Stock Market Analysis

BASE=http://127.0.0.1:8000
TOKEN=$(curl -s -X POST $BASE/api/v1/auth/login -H "Content-Type: application/json" \
  -d '{"username":"admin","password":"admin123","tenant_id":"default"}' \
  | python3 -c "import sys,json; print(json.load(sys.stdin).get('access_token',''))")
AUTH="Authorization: Bearer $TOKEN"
CT="Content-Type: application/json"

# 全市场扫描：11000+ 信号 → 精简候选
curl -s -H "$AUTH" "$BASE/api/v1/selection/daily"
# 返回: {meta:{trade_date, total_signals, strategy_config}, market_state:{state,should_enter,position_advice}, candidates, industry_signals}
# candidates 每项: {symbol, name, score, industry, trend, buy_reason, warnings}

# 指定策略 / 指定日期 / 忽略 MA20 空仓保护
curl -s -H "$AUTH" "$BASE/api/v1/selection/daily?strategy=aggressive"
curl -s -H "$AUTH" "$BASE/api/v1/selection/daily?date=2026-08-14&ignore_ma20=true"

# 选股历史 / 做空候选（负分分析）
curl -s -H "$AUTH" "$BASE/api/v1/selection/history"
curl -s -H "$AUTH" "$BASE/api/v1/selection/negative"

curl -s -H "$AUTH" "$BASE/api/v1/market-analysis/money-flow/period?period=5d&dimension=sector&category=shenwan&limit=25"
curl -s -H "$AUTH" "$BASE/api/v1/market-analysis/heatmap?trade_date=2026-08-14"
curl -s -H "$AUTH" "$BASE/api/v1/market-analysis/tags/by-tag?tag=半导体"

# ① 全维度特征（估值/技术/动量/波动/流动性/资金流/风格/行业/筹码/概念/微观结构/情绪）
#    注意：单股特征走 symbols/features（POST，body.symbols 数组；返回 data.items[] 快照）。
#    API 已换算单位（市值→亿元、flow*→百万元、totalMv→亿元），引用时注明
curl -s -X POST -H "$AUTH" -H "$CT" "$BASE/api/v1/research/symbols/features" \
  -d '{"symbols":["600519.SH"]}'

# ② K线（120 日，不复权价 + adj_factor）
curl -s -H "$AUTH" "$BASE/api/v1/research/kline/600519.SH?days=120"
# 多市场 K 线（A-HK-US，daily；A股 QuantDB 本地 parquet 优先）
curl -s -H "$AUTH" "$BASE/api/v1/market/kline?symbol=600519.SH&market=A&period=daily&days=120"

# ③ 模型信号（按 model_id 逐个拉，多模型共识；详见 3.7）
curl -s -H "$AUTH" "$BASE/api/v1/models"                    # 用户模型列表（items[]，含 id/metadata/is_default）
curl -s -H "$AUTH" "$BASE/api/v1/models/inference/stock/600519.SH/history?days=180"
curl -s -H "$AUTH" "$BASE/api/v1/models/inference/stock/600519.SH/history?days=180&model_id=xxx"

# ④ 风险初筛（无独立风险评分端点，用①特征 + ②K线本地判读后填§7.1 风险提示表；
#    策略级回测风险走 [[backtest-center]] §6 /qlib/risk/{backtest_id}/metrics）
#    本地看：估值分位（pe/pb vs 行业）、波动（volStd20）、流动性（金额/换手）、质押/商誉（财务层）、veto 项（ST/*ST/次新/长期停牌直接否决）

# ⑤ 大盘环境（L0，先于一切）
curl -s -H "$AUTH" "$BASE/api/v1/selection/daily"           # market_state 牛熊+仓位建议
curl -s -H "$AUTH" "$BASE/api/v1/market/index-kline?symbol=000001.SH&days=60"
curl -s -H "$AUTH" "$BASE/api/v1/market/overview"           # 多市场指数概览

