计),未模拟的费用与成交约束列入限制。
79
81
 
82
+ 趋势策略或用户要求肥尾收益诊断时,读取 [肥尾度量与趋势收益诊断](../bull-trend/references/fat-tails.md)的盈利集中度部分,基于平仓 `pnl` 计算大赢家贡献和删除敏感性。只有用户询问市场收益分布时才进入 κ 分支。
83
+
80
84
  未模拟的结算、涨跌停、复权或特定市场费用列入限制,结论限定为该模拟条件下的历史表现。
81
85
 
82
86
  ### 第四步:诊断
@@ -7,6 +7,8 @@ description: 智堡投研检索:查找机构研报、财报公告、电话会
7
7
 
8
8
  智堡 MCP 涵盖海外投行研报、国内券商研究、资管报告、央行/政府文献、电话会纪要和财报公告。首先检查这些 MCP 工具是否可用;不可用时直接改用 `search_comprehensive_intel`。
9
9
 
10
+ 港股估值功课调用 `get_valuation_history(symbol, years=10)`;配置 `LIXINGER_TOKEN` 时优先理杏仁非金融公司每日 PE-TTM/PB,失败回退百度。以返回的实际窗口和正值样本计算分位及 Band,不混接同一指标的不同来源,不将缺失解释为低估。
11
+
10
12
  ## 可用工具
11
13
 
12
14
  智堡 MCP 提供以下工具(通过 `mcp__wisburg-mcp-server__` 前缀调用):
@@ -7,11 +7,14 @@
7
7
  {% if risk_screening and risk_screening.veto_buy %}
8
8
  ⚠️ **一票否决** — 风险筛查触发,建议谨慎。
9
9
  {% endif %}
10
+ {% if risk_screening and (risk_screening.data_complete is sameas false or risk_screening.risk_level == 'unknown') %}
11
+ ⚠️ **数据不完整** — 当前风险筛查不能判定为低风险。
12
+ {% endif %}
10
13
 
11
14
  {% set dp = data_perspective or {} -%}
12
15
  {% set pp = dp.price_position or {} -%}
13
16
  {% set ts = dp.trend_status or {} -%}
14
- - 📈 **趋势:** {{ ts.ma_alignment or '-' }}{% if ts.is_bullish is not none %}({{ '看多' if ts.is_bullish else '看空' }}){% endif %}
17
+ - 📈 **趋势:** {{ ts.ma_alignment or '-' }}{% if ts.is_bullish is boolean %}({{ '看多' if ts.is_bullish else '看空' }}){% endif %}
15
18
  - 💰 **价位:** 现价 {{ m.fmt_num(pp.current_price) }} | 支撑 {{ m.fmt_num(pp.support_level) }} | 压力 {{ m.fmt_num(pp.resistance_level) }}
16
19
 
17
20
  {% set bp = battle_plan or {} -%}
@@ -26,12 +26,15 @@
26
26
  ## 风险筛查
27
27
 
28
28
  {% if risk_screening -%}
29
- - **风险等级:** {{ m.risk_emoji(risk_screening.risk_level) }} {{ risk_screening.risk_level or '-' }} | **风险分:** {{ risk_screening.risk_score }}/100
29
+ - **风险等级:** {{ m.risk_emoji(risk_screening.risk_level) }} {{ risk_screening.risk_level or '-' }} | **风险分:** {{ m.fmt_num(risk_screening.risk_score, 0) }}/100
30
+ {% if risk_screening.data_complete is sameas false or risk_screening.risk_level == 'unknown' -%}
31
+ - **数据不完整:** {{ risk_screening.warnings | join(';') }},当前结果不能判定为低风险。
32
+ {% endif %}
30
33
  {% if risk_screening.flags -%}
31
34
  - **风险标记:**
32
35
  {% for f in risk_screening.flags %} - [{{ f.severity }}] {{ f.category }}:{{ f.description }}
33
36
  {% endfor %}
34
- {%- else -%}
37
+ {%- elif risk_screening.data_complete is not sameas false and risk_screening.risk_level != 'unknown' -%}
35
38
  - **风险标记:** 无显著风险标记
36
39
  {%- endif %}
37
40
  {%- else -%}
@@ -45,7 +48,7 @@
45
48
  {% set pp = dp.price_position or {} -%}
46
49
  {% set vol = dp.volume_analysis or {} -%}
47
50
  {% set chip = dp.chip_structure or {} -%}
48
- - **趋势:** {{ ts.ma_alignment or '-' }}{% if ts.is_bullish is not none %}({{ '看多' if ts.is_bullish else '看空' }}){% endif %}{% if ts.trend_score is not none %} 趋势分 {{ m.fmt_num(ts.trend_score) }}{% endif %}
51
+ - **趋势:** {{ ts.ma_alignment or '-' }}{% if ts.is_bullish is boolean %}({{ '看多' if ts.is_bullish else '看空' }}){% endif %}{% if ts.trend_score is not none %} 趋势分 {{ m.fmt_num(ts.trend_score) }}{% endif %}
49
52
  - **价格结构:** 现价 {{ m.fmt_num(pp.current_price) }} | 偏离 MA5 {{ m.change_pct(pp.bias_ma5) }} | 支撑 {{ m.fmt_num(pp.support_level) }} | 压力 {{ m.fmt_num(pp.resistance_level) }}
50
53
  - **量价:** 量比 {{ m.fmt_num(vol.volume_ratio) }}{% if vol.volume_status %}({{ vol.volume_status }}){% endif %}{% if vol.turnover_rate is not none %} | 换手率 {{ m.change_pct(vol.turnover_rate) }}{% endif %}
51
54
  {% if chip and (chip.profit_ratio is not none or chip.avg_cost is not none) -%}
@@ -6,6 +6,7 @@ Does NOT store history, does NOT push, does NOT schedule. Host agent decides whe
6
6
 
