chore(master): release 1.3.7 (#41)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

.github/workflows/ci.yml

@@ -14,7 +14,7 @@ jobs:
   security:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
 
       - name: Run Bandit (Security Scan)
         uses: PyCQA/bandit-action@v1
@@ -25,9 +25,9 @@ jobs:
   test:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
 
-      - uses: actions/setup-python@v5
+      - uses: actions/setup-python@v6
         with:
           python-version: "3.10"
           cache: "pip"

.github/workflows/codacy.yml

@@ -36,11 +36,11 @@ jobs:
     steps:
       # Checkout the repository to the GitHub Actions runner
       - name: Checkout code
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
 
       # Execute Codacy Analysis CLI and generate a SARIF output with the security issues identified during the analysis
       - name: Run Codacy Analysis CLI
-        uses: codacy/codacy-analysis-cli-action@d840f886c4bd4edc059706d09c6a1586111c540b
+        uses: codacy/codacy-analysis-cli-action@562ee3e92b8e92df8b67e0a5ff8aa8e261919c08
         env:
           JAVA_TOOL_OPTIONS: "-Dfile.encoding=UTF-8"
         with:

.github/workflows/conventional-commits.yml

@@ -46,7 +46,7 @@ jobs:
     name: Validate Commit Messages
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
         with:
           fetch-depth: 0
 

.github/workflows/docs.yml

@@ -25,10 +25,10 @@ jobs:
 
     steps:
     - name: Checkout repository
-      uses: actions/checkout@v4
+      uses: actions/checkout@v6
 
     - name: Set up Python
-      uses: actions/setup-python@v5
+      uses: actions/setup-python@v6
       with:
         python-version: "3.10"
 
@@ -64,7 +64,7 @@ jobs:
 
     - name: Upload docs artifact
       if: github.ref == 'refs/heads/main'
-      uses: actions/upload-pages-artifact@v3
+      uses: actions/upload-pages-artifact@v5
       with:
         path: docs/_build/html
 

.github/workflows/lint.yml

@@ -16,10 +16,10 @@ jobs:
 
     steps:
     - name: Checkout repository
-      uses: actions/checkout@v4
+      uses: actions/checkout@v6
 
     - name: Set up Python
-      uses: actions/setup-python@v5
+      uses: actions/setup-python@v6
       with:
         python-version: "3.10"
 

.github/workflows/release.yml

@@ -12,7 +12,7 @@ jobs:
   release-please:
     runs-on: ubuntu-latest
     steps:
-      - uses: googleapis/release-please-action@v4
+      - uses: googleapis/release-please-action@v5
         with:
           token: ${{ secrets.GITHUB_TOKEN }}
           config-file: release-please-config.json

.github/workflows/scheduled-tests.yml

@@ -19,9 +19,9 @@ jobs:
         python-version: ["3.10", "3.11"]
 
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
 
-      - uses: actions/setup-python@v5
+      - uses: actions/setup-python@v6
         with:
           python-version: ${{ matrix.python-version }}
           cache: "pip"
@@ -49,9 +49,9 @@ jobs:
     name: Dependency Audit
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
 
-      - uses: actions/setup-python@v5
+      - uses: actions/setup-python@v6
         with:
           python-version: "3.10"
           cache: "pip"

.github/workflows/security.yml

@@ -19,10 +19,10 @@ jobs:
 
     steps:
     - name: Checkout repository
-      uses: actions/checkout@v4
+      uses: actions/checkout@v6
 
     - name: Set up Python
-      uses: actions/setup-python@v5
+      uses: actions/setup-python@v6
       with:
         python-version: "3.10"
 