# ⑥ 订单微结构截面分位（L4b，2026-08 新增；先确认最新分区）
docker exec quantmind ls /data/quantdb/6_ml_datasets/l2_factors/ | tail -1   # → dt=YYYYMMDD
docker exec quantmind python3 - <<'EOF'
import duckdb, pandas as pd
dt = "YYYYMMDD"   # 上一步最新分区
mkt = duckdb.connect().execute(f"SELECT * FROM read_parquet('/data/quantdb/6_ml_datasets/l2_factors/dt={dt}/data.parquet')").df()
s = mkt[mkt.symbol == "600519.SH"].iloc[0]
for f in ["vol_persistence","micro_vpin_vol_ratio","flow_buy_amount","flow_sell_amount",
          "micro_toxicity_persistence","flow_order_duration_p90","flow_cancel_lifetime",
          "flow_order_arrival_rate","micro_trade_interval_mean","vol_tick_density","vol_realized_jump",
          "micro_vpin_50","micro_vpin_ma_20","vol_realized_rrv"]:
    print(f, round(float(s[f]),4), f"{round(100*(mkt[f].astype(float)<float(s[f])).mean(),1)}%")
# 判读：负IC族(v_persist/toxicity_persist/flow_buy/sell/order_arrival/tick_density/realized_jump)≥70%高位=利空；
#       正IC族(vpin_vol_ratio/order_duration/cancel_lifetime/trade_interval_mean/vpin_50/ma_20)≤40%低位=缺资金
EOF

# ⑦ 新闻三步纵深（L7，方法论见 docs/news_sentiment_deep_report.md §13）
#    ①直接消息：Huntly 库按标题/正文搜 {名称}/{代码}（窗口 T-2~T）；②相关行业归类；③21 条规律对照
curl -s -H "$AUTH" "$BASE/api/v1/news/articles?tickers=600519&industries=白酒&limit=30&strong_only=true"
curl -s -H "$AUTH" "$BASE/api/v1/news/articles?tickers=600519&sort=sentiment_bullish&limit=10"

docker exec quantmind python3 - <<'EOF'
import pandas as pd
base = "/data/quantdb/3_financial_data"
code = "600519.SH"
inc  = pd.read_parquet(f"{base}/income/{code}.parquet")        # 利润表（元）
bal  = pd.read_parquet(f"{base}/balance/{code}.parquet")       # 资产负债表（元）
cf   = pd.read_parquet(f"{base}/cashflow/{code}.parquet")      # 现金流量表（元）
ps   = pd.read_parquet(f"{base}/pershare_index/{code}.parquet")# 每股指标（ROE 直接可用）
dv   = pd.read_parquet(f"{base}/dividend_factors/{code}.parquet")
hn   = pd.read_parquet(f"{base}/holder_num/{code}.parquet")
# 最新 8 期趋势（每期 m_timetag=YYYYMMDD）
for df in (inc, bal, cf, ps):
    print(df.tail(8)[["m_timetag"] + [c for c in df.columns if c in (
        "revenue","net_profit_incl_min_int_inc","net_cash_flows_oper_act","s_fa_eps_basic",
        "s_fa_bps","s_fa_ocfps","equity_roe","net_roe","sales_gross_profit",
        "inc_net_profit_rate","inventory_turnover","goodwill","tot_assets","tot_liab",
        "account_receivable","inventories","tot_shrhldr_eqy_excl_min_int")]])
print(hn.tail(4)[["endDate","shareholder"]])   # 股东户数（户）
print(dv.tail(4)[["time","interest"]])         # interest=每10股派息（元），每股股息=interest/10
EOF

docker exec quantmind python3 - <<'EOF'
import pandas as pd, glob
# 拉近 5 年 valuation 序列算历史分位（Hive 分区 dt=YYYYMMDD/data.parquet）
files = sorted(glob.glob("/data/quantdb/5_technical_derived/valuation/dt=*/data.parquet"))[-1260:]
rows = []
for f in files:
    df = pd.read_parquet(f, columns=["symbol","time","pe_ttm","pb","ps_ttm","dividend_rate","total_mv","float_mv"])
    r = df[df.symbol == "600519.SH"]
    if len(r): rows.append(r.iloc[0])
v = pd.DataFrame(rows).sort_values("time").dropna(subset=["pe_ttm"])
cur = v.iloc[-1]
print("当前 PE %.2f 处于近5年 %.0f%% 分位（%d 个交易日）" % (cur.pe_ttm, (v.pe_ttm <= cur.pe_ttm).mean()*100, len(v)))
print("PB %.2f 分位 %.0f%%，股息率 %.3f%%" % (cur.pb, (v.pb <= cur.pb).mean()*100, cur.dividend_rate))
print("总市值 %.0f 亿元 / 流通市值 %.0f 亿元" % (cur.total_mv/1e8, cur.float_mv/1e8))
EOF

