feat: zentrale data_config.yaml + apply_config.py für dynamische Datenkonfiguration

apply_config.py

@@ -0,0 +1,91 @@
+#!/usr/bin/env python3
+"""
+Liest data_config.yaml und schreibt alle Werte in:
+- .env (Zeiträume, Pfade)
+- generate.py (Qlib Datengenerierung)
+"""
+import yaml
+import re
+from pathlib import Path
+
+CONFIG = Path(__file__).parent / "data_config.yaml"
+ENV = Path(__file__).parent / ".env"
+GENERATE = Path("/home/nico/miniconda3/envs/rdagent/lib/python3.10/site-packages/rdagent/scenarios/qlib/experiment/factor_data_template/generate.py")
+
+with open(CONFIG) as f:
+    cfg = yaml.safe_load(f)
+
+# --- .env updaten ---
+env_text = ENV.read_text()
+
+replacements = {
+    r"QLIB_DATA_DIR=.*": f"QLIB_DATA_DIR={cfg['data_path'].replace('~', str(Path.home()))}",
+    r"QLIB_FREQ=.*": f"QLIB_FREQ={cfg['frequency']}",
+    r"QLIB_FACTOR_TRAIN_START=.*": f"QLIB_FACTOR_TRAIN_START={cfg['train_start']}",
+    r"QLIB_FACTOR_TRAIN_END=.*": f"QLIB_FACTOR_TRAIN_END={cfg['train_end']}",
+    r"QLIB_FACTOR_VALID_START=.*": f"QLIB_FACTOR_VALID_START={cfg['valid_start']}",
+    r"QLIB_FACTOR_VALID_END=.*": f"QLIB_FACTOR_VALID_END={cfg['valid_end']}",
+    r"QLIB_FACTOR_TEST_START=.*": f"QLIB_FACTOR_TEST_START={cfg['test_start']}",
+    r"QLIB_FACTOR_TEST_END=.*": f"QLIB_FACTOR_TEST_END={cfg['test_end']}",
+    r"QLIB_MODEL_TRAIN_START=.*": f"QLIB_MODEL_TRAIN_START={cfg['train_start']}",
+    r"QLIB_MODEL_TRAIN_END=.*": f"QLIB_MODEL_TRAIN_END={cfg['train_end']}",
+    r"QLIB_MODEL_VALID_START=.*": f"QLIB_MODEL_VALID_START={cfg['valid_start']}",
+    r"QLIB_MODEL_VALID_END=.*": f"QLIB_MODEL_VALID_END={cfg['valid_end']}",
+    r"QLIB_MODEL_TEST_START=.*": f"QLIB_MODEL_TEST_START={cfg['test_start']}",
+    r"QLIB_MODEL_TEST_END=.*": f"QLIB_MODEL_TEST_END={cfg['test_end']}",
+    r"QLIB_QUANT_TRAIN_START=.*": f"QLIB_QUANT_TRAIN_START={cfg['train_start']}",
+    r"QLIB_QUANT_TRAIN_END=.*": f"QLIB_QUANT_TRAIN_END={cfg['train_end']}",
+    r"QLIB_QUANT_VALID_START=.*": f"QLIB_QUANT_VALID_START={cfg['valid_start']}",
+    r"QLIB_QUANT_VALID_END=.*": f"QLIB_QUANT_VALID_END={cfg['valid_end']}",
+    r"QLIB_QUANT_TEST_START=.*": f"QLIB_QUANT_TEST_START={cfg['test_start']}",
+    r"QLIB_QUANT_TEST_END=.*": f"QLIB_QUANT_TEST_END={cfg['test_end']}",
+}
+
+for pattern, replacement in replacements.items():
+    env_text = re.sub(pattern, replacement, env_text)
+
+ENV.write_text(env_text)
+print("✓ .env aktualisiert")
+
+# --- generate.py updaten ---
+data_path = cfg['data_path']
+freq = cfg['frequency']
+train_start = cfg['train_start']
+test_end = cfg['test_end']
+valid_start = cfg['valid_start']
+cols = str(cfg['columns'])
+
+generate_text = f'''import qlib
+import pandas as pd
+
+qlib.init(provider_uri="{data_path}", freq="{freq}")
+
+from qlib.data import D
+
+instruments = D.instruments(market="all")
+fields = {cols}
+
+data = (
+    D.features(instruments, fields, freq="{freq}")
+    .swaplevel()
+    .sort_index()
+    .loc["{train_start}":]
+    .sort_index()
+)
+data.to_hdf("./daily_pv_all.h5", key="data")
+
+data_debug = (
+    D.features(instruments, fields, start_time="{valid_start}", end_time="{test_end}", freq="{freq}")
+    .swaplevel()
+    .sort_index()
+)
+data_debug.to_hdf("./daily_pv_debug.h5", key="data")
+'''
+
+GENERATE.write_text(generate_text)
+print("✓ generate.py aktualisiert")
+print(f"\nKonfiguration angewendet:")
+print(f"  Instrument: {cfg['instrument']}")
+print(f"  Frequenz:   {cfg['frequency']}")
+print(f"  Train:      {cfg['train_start']} → {cfg['train_end']}")
+print(f"  Test:       {cfg['test_start']} → {cfg['test_end']}")

data_config.yaml

@@ -0,0 +1,44 @@
+# ============================================================
+#  Predix Data Configuration
+#  Ändere hier Instrument, Frequenz und Zeiträume
+#  Alle anderen Komponenten lesen aus dieser Datei
+# ============================================================
+
+instrument: EURUSD
+frequency: 15min          # 1min, 5min, 15min, 1h, 1d
+data_path: ~/.qlib/qlib_data/eurusd_data
+
+# Verfügbare Spalten (keine $factor Spalte!)
+columns:
+  - $open
+  - $close
+  - $high
+  - $low
+  - $volume
+
+# Walk-Forward Split
+train_start: "2022-03-14"
+train_end:   "2024-06-30"
+valid_start: "2024-07-01"
+valid_end:   "2024-12-31"
+test_start:  "2025-01-01"
+test_end:    "2026-03-20"
+
+# Markt-Kontext für LLM Prompts
+market_context:
+  spread_bps: 1.5
+  sessions:
+    asian:   "00:00-08:00 UTC"
+    london:  "08:00-16:00 UTC"
+    ny:      "13:00-21:00 UTC"
+    overlap: "13:00-16:00 UTC"
+  target_arr: 9.62          # % ARR zu schlagen
+  max_drawdown: 20          # % maximaler Drawdown
+
+# Lookback Referenz (in Bars)
+lookback:
+  1h:  4
+  2h:  8
+  4h:  16
+  8h:  32
+  1d:  96