.release-please-manifest.json

@@ -1,3 +1,3 @@
 {
-  ".": "1.2.2"
+  ".": "1.3.7"
 }

CHANGELOG.md

@@ -1,5 +1,105 @@
 # Changelog
 
+## [1.3.7](https://github.com/TPTBusiness/Predix/compare/v1.3.6...v1.3.7) (2026-04-30)
+
+
+### Bug Fixes
+
+* **security:** nosec for B608/B701 false positives in UI and template code ([5eb5d7e](https://github.com/TPTBusiness/Predix/commit/5eb5d7e8fdbe90e0dced83fef4e09f5a33e96b2b))
+* **security:** replace eval() with ast.literal_eval and add request timeouts (B307, B113) ([3301ada](https://github.com/TPTBusiness/Predix/commit/3301ada697ca7d3afa1a188d2a76a87ae98b4529))
+* **security:** replace shell=True subprocess calls with list args (B602) ([13c08f4](https://github.com/TPTBusiness/Predix/commit/13c08f4ce6813eb7c314087921ec8c0f40074bd7))
+
+## [1.3.6](https://github.com/TPTBusiness/Predix/compare/v1.3.5...v1.3.6) (2026-04-30)
+
+
+### Bug Fixes
+
+* **security:** real fix for B110 (logging in factor_proposal.py [#746](https://github.com/TPTBusiness/Predix/issues/746)) ([16624e0](https://github.com/TPTBusiness/Predix/commit/16624e0bd966ae4d24c4a3eb42bbc31c11da3136))
+* **security:** real fix for B110 (logging in factor_runner.py [#744](https://github.com/TPTBusiness/Predix/issues/744)) ([88cf0fb](https://github.com/TPTBusiness/Predix/commit/88cf0fb8828b11c97f2f3ae2881a4900b020c6f0))
+* **security:** real fix for B110 (logging in quant_proposal.py [#741](https://github.com/TPTBusiness/Predix/issues/741)) ([7cf2a64](https://github.com/TPTBusiness/Predix/commit/7cf2a644f553b054bd4b0607ea51e5372e68d90a))
+* **security:** real fix for B110 (logging in quant_proposal.py [#741](https://github.com/TPTBusiness/Predix/issues/741)) ([ef985f8](https://github.com/TPTBusiness/Predix/commit/ef985f86035d8dca707c60137e6508349a0c4ae6))
+* **security:** real fix for B404/B603 (sys.executable in factor_runner.py [#745](https://github.com/TPTBusiness/Predix/issues/745)) ([819655a](https://github.com/TPTBusiness/Predix/commit/819655aaa3efa76596d60501d0e8ca365df3e5e2))
+* **security:** revert broken read_pickle encoding arg in kaggle template (B301) ([3574907](https://github.com/TPTBusiness/Predix/commit/35749073c91e69f63ddaad61dae3f2b799327e63))
+* **security:** validate SQL identifiers in _add_column_if_not_exists (B608) ([e10dfa2](https://github.com/TPTBusiness/Predix/commit/e10dfa2576038e911f83595d3b466c261bc0cd54))
+* **security:** whitelist-validate metric column in get_top_factors (B608) ([e50519f](https://github.com/TPTBusiness/Predix/commit/e50519fe066e68aec2f19b83df4f643c3c22053d))
+
+## [1.3.5](https://github.com/TPTBusiness/Predix/compare/v1.3.4...v1.3.5) (2026-04-27)
+
+
+### Bug Fixes
+
+* **auto-fixer:** add five new factor code fixes for groupby/apply errors ([449c8fd](https://github.com/TPTBusiness/Predix/commit/449c8fd70a327e604dcca122e4a134f0cca918e4))
+* **auto-fixer:** add four new factor code fixes for common runtime errors ([40484f6](https://github.com/TPTBusiness/Predix/commit/40484f6d300425da481f1edd325da4acbc06ec7d))
+* **auto-fixer:** add groupby([level=N,'date']) SyntaxError fix ([ca77c00](https://github.com/TPTBusiness/Predix/commit/ca77c005bea4abdd8854c1de2b0e8d03b7742161))
+* **auto-fixer:** disable _fix_min_periods for intraday data ([77b0740](https://github.com/TPTBusiness/Predix/commit/77b0740f059349df7e769a378af728aa33b2070e))
+* **auto-fixer:** fix chained groupby(level=N).groupby('date') pattern ([7d5fe32](https://github.com/TPTBusiness/Predix/commit/7d5fe32b31a19ce8b04bd8f5a430720fdb748f7a))
+* **auto-fixer:** fix df.loc[instrument] DateParseError on MultiIndex frames ([b7860ea](https://github.com/TPTBusiness/Predix/commit/b7860eafc0ad26384947ce0510ecf4e9f3425807))
+* **auto-fixer:** fix df['instrument'] KeyError on MultiIndex frames ([aad6bd1](https://github.com/TPTBusiness/Predix/commit/aad6bd1c7c720b3d486e0cf248337f32394773b1))
+* **auto-fixer:** fix two assignment-target bugs in instrument column fixers ([421eedf](https://github.com/TPTBusiness/Predix/commit/421eedffed4b883c24397dc5581c019a3985277f))
+* **auto-fixer:** preserve date dimension in groupby(['instrument','date']) fix ([b58fdd8](https://github.com/TPTBusiness/Predix/commit/b58fdd8be43720b5d4363e0f8de9a01591d4d2dc))
+* **auto-fixer:** remove ddof from rolling() args, not only from std()/var() ([b0fc328](https://github.com/TPTBusiness/Predix/commit/b0fc328d0d4a041c65d8eeb32cb3f2bb86568406))
+* **auto-fixer:** strip spurious .reset_index() after .transform() calls ([8708aae](https://github.com/TPTBusiness/Predix/commit/8708aae6e08728cda1875c775a76dc92e43576f3))
+* **loop:** prevent step_idx advance on unhandled exceptions + fix consecutive assistant messages ([5ec4ad1](https://github.com/TPTBusiness/Predix/commit/5ec4ad1b96b5b99ef42bea7bb828cb1ef709a688))
+
+## [1.3.4](https://github.com/TPTBusiness/Predix/compare/v1.3.3...v1.3.4) (2026-04-27)
+
+
+### Bug Fixes
+