docker exec quantmind python3 - <<'EOF'
import pandas as pd, glob
files = sorted(glob.glob("/data/quantdb/2_base_sector/margin_trading/dt=*/data.parquet"))[-30:]
rows = []
for f in files:
    df = pd.read_parquet(f, columns=["symbol","time","finance_balance","finance_buy","finance_repay","finance_net","slo_volume","slo_net"])
    r = df[df.symbol == "600519.SH"]
    if len(r): rows.append(r.iloc[0])
m = pd.DataFrame(rows).sort_values("time")
cur = m.iloc[-1]
print("融资余额 %.1f 亿元（近30日 %+.1f%%）" % (cur.finance_balance/1e4, (cur.finance_balance/m.iloc[0].finance_balance-1)*100))
print("近30日融资净买入累计 %.1f 万元，融券净卖出 %.0f 股" % (m.finance_net.tail(20).sum(), m.slo_net.tail(20).sum()))
EOF

curl -s -H "$AUTH" "$BASE/api/v1/news/articles?tickers=600519&limit=30"
curl -s -H "$AUTH" "$BASE/api/v1/news/articles?tickers=600519&industries=白酒&strong_only=true"
curl -s -H "$AUTH" "$BASE/api/v1/news/articles?tickers=600519&sort=sentiment_bullish&limit=10"

curl -s -H "$AUTH" "$BASE/api/v1/admin/data-platform/quantdb/catalog"
curl -s -H "$AUTH" "$BASE/api/v1/admin/data-platform/quantdb/preview?dataset=l1_factors&limit=5"

curl -s -H "$AUTH" "$BASE/api/v1/selection/daily" -o /tmp/selection.json
python3 <<'EOF'
import json, csv
d = json.load(open('/tmp/selection.json'))
meta = d.get('meta', {}); ms = d.get('market_state', {})
cands = d.get('candidates', [])
with open('/tmp/selection.csv', 'w', newline='', encoding='utf-8-sig') as f:
    w = csv.writer(f)
    w.writerow(['代码','名称','分数','行业','趋势','买入理由'])
    for c in cands:
        w.writerow([c.get('symbol'), c.get('name'), round(c.get('score',0),4), c.get('industry'), c.get('trend'), c.get('buy_reason')])
print(f'导出 {len(cands)} 只选股 → /tmp/selection.csv（交易日 {meta.get("trade_date")}，市场状态 {ms.get("state")}）')
EOF

curl -s -X POST -H "$AUTH" -H "$CT" "$BASE/api/v1/research/symbols/features" \
  -d '{"symbols":["600519.SH"]}' -o /tmp/stock_features.json
python3 <<'EOF'
import json
d = json.load(open('/tmp/stock_features.json')).get('data', {})
snap = (d.get('items') or [d])[0]  # 快照：嵌套分类 {类: {字段: 值}} 或扁平字段
rows = []
for cat, fields in snap.items():
    if isinstance(fields, dict):
        for k, v in fields.items():
            rows.append([cat, k, v])
    elif cat not in ('code', 'symbol'):
        rows.append(['base', cat, fields])
with open('/tmp/stock_features.csv', 'w', newline='', encoding='utf-8-sig') as f:
    w = csv.writer(f)
    w.writerow(['类别','字段','值'])
    for r in rows: w.writerow(r)
print(f'导出 {len(rows)} 个字段 → /tmp/stock_features.csv（API 已换算：市值=亿元、flow=百万元）')
EOF

curl -s -X POST -H "$AUTH" -H "$CT" "$BASE/api/v1/research/batch-features" \
  -d '{"symbols":["600519.SH","000858.SZ","601318.SH"],"fields":["pe","pb","roe","totalMv","momRet20d","volStd20","mainFlow"]}' \
  -o /tmp/batch_features.json
python3 <<'EOF'
import json, csv
d = json.load(open('/tmp/batch_features.json')).get('data', {}).get('items', [])
with open('/tmp/stock_compare.csv', 'w', newline='', encoding='utf-8-sig') as f:
    w = csv.writer(f)
    if d:
        w.writerow(['代码'] + list(d[0].get('values', {}).keys()))
        for it in d:
            w.writerow([it.get('symbol')] + list(it.get('values', {}).values()))
print(f'导出 {len(d)} 只股票对比 → /tmp/stock_compare.csv')
EOF