7
7
  import argparse
8
8
  import json
9
+ import math
9
10
  import sys
10
11
  from concurrent.futures import ThreadPoolExecutor
11
12
  from datetime import datetime, timezone
@@ -31,8 +32,14 @@ RULE_TYPES = {
31
32
  def _check_rule(rule: dict, quote: dict, tech: dict, anom: dict, risk: dict) -> dict | None:
32
33
  rtype = rule.get("type")
33
34
  value = rule.get("value")
34
- if rtype not in RULE_TYPES:
35
+ if not isinstance(rtype, str) or rtype not in RULE_TYPES:
35
36
  return {"rule": rtype, "triggered": False, "error": f"unknown rule type: {rtype}"}
37
+ if rtype in ("price_below", "price_above", "change_pct_above", "change_pct_below", "volume_ratio_above"):
38
+ try:
39
+ if isinstance(value, bool) or not math.isfinite(float(value)):
40
+ raise ValueError
41
+ except (ValueError, TypeError):
42
+ return {"rule": rtype, "triggered": False, "error": "Rule value must be a finite number"}
36
43
 
37
44
  price = quote.get("price")
38
45
  change_pct = quote.get("change_pct")
@@ -90,6 +97,8 @@ def _check_rule(rule: dict, quote: dict, tech: dict, anom: dict, risk: dict) ->
90
97
  "message": f"量比 {ratio} 超过 {value}",
91
98
  }
92
99
  elif rtype == "anomaly":
100
+ if not isinstance(value, str) or not value:
101
+ return {"rule": rtype, "triggered": False, "error": "Anomaly type must be a nonempty string"}
93
102
  # value is an anomaly type string, e.g. "macd_golden_cross"
94
103
  anomalies = (anom or {}).get("anomalies", [])
95
104
  for a in anomalies:
@@ -106,6 +115,8 @@ def _check_rule(rule: dict, quote: dict, tech: dict, anom: dict, risk: dict) ->
106
115
  return {"rule": rtype, "triggered": True, "severity": "high", "message": "风险筛查触发一票否决"}
107
116
  elif rtype == "risk_level_at_least":
108
117
  order = {"low": 0, "medium": 1, "high": 2}
118
+ if not isinstance(value, str) or value not in order:
119
+ return {"rule": rtype, "triggered": False, "error": "Risk threshold must be low/medium/high"}
109
120
  cur = order.get((risk or {}).get("risk_level"))
110
121
  want = order.get(value)
111
122
  if cur is not None and want is not None and cur >= want:
@@ -122,7 +133,7 @@ def _check_rule(rule: dict, quote: dict, tech: dict, anom: dict, risk: dict) ->
122
133
 
123
134
 
124
135
  def check_rules(symbol: str, rules: list[dict]) -> dict:
125
- if not rules:
136
+ if not isinstance(rules, list) or not rules or any(not isinstance(rule, dict) for rule in rules):
126
137
  return {"symbol": symbol, "error": "no rules provided"}
127
138
 