+* **auto-fixer:** add five new factor code fixes for groupby/apply errors ([449c8fd](https://github.com/TPTBusiness/Predix/commit/449c8fd70a327e604dcca122e4a134f0cca918e4))
+* **auto-fixer:** add four new factor code fixes for common runtime errors ([40484f6](https://github.com/TPTBusiness/Predix/commit/40484f6d300425da481f1edd325da4acbc06ec7d))
+* **auto-fixer:** add groupby([level=N,'date']) SyntaxError fix ([ca77c00](https://github.com/TPTBusiness/Predix/commit/ca77c005bea4abdd8854c1de2b0e8d03b7742161))
+* **auto-fixer:** disable _fix_min_periods for intraday data ([77b0740](https://github.com/TPTBusiness/Predix/commit/77b0740f059349df7e769a378af728aa33b2070e))
+* **auto-fixer:** fix chained groupby(level=N).groupby('date') pattern ([7d5fe32](https://github.com/TPTBusiness/Predix/commit/7d5fe32b31a19ce8b04bd8f5a430720fdb748f7a))
+* **auto-fixer:** fix df.loc[instrument] DateParseError on MultiIndex frames ([b7860ea](https://github.com/TPTBusiness/Predix/commit/b7860eafc0ad26384947ce0510ecf4e9f3425807))
+* **auto-fixer:** fix df['instrument'] KeyError on MultiIndex frames ([aad6bd1](https://github.com/TPTBusiness/Predix/commit/aad6bd1c7c720b3d486e0cf248337f32394773b1))
+* **auto-fixer:** preserve date dimension in groupby(['instrument','date']) fix ([b58fdd8](https://github.com/TPTBusiness/Predix/commit/b58fdd8be43720b5d4363e0f8de9a01591d4d2dc))
+* **auto-fixer:** remove ddof from rolling() args, not only from std()/var() ([b0fc328](https://github.com/TPTBusiness/Predix/commit/b0fc328d0d4a041c65d8eeb32cb3f2bb86568406))
+* **backtest:** replace broken MC permutation test with binomial win-rate test ([c38d894](https://github.com/TPTBusiness/Predix/commit/c38d89478f586825bfca5715a96ca70ccd8791a3))
+* **factors:** detect and correct look-ahead bias in daily-constant factors ([eb490a4](https://github.com/TPTBusiness/Predix/commit/eb490a461b66cbd815ae53ac5205115754712432))
+* **factors:** extend look-ahead rules to session factors and add intraday-factor guidance ([c24c100](https://github.com/TPTBusiness/Predix/commit/c24c100442d6487686c0578de0b32d240fcbf215))
+* **loop:** compress old experiment history in proposal prompt to reduce context size ([4bf90a9](https://github.com/TPTBusiness/Predix/commit/4bf90a905ba8b2aba2a818191c19998088cccaaf))
+* **loop:** prevent step_idx advance on unhandled exceptions + fix consecutive assistant messages ([5ec4ad1](https://github.com/TPTBusiness/Predix/commit/5ec4ad1b96b5b99ef42bea7bb828cb1ef709a688))
+
+## [1.3.3](https://github.com/TPTBusiness/Predix/compare/v1.3.2...v1.3.3) (2026-04-25)
+
+
+### Bug Fixes
+
+* **backtest:** replace broken MC permutation test with binomial win-rate test ([c38d894](https://github.com/TPTBusiness/Predix/commit/c38d89478f586825bfca5715a96ca70ccd8791a3))
+* **factors:** detect and correct look-ahead bias in daily-constant factors ([eb490a4](https://github.com/TPTBusiness/Predix/commit/eb490a461b66cbd815ae53ac5205115754712432))
+* **factors:** extend look-ahead rules to session factors and add intraday-factor guidance ([c24c100](https://github.com/TPTBusiness/Predix/commit/c24c100442d6487686c0578de0b32d240fcbf215))
+* **loop:** compress old experiment history in proposal prompt to reduce context size ([4bf90a9](https://github.com/TPTBusiness/Predix/commit/4bf90a905ba8b2aba2a818191c19998088cccaaf))
+* **strategies:** guard against None IC in acceptance check, disable slow wf_rolling ([2197f52](https://github.com/TPTBusiness/Predix/commit/2197f52150a50ef38d9e70991d7e48c8c30caec4))
+* **strategies:** handle None ic/sharpe/dd in rejected strategy log output ([ad2ad3a](https://github.com/TPTBusiness/Predix/commit/ad2ad3ab3360ea75ed3bbc90c12098b9c5cc0114))
+
+## [1.3.2](https://github.com/TPTBusiness/Predix/compare/v1.3.1...v1.3.2) (2026-04-23)
+
+
+### Bug Fixes
+
+* **strategies:** guard against None IC in acceptance check, disable slow wf_rolling ([2197f52](https://github.com/TPTBusiness/Predix/commit/2197f52150a50ef38d9e70991d7e48c8c30caec4))
+* **strategies:** handle None ic/sharpe/dd in rejected strategy log output ([ad2ad3a](https://github.com/TPTBusiness/Predix/commit/ad2ad3ab3360ea75ed3bbc90c12098b9c5cc0114))
+
+## [1.3.1](https://github.com/TPTBusiness/Predix/compare/v1.3.0...v1.3.1) (2026-04-21)
+
+
+### Bug Fixes
+
+* **deps:** bump python-dotenv to >=1.2.2 (CVE symlink overwrite) ([126ae7d](https://github.com/TPTBusiness/Predix/commit/126ae7d5fb556b677d09d10221862a0d648d697a))
+
+## [1.3.0](https://github.com/TPTBusiness/Predix/compare/v1.2.2...v1.3.0) (2026-04-21)
+
+
+### Features
+
+* **backtest:** add rolling walk-forward validation and Monte Carlo trade permutation test ([637a94c](https://github.com/TPTBusiness/Predix/commit/637a94c1d987da763869f4f9b73372a3f37d873c))
+
+
+### Bug Fixes
+
+* **security:** resolve all 30 Bandit security alerts (B301, B614, B104) ([ce5983d](https://github.com/TPTBusiness/Predix/commit/ce5983d9d59c4c34341fb1ec749e44bbcfc4a1c4))
+
 ## [1.2.2](https://github.com/TPTBusiness/Predix/compare/v1.2.1...v1.2.2) (2026-04-19)
 
