"""
config.py — 全局常量、特征定义、路径、API Key
=================================================
所有模块共享的配置集中在此。
"""
import os

# ── 工作目录 ──
BASE_DIR = os.environ.get('OILVERSE_BASE_DIR', os.path.dirname(os.path.abspath(__file__)))
OUTPUT_DIR = os.path.join(BASE_DIR, 'output')
DATA_DIR = os.path.join(BASE_DIR, 'data', 'csv_raw')

# ── API Keys ──
FRED_API_KEY = 'fc02a6e6a359a4cc16f0f1752d258011'
EIA_API_KEY = '9Nv5PhLREMmmKeo0zJ2U3Zu21Bntf8DfhEKBpi55'
SILICONFLOW_API_KEY = 'sk-cgllfrsuchzzwerkxcegkhbroboqcnpuubhreyrfudfstqjv'
SILICONFLOW_BASE_URL = 'https://api.siliconflow.cn/v1'
SILICONFLOW_MODEL = 'Qwen/Qwen2.5-7B-Instruct'

# ── 价格列 ──
PRICE_COLS = {
    'WTI':   'WTI_spot',
    'Brent': 'Brent_spot',
}
PRICE_COL = 'WTI_spot'  # backward compat

# ── 特征列表 (20个，含GPR + 新闻情绪) ──
FEATURES = [
    'Brent_spot', 'natgas_spot_henry', 'iron_ore_spot',              # Price
    'rig_count_us_new', 'supply_saudi', 'us_oil_inventory_total',    # Supply
    'pmi_us_mfg', 'usd_index', 'nonfarm_us', 'ipi_us',             # Demand
    'vix_lag1', 'vix_lag2', 'geo_shock_count', 'geo_active_events',  # Risk+Geo
    'mom1m_lag1', 'hist_vol_12m', 'rsi12m',                          # Technical
    'news_oil_sentiment', 'news_geo_tone', 'news_article_volume',    # Alternative (GDELT)
]

# ── 因子分组 ──
FACTOR_GROUPS = {
    'Price':       ['Brent_spot', 'natgas_spot_henry', 'iron_ore_spot'],
    'Supply':      ['rig_count_us_new', 'supply_saudi', 'us_oil_inventory_total'],
    'Demand':      ['pmi_us_mfg', 'usd_index', 'nonfarm_us', 'ipi_us'],
    'Risk_Geo':    ['vix_lag1', 'vix_lag2', 'geo_shock_count', 'geo_active_events'],
    'Technical':   ['mom1m_lag1', 'hist_vol_12m', 'rsi12m'],
    'Alternative': ['news_oil_sentiment', 'news_geo_tone', 'news_article_volume'],
}

# ── Walk-Forward 参数 ──
TRAIN_WINDOW = 120
MAX_FEATURES = 10
MIN_TRAIN_SAMPLES = 20

# ── Regime 签名 (历史时期特征向量) ──
REGIME_SIGNATURES = {
    '2008 金融危机':   {'supply_stress': 0.1,  'demand_stress': 0.6,  'geopolitical_stress': 0.1,  'price_momentum': 0.15, 'vol_level': 0.25},
    '2014 页岩油冲击': {'supply_stress': 0.5,  'demand_stress': 0.2,  'geopolitical_stress': 0.05, 'price_momentum': 0.2,  'vol_level': 0.12},
    '2020 COVID':     {'supply_stress': 0.3,  'demand_stress': 0.5,  'geopolitical_stress': 0.05, 'price_momentum': 0.1,  'vol_level': 0.30},
    '2022 俄乌冲突':   {'supply_stress': 0.3,  'demand_stress': 0.1,  'geopolitical_stress': 0.5,  'price_momentum': 0.05, 'vol_level': 0.15},
    '2023 OPEC减产':  {'supply_stress': 0.5,  'demand_stress': 0.15, 'geopolitical_stress': 0.15, 'price_momentum': 0.1,  'vol_level': 0.08},
    '常态/低波动':     {'supply_stress': 0.15, 'demand_stress': 0.15, 'geopolitical_stress': 0.1,  'price_momentum': 0.1,  'vol_level': 0.04},
}

REGIME_TYPE_MAP = {
    '2008 金融危机': 'demand_collapse',
    '2014 页岩油冲击': 'supply_glut',
    '2020 COVID': 'demand_collapse',
    '2022 俄乌冲突': 'geopolitical',
    '2023 OPEC减产': 'supply_cut',
    '常态/低波动': 'normal',
}

# ── 行业映射 ──
INDUSTRIES = ['Aviation', 'Logistics', 'Chemicals', 'Manufacturing', 'Upstream_OG']
INDUSTRY_ZH = {
    'Aviation': '航空', 'Logistics': '物流', 'Chemicals': '化工',
    'Manufacturing': '制造', 'Upstream_OG': '上游油气',
}

# ── 输出文件路径 ──
OUTPUT_FILES = {
    'results':     os.path.join(OUTPUT_DIR, 'v2_championship_results.csv'),
    'shap':        os.path.join(OUTPUT_DIR, 'v2_shap_records.json'),
    'nlg':         os.path.join(OUTPUT_DIR, 'v2_nlg_reports.json'),
    'scenarios':   os.path.join(OUTPUT_DIR, 'v2_scenarios.json'),
    'regime':      os.path.join(OUTPUT_DIR, 'v2_regime_data.json'),
    'ablation':    os.path.join(OUTPUT_DIR, 'v2_ablation.json'),
    'hedging':     os.path.join(OUTPUT_DIR, 'v2_hedging.json'),
    'backtest':    os.path.join(OUTPUT_DIR, 'v2_hedge_backtest.json'),
    'feat_sel':    os.path.join(OUTPUT_DIR, 'v2_feature_selection.json'),
    'dashboard':   os.path.join(OUTPUT_DIR, 'risk_dashboard.html'),
    'panel_live':  os.path.join(OUTPUT_DIR, 'panel_monthly_live.csv'),
    'lineage':     os.path.join(OUTPUT_DIR, 'data_lineage.json'),
    'quality':     os.path.join(OUTPUT_DIR, 'data_quality.json'),
    'causal':      os.path.join(OUTPUT_DIR, 'causal_analysis.json'),
    'events':      os.path.join(OUTPUT_DIR, 'event_timeline.json'),
}