128
139
  needs_tech = any(r.get("type") == "volume_ratio_above" for r in rules)
@@ -141,6 +152,9 @@ def check_rules(symbol: str, rules: list[dict]) -> dict:
141
152
  with ThreadPoolExecutor(max_workers=len(jobs)) as pool:
142
153
  futures = {key: pool.submit(_run_json, script, args) for key, (script, args) in jobs.items()}
143
154
  fetched = {key: future.result() for key, future in futures.items()}
155
+ fetched = {
156
+ key: value if isinstance(value, dict) and "error" not in value else None for key, value in fetched.items()
157
+ }
144
158
  quote = fetched["quote"] or {}
145
159
  tech = fetched.get("tech")
146
160
  anom = fetched.get("anom")
@@ -153,13 +167,25 @@ def check_rules(symbol: str, rules: list[dict]) -> dict:
153
167
  warnings.append("technical analysis fetch failed — volume_ratio rules cannot fire")
154
168
  if needs_anom and not anom:
155
169
  warnings.append("anomaly fetch failed — anomaly rules cannot fire")
156
- if needs_risk and not risk:
170
+ if needs_risk and (not risk or risk.get("data_complete") is False or risk.get("risk_level") == "unknown"):
157
171
  warnings.append("risk screening fetch failed — risk rules cannot fire")
158
172
 
159
173
  hits = []
160
174
  evaluated = []
161
175
  for rule in rules:
162
- result = _check_rule(rule, quote, tech, anom, risk)
176
+ rtype = rule.get("type")
177
+ missing = (
178
+ (rtype in ("price_below", "price_above") and quote.get("price") is None)
179
+ or (rtype in ("change_pct_below", "change_pct_above") and quote.get("change_pct") is None)
180
+ or (rtype == "volume_ratio_above" and (tech or {}).get("volume", {}).get("volume_ratio") is None)
181
+ or (rtype == "anomaly" and not isinstance((anom or {}).get("anomalies"), list))
182
+ or (rtype == "risk_veto" and (risk or {}).get("veto_buy") is None)
183
+ or (rtype == "risk_level_at_least" and (risk or {}).get("risk_level") not in ("low", "medium", "high"))
184
+ )
185
+ if missing:
186
+ result = {"rule": rtype, "triggered": False, "error": "Required source data unavailable"}
187
+ else:
188
+ result = _check_rule(rule, quote, tech, anom, risk)
163
189
  if result is None:
164
190
  evaluated.append({"rule": rule.get("type"), "triggered": False})
165
191
  else:
@@ -214,6 +240,8 @@ def main():
214
240
  result = check_rules(args.symbol, rules)
215
241
  json.dump(result, sys.stdout, ensure_ascii=False, indent=2, default=str)
216
242
  print()
243
+ if "error" in result:
244
+ sys.exit(1)
217
245
 
218
246
 
219
247
  if __name__ == "__main__":
package/tools/backtest.py CHANGED
@@ -16,6 +16,7 @@ import yaml
16
16
  sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
17
17
 
18
18
  from _subproc import json_safe, run_tool, utf8_stdio
19
+ from stock_data import detect_market
19
20
  from technical import bollinger_series, ema_series, ma_series, macd_series, rsi_series
20
21
 
21
22
  # --------------- Strategy Loading ---------------
@@ -67,7 +68,7 @@ def fetch_kline(symbol: str, start: str | None, end: str | None) -> pd.DataFrame
67
68
  if start and end:
68
69
  d0 = datetime.strptime(start, "%Y-%m-%d")
69
70
  d1 = datetime.strptime(end, "%Y-%m-%d")
70
- count = max((d1 - d0).days + 60, 300)
71
+ count = max((datetime.now() - d0).days + 60, (d1 - d0).days + 60, 300)
71
72
 
72
73
  # run_tool gives venv-python + a bounded timeout + None on any failure
73
74
  # (non-zero exit, empty/invalid stdout); callers turn that into clean JSON.
@@ -251,7 +252,11 @@ def evaluate_conditions(df: pd.DataFrame, conditions: list, logic: str, idx: int
251
252
 