 

pyproject.toml

@@ -68,7 +68,7 @@ ignore_missing_imports = true
 module = "llama"
 
 [tool.pytest.ini_options]
-addopts = "-l -s --durations=0"
+addopts = "-l -s --durations=0 -m 'not slow'"
 log_cli = true
 log_cli_level = "info"
 log_date_format = "%Y-%m-%d %H:%M:%S"

rdagent/app/utils/info.py

@@ -54,11 +54,11 @@ def rdagent_info():
     current_version = importlib.metadata.version("rdagent")
     logger.info(f"RD-Agent version: {current_version}")
     api_url = f"https://api.github.com/repos/microsoft/RD-Agent/contents/requirements.txt?ref=main"
-    response = requests.get(api_url)
+    response = requests.get(api_url, timeout=30)
     if response.status_code == 200:
         files = response.json()
         file_url = files["download_url"]
-        file_response = requests.get(file_url)
+        file_response = requests.get(file_url, timeout=30)
         if file_response.status_code == 200:
             all_file_contents = file_response.text.split("\n")
         else:

rdagent/components/backtesting/__init__.py

@@ -11,16 +11,23 @@ from .vbt_backtest import (
     FTMO_MAX_LEVERAGE,
     FTMO_RISK_PER_TRADE,
     OOS_START_DEFAULT,
+    WF_IS_YEARS,
+    WF_OOS_YEARS,
+    WF_STEP_YEARS,
     backtest_from_forward_returns,
     backtest_signal,
     backtest_signal_ftmo,
+    monte_carlo_trade_pvalue,
+    walk_forward_rolling,
 )
 
 __all__ = [
     'BacktestMetrics', 'FactorBacktester', 'ResultsDatabase',
     'CorrelationAnalyzer', 'PortfolioOptimizer', 'AdvancedRiskManager',
     'backtest_signal', 'backtest_signal_ftmo', 'backtest_from_forward_returns',
+    'monte_carlo_trade_pvalue', 'walk_forward_rolling',
     'DEFAULT_BARS_PER_YEAR', 'DEFAULT_TXN_COST_BPS',
     'FTMO_INITIAL_CAPITAL', 'FTMO_MAX_DAILY_LOSS', 'FTMO_MAX_TOTAL_LOSS',
     'FTMO_MAX_LEVERAGE', 'FTMO_RISK_PER_TRADE', 'OOS_START_DEFAULT',
+    'WF_IS_YEARS', 'WF_OOS_YEARS', 'WF_STEP_YEARS',
 ]

rdagent/components/backtesting/results_db.py

@@ -71,6 +71,9 @@ class ResultsDatabase:
 
         self.conn.commit()
 
+    _ALLOWED_TABLES = frozenset({"factors", "backtest_runs", "loop_results"})
+    _ALLOWED_COL_TYPES = frozenset({"REAL", "TEXT", "INTEGER", "BLOB"})
+
     def _add_column_if_not_exists(self, table: str, column: str, col_type: str) -> None:
         """
         Add a column to a table if it doesn't already exist.
@@ -78,20 +81,24 @@ class ResultsDatabase:
         Parameters
         ----------
         table : str
-            Table name
+            Table name (must be in _ALLOWED_TABLES)
         column : str
-            Column name to add
+            Column name to add (alphanumeric + underscore only)
         col_type : str
-            SQL column type (e.g., 'REAL', 'TEXT')
+            SQL column type (must be in _ALLOWED_COL_TYPES)
         """
+        if table not in self._ALLOWED_TABLES:
+            raise ValueError(f"Unknown table: {table!r}")
+        if not column.replace("_", "").isalnum():
+            raise ValueError(f"Invalid column name: {column!r}")
+        if col_type not in self._ALLOWED_COL_TYPES:
+            raise ValueError(f"Invalid column type: {col_type!r}")
+
         c = self.conn.cursor()
-        try:
-            # Try to query the column - if it fails, it doesn't exist
-            # nosec B608: Internal schema migration, column names are controlled
-            c.execute(f"SELECT {column} FROM {table} LIMIT 1")  # nosec B608
-        except sqlite3.OperationalError:
-            # Column doesn't exist, add it
-            c.execute(f"ALTER TABLE {table} ADD COLUMN {column} {col_type}")  # nosec B608
+        c.execute("SELECT name FROM pragma_table_info(?)", (table,))
+        existing = {row[0] for row in c.fetchall()}
+        if column not in existing:
+            c.execute(f"ALTER TABLE {table} ADD COLUMN {column} {col_type}")
     