curl -s -X POST -H "$AUTH" -H "$CT" "$BASE/api/v1/research/batch-features" \
  -d '{"symbols":["600519.SH","000858.SZ","601318.SH"],"fields":["pe","pb","roe","totalMv","momRet20d","volStd20","mainFlow"]}' \
  -o /tmp/risk_fields.json
python3 <<'EOF'
import json, csv
d = json.load(open('/tmp/risk_fields.json')).get('data', {}).get('items', [])
with open('/tmp/risk_fields.csv', 'w', newline='', encoding='utf-8-sig') as f:
    w = csv.writer(f)
    if d:
        w.writerow(['代码'] + list(d[0].get('values', {}).keys()))
        for it in d:
            w.writerow([it.get('symbol')] + list(it.get('values', {}).values()))
print(f'导出 {len(d)} 只风险字段 → /tmp/risk_fields.csv（评分按§7.1 风险提示表人工/模型判读，不虚构端点）')
EOF

# 某次推理批次的候选池（含各股行业/概念/指数/分数）
curl -s -H "$AUTH" "$BASE/api/v1/research/overview?limit=100"
# 指定 run 的全池数据（筛选/排序需要全池）
curl -s -H "$AUTH" "$BASE/api/v1/research/universe?run_id=run_20260805_xxx&limit=2000"

# {股票名}（{代码}）深度分析报告

> **报告日期**：{YYYY-MM-DD}　**数据截至**：{最新交易日，注明各数据集实际日期}
> **分析框架**：市场环境 → 基本面 → 估值 → 技术 → 资金筹码 → 行业 → 模型 → 舆情
> **免责声明**：本报告由 AI 自动生成，仅供研究参考，不构成投资建议。

## 一、投资要点（3-5 条，每条 ≤ 1 行，多空都写）

## 二、核心结论

| 维度 | 评级 | 核心依据（带数值） |
|---|---|---|
| 市场环境 | 中性/偏多/偏空 | 上证 3927 vs MA20 3890，建议仓位 40% |
| 基本面 | 优秀/良好/一般/恶化 | 单季营收 +12.3%，现金流/净利润 1.15 |
| 估值 | 低估/合理/高估 | PE 23.1x = 近5年 18% 分位 |
| 技术面 | 多头/空头/震荡 | 收盘站上 MA20/60，MACD 金叉后柱状放大 |
| 资金面 | 流入/流出/分歧 | 近5日大单净流入 +2.1 亿元 |
| 综合评级 | 买入/增持/中性/减持 | 好公司 + 好价格 + 时机待确认 |

## 三、公司概况与业务透视
（名称、代码、申万行业、上市日期、总市值/流通市值、两融标的与否）

## 四、财务分析（最近 8 个报告期表格 + 趋势解读）
| 报告期 | 营收(亿元) | 单季同比% | 归母净利(亿元) | 毛利率% | 净利率% | ROE% | EPS(元) | 经营现金流(亿元) |
|---|---|---|---|---|---|---|---|---|
（每个 % 值注明：毛利=利润表 sales_gross_profit，同比=本期/去年同期−1，现金流/净利=利润质量）

## 五、估值分析
| 指标 | 当前值 | 近5年分位 | 行业相对 | 判断 |
|---|---|---|---|---|
| PE(TTM) | 23.1x | 18% | 0.72 | 低估 |
| PB | … | … | — | … |
| 股息率 | 0.39% | … | — | … |

## 六、技术分析
（趋势/动能/波动/量价分层 + 关键支撑压力位 + 各指标当前值与判断依据）

## 七、资金面与筹码
（超大单/大单/散户三口径 + 近5日/20日趋势 + 融资融券 + 获利盘 + 派发/吸筹判断）

## 八、行业与概念
（行业强度/拥挤度/资金流排名 + 共振/逆势/掉队分类 + 概念热度）

## 九、AI 模型信号
| 模型 | 周期 | 最新分数 | 180日趋势 | 信号 | 排名 |
|---|---|---|---|---|---|
（共识度 + 与量价背离说明）

## 十、多空证据对照
| 层 | 多方证据 | 空方证据 |
|---|---|---|
（每格带数值，无证据写「—」）

## 十一、风险提示（按严重度排序）
1. **{风险}**：{触发条件 + 影响 + 监测指标}

## 十二、操作建议
| 持仓状态 | 建议 | 触发条件 |
|---|---|---|
| 未持仓 | 等待回踩 62 元或模型转 BUY | 放量突破 68 元可追 |
| 已持仓 | 持有，止损 58 元 | 跌破 MA60 或大单连续 3 日流出减仓 |