252
253
 
253
254
  def simulate(df: pd.DataFrame, strategy: dict, capital: float, symbol: str) -> dict:
254
- market = "A" if re.match(r"^\d{6}$", symbol) else "OTHER"
255
+ if not math.isfinite(capital) or capital <= 0:
256
+ raise ValueError("capital must be finite and positive")
257
+ if not np.isfinite(df["close"]).all() or (df["close"] <= 0).any():
258
+ raise ValueError("backtest prices must be finite and positive")
259
+ market = detect_market(symbol)
255
260
  lot_size = 100 if market == "A" else 1
256
261
  sim_cfg = strategy.get("simulation", {})
257
262
  slippage = float(sim_cfg.get("slippage", 0.001))
@@ -268,6 +273,8 @@ def simulate(df: pd.DataFrame, strategy: dict, capital: float, symbol: str) -> d
268
273
  exit_ = strategy.get("exit", {})
269
274
  position_cfg = strategy.get("position", {})
270
275
  size_frac = float(position_cfg.get("size", 1.0))
276
+ if not 0 <= size_frac <= 1:
277
+ raise ValueError("position.size must be a fraction in [0, 1]")
271
278
  stop_loss = float(exit_.get("stop_loss", -0.10))
272
279
  take_profit = float(exit_.get("take_profit", 0.50))
273
280
  ts_val = exit_.get("trailing_stop")
@@ -297,7 +304,7 @@ def simulate(df: pd.DataFrame, strategy: dict, capital: float, symbol: str) -> d
297
304
  if hit:
298
305
  buy_price = price * (1 + slippage)
299
306
  buy_amount = cash * size_frac
300
- shares = int(buy_amount / buy_price / lot_size) * lot_size
307
+ shares = int(buy_amount / (buy_price * (1 + commission)) / lot_size) * lot_size
301
308
  if shares <= 0:
302
309
  continue
303
310
  cost = shares * buy_price * (1 + commission)
@@ -417,7 +424,7 @@ def compute_metrics(trades: list, equity_curve: list, initial: float, final: flo
417
424
  years = 1
418
425
  annual_return = (1 + total_return) ** (1 / max(years, 0.01)) - 1 if total_return > -1 else total_return
419
426
 
420
- equities = pd.Series([e["equity"] for e in equity_curve], dtype=float)
427
+ equities = pd.Series([initial, *[e["equity"] for e in equity_curve]], dtype=float)
421
428
  peak = equities.cummax()
422
429
  dd = (equities - peak) / peak
423
430
  max_drawdown = float(dd.min()) if len(dd) > 0 else 0
@@ -653,6 +660,13 @@ def run_backtest(strategy_path: str, symbol: str, start: str | None, end: str |
653
660
  if len(df) < 30:
654
661
  return {"error": f"Insufficient data: {len(df)} rows (need at least 30)"}
655
662
 
663
+ requested_period = f"{start} to {end}"
664
+ actual_start = df["date"].iloc[0].strftime("%Y-%m-%d")
665
+ actual_end = df["date"].iloc[-1].strftime("%Y-%m-%d")
666
+ warnings = []
667
+ if actual_start != start or actual_end != end:
668
+ warnings.append(f"Actual data covers {actual_start} to {actual_end}; requested {requested_period}")
669
+ start, end = actual_start, actual_end
656
670
  sim = simulate(df, strategy, capital, symbol)
657
671
  final = sim["final_equity"]
658
672
  metrics = compute_metrics(sim["trades"], sim["equity_curve"], capital, final, start, end)
@@ -661,6 +675,8 @@ def run_backtest(strategy_path: str, symbol: str, start: str | None, end: str |
661
675
  "strategy": strategy.get("name", "unnamed"),
662
676
  "symbol": symbol,
663
677
  "period": f"{start} to {end}",
678
+ "requested_period": requested_period,
679
+ "warnings": warnings,
664
680
  "initial_capital": capital,
665
681
  "final_capital": final,
666
682
  "costs": sim["costs"],
@@ -837,6 +853,8 @@ def main():
837
853
  sys.exit(1)
838
854
 
839
855
  print(json.dumps(json_safe(result), ensure_ascii=False, default=str))
856
+ if "error" in result:
857
+ sys.exit(1)
840
858
 
841
859
 
842
860
  if __name__ == "__main__":
@@ -18,6 +18,9 @@ ALL_MARKETS = ["A", "HK", "US", "JP", "KR", "TW"]
18
18
 
19
19
  # tool_name -> {markets: [supported markets], reason: str for unsupported}
20
20
  TOOL_MATRIX = {
21
+ "get_announcements": {"markets": ["A"], "reason": "cninfo A-share disclosures only"},
22
+ "get_us_holdings": {"markets": ["US"], "reason": "SEC manager holdings or verified issuer-CUSIP reverse lookup"},
23
+ "get_insider_trades": {"markets": ["US"], "reason": "SEC issuer Form 4 only"},
21
24
  "get_valuation_history": {"markets": ["A", "HK"], "reason": "A/HK historical valuation providers only"},
22
25
  "get_index_valuation": {"markets": ["A"], "reason": "Chinese index valuation only"},
23
26
  "get_earnings_calendar": {
package/tools/gather.py CHANGED
@@ -10,6 +10,7 @@ from pathlib import Path
10
10
  sys.path.insert(0, str(Path(__file__).resolve().parent))
11
11
  from _subproc import json_safe, run_tool, utf8_stdio # noqa: E402
12
12
  from stock_data import detect_market # noqa: E402
13
+ from technical import analyze_records # noqa: E402
13
14
 