     def add_factor(self, name: str, type: str = "unknown") -> int:
         c = self.conn.cursor()
@@ -183,16 +190,18 @@ class ResultsDatabase:
         pd.DataFrame
             DataFrame with factor names and metrics
         """
-        # Map shorthand to full column name
+        _ALLOWED_METRICS = frozenset({
+            'sharpe', 'ic', 'annual_return', 'max_drawdown',
+            'win_rate', 'information_ratio', 'volatility',
+        })
         metric_map = {
-            'sharpe': 'sharpe',
-            'ic': 'ic',
-            'return': 'annual_return',
-            'drawdown': 'max_drawdown',
-            'win_rate': 'win_rate',
+            'sharpe': 'sharpe', 'ic': 'ic', 'return': 'annual_return',
+            'drawdown': 'max_drawdown', 'win_rate': 'win_rate',
             'information_ratio': 'information_ratio',
         }
         col = metric_map.get(metric, metric)
+        if col not in _ALLOWED_METRICS:
+            raise ValueError(f"Unknown metric: {metric!r}")
 
         return pd.read_sql_query(
             f"""SELECT factor_name, ic, sharpe, annual_return, max_drawdown,
@@ -201,7 +210,7 @@ class ResultsDatabase:
                 JOIN factors ON factor_id = factors.id
                 WHERE {col} IS NOT NULL
                 ORDER BY {col} DESC
-                LIMIT ?""",
+                LIMIT ?""",  # nosec B608 — col is validated against _ALLOWED_METRICS above
             self.conn,
             params=[limit]
         )

rdagent/components/backtesting/vbt_backtest.py

@@ -342,6 +342,129 @@ def _apply_ftmo_mask(
 
 OOS_START_DEFAULT = "2024-01-01"
 
+# Rolling walk-forward default windows (IS years, OOS years, step years)
+WF_IS_YEARS  = 3
+WF_OOS_YEARS = 1
+WF_STEP_YEARS = 1
+
+
+def monte_carlo_trade_pvalue(
+    trade_pnl: pd.Series,
+    n_permutations: int = 1000,
+    seed: int = 0,
+) -> float:
+    """
+    Monte Carlo permutation test on trade-level P&L.
+
+    Runs a one-sided binomial test on trade-level win rate.
+
+    Tests H0: win_rate = 0.5 (random trading) against H1: win_rate > 0.5.
+    The ``n_permutations`` parameter is kept for API compatibility but is unused.
+
+    p < 0.05 → win rate is significantly above 50%, indicating a genuine per-trade edge.
+
+    Parameters
+    ----------
+    trade_pnl : pd.Series
+        Per-trade net returns (output of ``_compute_trade_pnl``).
+    n_permutations : int
+        Number of random permutations (default 1000).
+    seed : int
+        RNG seed for reproducibility.
+
+    Returns
+    -------
+    float
+        p-value in [0, 1]. Lower is better.
+    """
+    if len(trade_pnl) < 2:
+        return 1.0
+    trades = trade_pnl.values.copy()
+    # Binomial test: is the win rate significantly above 50%?
+    # p = probability of observing >= n_wins out of n_trades under null (win_rate=0.5).
+    # Low p → strategy has a significant positive edge per trade.
+    from scipy.stats import binomtest
+    n_wins = int((trades > 0).sum())
+    n_total = len(trades)
+    result = binomtest(n_wins, n_total, p=0.5, alternative="greater")
+    return float(result.pvalue)
+
+
+def walk_forward_rolling(
+    close: pd.Series,
+    signal: pd.Series,
+    leverage: float,
+    txn_cost_bps: float = DEFAULT_TXN_COST_BPS,
+    bars_per_year: int = DEFAULT_BARS_PER_YEAR,
+    is_years: int = WF_IS_YEARS,
+    oos_years: int = WF_OOS_YEARS,
+    step_years: int = WF_STEP_YEARS,
+) -> Dict[str, Any]:
+    """
+    Rolling walk-forward validation: multiple IS/OOS windows shifted by ``step_years``.
+
+    Each window runs an independent FTMO simulation on the IS and OOS slices.
+    Produces aggregate OOS statistics to measure cross-time consistency.
+
+    Returns
+    -------
+    dict with keys:
+        wf_n_windows, wf_oos_sharpe_mean, wf_oos_sharpe_std,
+        wf_oos_monthly_return_mean, wf_oos_consistency (fraction of windows
+        with OOS Sharpe > 0), wf_windows (list of per-window dicts)
+    """
+    if not isinstance(close.index, pd.DatetimeIndex):
+        return {"wf_n_windows": 0}
+
+    start_year = close.index[0].year
+    end_year   = close.index[-1].year
+
+    windows = []
+    yr = start_year
+    while True:
+        is_start  = pd.Timestamp(f"{yr}-01-01")
+        is_end    = pd.Timestamp(f"{yr + is_years}-01-01")
+        oos_end   = pd.Timestamp(f"{yr + is_years + oos_years}-01-01")
+        if oos_end.year > end_year + 1:
+            break
+        is_mask  = (close.index >= is_start) & (close.index < is_end)
+        oos_mask = (close.index >= is_end)   & (close.index < oos_end)
+        if is_mask.sum() < 1000 or oos_mask.sum() < 1000:
+            yr += step_years
+            continue
+
+        window: Dict[str, Any] = {
+            "is_start":  str(is_start.date()),
+            "is_end":    str(is_end.date()),
+            "oos_start": str(is_end.date()),
+            "oos_end":   str(oos_end.date()),
+        }
+        for mask, prefix in [(is_mask, "is"), (oos_mask, "oos")]:
+            close_s  = close.loc[mask]
+            signal_s = signal.loc[mask]
+            masked_s, _ = _apply_ftmo_mask(signal_s, close_s, leverage, txn_cost_bps)
+            r = backtest_signal(close=close_s, signal=masked_s,
+                                txn_cost_bps=txn_cost_bps, bars_per_year=bars_per_year)
+            window[f"{prefix}_sharpe"]       = r.get("sharpe", 0.0)
+            window[f"{prefix}_monthly_return_pct"] = r.get("monthly_return_pct", 0.0)
+            window[f"{prefix}_n_trades"]     = r.get("n_trades", 0)
+        windows.append(window)
+        yr += step_years
+
+    if not windows:
+        return {"wf_n_windows": 0}
+
+    oos_sharpes  = [w["oos_sharpe"] for w in windows]
+    oos_monthly  = [w["oos_monthly_return_pct"] for w in windows]
+    return {
+        "wf_n_windows":              len(windows),
+        "wf_oos_sharpe_mean":        float(np.mean(oos_sharpes)),
+        "wf_oos_sharpe_std":         float(np.std(oos_sharpes)),
+        "wf_oos_monthly_return_mean": float(np.mean(oos_monthly)),
+        "wf_oos_consistency":        float(np.mean([s > 0 for s in oos_sharpes])),
+        "wf_windows":                windows,
+    }
+
 
 def backtest_signal_ftmo(
     close: pd.Series,
@@ -354,6 +477,8 @@ def backtest_signal_ftmo(
     bars_per_year: int = DEFAULT_BARS_PER_YEAR,
     forward_returns: Optional[pd.Series] = None,
     oos_start: Optional[str] = OOS_START_DEFAULT,
+    wf_rolling: bool = False,
+    mc_n_permutations: int = 0,
 ) -> Dict[str, Any]:
     """
     FTMO-compliant backtest of a strategy signal on EUR/USD.
@@ -385,6 +510,13 @@ def backtest_signal_ftmo(
         Maximum leverage (default 30 = FTMO 1:30).
     oos_start : str or None
         Start of out-of-sample period (ISO date). None disables OOS split.
+    wf_rolling : bool
+        If True, run rolling walk-forward validation (multiple IS/OOS windows).
+        Results are stored under ``wf_*`` keys. Default False.
+    mc_n_permutations : int
+        Number of Monte Carlo trade permutations. 0 = disabled (default).
+        When > 0, computes ``mc_pvalue``: fraction of permuted sequences whose
+        total return >= real total return. p < 0.05 indicates a genuine edge.
     """
     stop_price = stop_pips * FTMO_PIP
     leverage_by_risk = risk_pct / (stop_price / eurusd_price)
@@ -440,6 +572,28 @@ def backtest_signal_ftmo(
         result["is_n_bars"]  = int(is_mask.sum())
         result["oos_n_bars"] = int(oos_mask.sum())
 
+    # Rolling walk-forward validation
+    if wf_rolling:
+        wf = walk_forward_rolling(
+            close=close,
+            signal=signal,
+            leverage=leverage,
+            txn_cost_bps=txn_cost_bps,
+            bars_per_year=bars_per_year,
+        )
+        result.update(wf)
+
+    # Monte Carlo trade permutation test
+    if mc_n_permutations > 0:
+        position = masked_signal.shift(1).fillna(0)
+        bar_ret  = close.pct_change().fillna(0)
+        txn_cost = txn_cost_bps / 10_000.0
+        position_change = position.diff().abs().fillna(position.abs())
+        strat_ret = position * bar_ret - position_change * txn_cost
+        trade_pnl = _compute_trade_pnl(position, strat_ret)
+        result["mc_pvalue"]        = monte_carlo_trade_pvalue(trade_pnl, mc_n_permutations)
+        result["mc_n_permutations"] = mc_n_permutations
+
     return result
 
 

rdagent/components/coder/factor_coder/auto_fixer.py

@@ -51,13 +51,23 @@ class FactorAutoFixer:
         self.fixes_applied = []
         fixed_code = code
 