14
15
 
15
16
  # gather fans out the heavier CLIs, so the default timeout stays 60s; raw stdout
@@ -32,36 +33,41 @@ def _parse_json(raw: str | None):
32
33
  return raw
33
34
 
34
35
 
35
- def _run_all(tasks: dict) -> dict:
36
+ def _run_all(tasks: dict, symbol: str) -> dict:
36
37
  with ThreadPoolExecutor(max_workers=len(tasks)) as pool:
37
38
  futures = {pool.submit(_run, script, args): key for key, (script, args) in tasks.items()}
38
- return {key: _parse_json(f.result()) for f, key in futures.items()}
39
+ result = {key: _parse_json(f.result()) for f, key in futures.items()}
40
+ result["technical"] = analyze_records(symbol, result["kline"]) if result["kline"] is not None else None
41
+ return result
39
42
 
40
43
 
41
44
  def gather_analysis(symbol: str) -> dict:
45
+ market = detect_market(symbol)
42
46
  tasks = {
43
47
  "quote": ("stock_data.py", ["quote", symbol]),
44
48
  "kline": ("stock_data.py", ["kline", symbol, "--period", "daily", "--count", "120"]),
45
- "technical": ("technical.py", ["analyze", symbol, "--period", "daily", "--count", "120"]),
46
49
  "financials": ("stock_data.py", ["financials", symbol]),
47
50
  "capital_flow": ("stock_data.py", ["capital_flow", symbol]),
48
51
  "chip_distribution": ("stock_data.py", ["chip_distribution", symbol]),
49
52
  "news": ("search_intel.py", ["search", _news_query(symbol)]),
50
53
  "social_sentiment": ("search_intel.py", ["sentiment", symbol]),
51
54
  "risk": ("risk_screening.py", ["screen", symbol]),
52
- "regime": ("market_regime.py", ["detect"]),
53
55
  }
54
- return _run_all(tasks)
56
+ if market in ("A", "HK", "US"):
57
+ tasks["regime"] = ("market_regime.py", ["detect", market])
58
+ result = _run_all(tasks, symbol)
59
+ if market not in ("A", "HK", "US"):
60
+ result["regime"] = {"error": f"Market regime unavailable for {market}"}
61
+ return result
55
62
 
56
63
 
57
64
  def gather_technical(symbol: str, kline_count: int = 120, with_quote: bool = False) -> dict:
58
65
  tasks = {
59
66
  "kline": ("stock_data.py", ["kline", symbol, "--period", "daily", "--count", str(kline_count)]),
60
- "technical": ("technical.py", ["analyze", symbol, "--period", "daily", "--count", str(kline_count)]),
61
67
  }
62
68
  if with_quote:
63
69
  tasks["quote"] = ("stock_data.py", ["quote", symbol])
64
- return _run_all(tasks)
70
+ return _run_all(tasks, symbol)
65
71
 
66
72
 
67
73
  def gather_screen(market: str = "A", top: int = 20, config: str | None = None) -> dict:
@@ -73,16 +79,20 @@ def gather_screen(market: str = "A", top: int = 20, config: str | None = None) -
73
79
 
74
80
 
75
81
  def gather_fundamental(symbol: str) -> dict:
82
+ market = detect_market(symbol)
76
83
  tasks = {
77
84
  "quote": ("stock_data.py", ["quote", symbol]),
78
85
  "kline": ("stock_data.py", ["kline", symbol, "--period", "daily", "--count", "60"]),
79
- "technical": ("technical.py", ["analyze", symbol, "--period", "daily", "--count", "60"]),
80
86
  "financials": ("stock_data.py", ["financials", symbol]),
81
87
  "news": ("search_intel.py", ["search", _news_query(symbol)]),
82
88
  "stock_info": ("stock_data.py", ["stock_info", symbol]),
83
- "sector_rankings": ("stock_data.py", ["sector_rankings", "--top", "5", "--direction", "both"]),
84
89
  }
85
- return _run_all(tasks)
90
+ if market == "A":
91
+ tasks["sector_rankings"] = ("stock_data.py", ["sector_rankings", "--top", "5", "--direction", "both"])
92
+ result = _run_all(tasks, symbol)
93
+ if market != "A":
94
+ result["sector_rankings"] = {"error": f"Sector rankings unavailable for {market}"}
95
+ return result
86
96
 
87
97
 
88
98
  def main():