-        # Apply fixes in order - groupby fixes MUST come before min_periods fixes
+        # Apply fixes in order
+        # NOTE: _fix_min_periods is intentionally excluded — it increased min_periods to
+        # match window size, which causes all-NaN output for intraday data with 96 bars/day
+        # (window=240 > 96 means zero valid bars per day). The LLM sets its own min_periods.
         fix_methods = [
-            self._fix_groupby_apply_to_transform,  # First: fix groupby patterns
-            self._fix_min_periods,                  # Second: fix min_periods in resulting rolling calls
-            self._fix_inf_nan_handling,             # Third: add inf/nan handling
-            self._fix_data_range_processing,        # Fourth: ensure full data range
-            self._fix_multiindex_groupby,           # Fifth: ensure groupby on MultiIndex
+            self._fix_instrument_column_access,     # First: fix df['instrument'] on MultiIndex
+            self._fix_instrument_loc_multiindex,    # Second: fix df.loc[instrument_var] on MultiIndex
+            self._fix_zero_volume_proxy,            # Third: replace zero $volume with range proxy
+            self._fix_reset_index_groupby,          # Fourth: fix groupby(level=N) after reset_index()
+            self._fix_groupby_mixed_levels,         # Fifth: fix groupby(level=[int, str])
+            self._fix_groupby_column_on_multiindex, # Sixth: fix groupby(['instrument','date']) on MultiIndex
+            self._fix_chained_groupby,              # Seventh: fix groupby(level=N).groupby('date') chain
+            self._fix_rolling_ddof,                 # Eighth: remove unsupported ddof kwarg
+            self._fix_groupby_apply_to_transform,   # Ninth: fix groupby patterns
+            self._fix_inf_nan_handling,             # Tenth: add inf/nan handling
+            self._fix_data_range_processing,        # Eleventh: ensure full data range
+            self._fix_multiindex_groupby,           # Twelfth: ensure groupby on MultiIndex
         ]
 
         for fix_method in fix_methods:
@@ -75,6 +85,352 @@ class FactorAutoFixer:
 
         return fixed_code
 
+    def _fix_instrument_column_access(self, code: str) -> str:
+        """
+        Fix: df['instrument'] raises KeyError on a MultiIndex DataFrame because
+        'instrument' is an index level (level 1), not a column.
+
+        Replace df['instrument'] with df.index.get_level_values('instrument')
+        but only when the DataFrame has a MultiIndex (not after reset_index which
+        would have promoted it to a real column).
+
+        Also fixes df.reset_index()['instrument'] correctly since after reset_index
+        the column exists.
+        """
+        fixed_code = code
+
+        # Skip if already fixed or if reset_index() is being used before the access
+        # We only fix bare df['instrument'] where df is the original MultiIndex frame.
+        # Heuristic: if the assignment lhs or context shows reset_index, leave it alone.
+
+        # Pattern: <varname>['instrument'] where varname is NOT a reset_index result
+        reset_vars = set(re.findall(r'(\w+)\s*=\s*\w[^=\n]*\.reset_index\(', fixed_code))
+
+        def _replace_instrument_access(m: re.Match) -> str:
+            var = m.group(1)
+            if var in reset_vars:
+                return m.group(0)  # leave reset_index vars alone — column exists
+            self.fixes_applied.append(f"instrument_column: {var}['instrument'] → get_level_values(1)")
+            return f"{var}.index.get_level_values(1)"
+
+        # Exclude assignment targets: var['instrument'] = ... must not become
+        # var.index.get_level_values(1) = ... (SyntaxError: cannot assign to function call)
+        fixed_code = re.sub(r"(\w+)\['instrument'\](?!\s*=)", _replace_instrument_access, fixed_code)
+
+        return fixed_code
+
+    def _fix_instrument_loc_multiindex(self, code: str) -> str:
+        """
+        Fix: df.loc[instrument_var] raises DateParseError on a (datetime, instrument)
+        MultiIndex because pandas tries to match the instrument string against the
+        datetime level (level 0).
+
+        Pattern detected: for-loops iterating over get_level_values('instrument') or
+        get_level_values(1) where the loop variable is then used as df.loc[loop_var].
+
+        Replacement: df.loc[instrument_var] → df.xs(instrument_var, level=1)
+        """
+        fixed_code = code
+
+        # Find variables iterated from get_level_values('instrument') or get_level_values(1)
+        inst_vars = set(
+            re.findall(
+                r"for\s+(\w+)\s+in\s+.+?\.get_level_values\s*\(\s*(?:1|['\"]instrument['\"])\s*\)[^:\n]*:",
+                code,
+            )
+        )
+
+        if not inst_vars:
+            return fixed_code
+
+        for var in inst_vars:
+            # Replace DF.loc[var] (read) with DF.xs(var, level=1)
+            # Exclude write-back patterns (DF.loc[var] = ...) — leave those as-is
+            def _make_replacer(v: str):
+                def _replace(m: re.Match) -> str:
+                    df_var = m.group(1)
+                    self.fixes_applied.append(
+                        f"instrument_loc: {df_var}.loc[{v}] → {df_var}.xs({v}, level=1)"
+                    )
+                    return f"{df_var}.xs({v}, level=1)"
+
+                return _replace
+
+            # Only match when NOT followed by ' =' (assignment)
+            fixed_code = re.sub(
+                rf"(\w+)\.loc\[\s*{re.escape(var)}\s*\](?!\s*=)",
+                _make_replacer(var),
+                fixed_code,
+            )
+
+        return fixed_code
+
+    def _fix_zero_volume_proxy(self, code: str) -> str:
+        """
+        Fix: $volume is always 0 in our EUR/USD dataset (FX has no real volume).
+        Any factor using $volume (VWAP, volume-weighted returns, etc.) produces
+        all-NaN output because 0*price=0 and sum(0)/sum(0)=NaN.
+
+        Insert a guard right after pd.read_hdf() that replaces zero volume with
+        the intraday price-range proxy ($high - $low) so volume-weighted factors
+        produce meaningful signals.
+        """
+        if "'$volume'" not in code and '"$volume"' not in code:
+            return code
+
+        # Already patched
+        if "volume proxy" in code:
+            return code
+
+        lines = code.splitlines()
+        insert_after = -1
+        df_var = "df"
+        indent = "    "
+
+        for i, line in enumerate(lines):
+            if "read_hdf(" in line:
+                m = re.match(r"(\s*)(\w+)\s*=\s*", line)
+                if m:
+                    indent = m.group(1)
+                    df_var = m.group(2)
+                else:
+                    m2 = re.match(r"(\s*)", line)
+                    indent = m2.group(1) if m2 else "    "
+                insert_after = i
+                break
+
+        if insert_after == -1:
+            return code
+
+        proxy_lines = [
+            f"{indent}# volume proxy: $volume is always 0 in FX data — use price-range as proxy",
+            f"{indent}if ({df_var}['$volume'] == 0).all():",
+            f"{indent}    {df_var}['$volume'] = {df_var}['$high'] - {df_var}['$low']",
+        ]
+        lines = lines[: insert_after + 1] + proxy_lines + lines[insert_after + 1 :]
+        self.fixes_applied.append("volume_proxy: replaced zero $volume with ($high - $low)")
+        return "\n".join(lines)
+
+    def _fix_reset_index_groupby(self, code: str) -> str:
+        """
+        Fix: groupby(level=N) on a variable created by .reset_index() fails because
+        reset_index() converts the MultiIndex into regular columns, leaving a plain
+        RangeIndex.  Replace groupby(level=N) on such variables with
+        groupby('instrument').
+
+        Detected pattern:
+            varname = <anything>.reset_index(...)
+            ...
+            varname.groupby(level=0|1)
+        """
+        fixed_code = code
+
+        # Find all variables assigned via reset_index()
+        reset_vars = set(re.findall(r'(\w+)\s*=\s*\w[^=\n]*\.reset_index\(', fixed_code))
+
+        for var in reset_vars:
+            # Replace var.groupby(level=N) with var.groupby('instrument')
+            pattern = rf'{re.escape(var)}\.groupby\(level\s*=\s*\d+\)'
+            if re.search(pattern, fixed_code):
+                fixed_code = re.sub(pattern, f"{var}.groupby('instrument')", fixed_code)
+                self.fixes_applied.append(f"reset_index_groupby: {var}.groupby(level=N) → groupby('instrument')")
+
+        return fixed_code
+
+    def _fix_groupby_mixed_levels(self, code: str) -> str:
+        """
+        Fix: groupby(level=[int, 'str']) raises AssertionError because string level
+        names don't exist on an unnamed MultiIndex.  Keep only integer levels.
+
+        Pattern:  .groupby(level=[0, 'date'])  →  .groupby(level=0)
+                  .groupby(level=[1, 'date'])  →  .groupby(level=1)
+        """
+        fixed_code = code
+
+        def _keep_int_levels(m):
+            inner = m.group(1)
+            ints = re.findall(r'\b(\d+)\b', inner)
+            if not ints:
+                return m.group(0)
+            replacement = f'.groupby(level={ints[0]})' if len(ints) == 1 else f'.groupby(level=[{", ".join(ints)}])'
+            self.fixes_applied.append(f"mixed_levels: groupby(level=[...,str]) → {replacement}")
+            return replacement
+
+        fixed_code = re.sub(r'\.groupby\(level=\[([^\]]+)\]\)', _keep_int_levels, fixed_code)
+        return fixed_code
+
+    def _fix_groupby_column_on_multiindex(self, code: str) -> str:
+        """
+        Fix: groupby(['instrument', 'date']) on a MultiIndex (datetime, instrument)
+        DataFrame fails with KeyError because those are index levels, not columns.
+
+        Correct replacement preserves BOTH dimensions so intraday calculations reset
+        per day:
+            var.groupby(['instrument', 'date'])
+            → var.groupby([var.index.get_level_values(1), var.index.get_level_values(0).normalize()])
+
+        Single-column groupby(['instrument']) is correctly replaced with groupby(level=1).
+        Note: do NOT convert groupby('instrument') → groupby(level=1) here — that would
+        undo the reset_index_groupby fix which correctly emits groupby('instrument').
+        """
+        fixed_code = code
+
+        # Variables created via reset_index() have a plain RangeIndex — applying
+        # get_level_values() on them would raise AttributeError. Skip those.
+        reset_vars = set(re.findall(r'(\w+)\s*=\s*\w[^=\n]*\.reset_index\(', fixed_code))
+
+        def _replace_two_col_groupby(m: re.Match, order: str) -> str:
+            var = m.group(1)
+            if var in reset_vars:
+                return m.group(0)  # leave reset_index vars alone — RangeIndex, not MultiIndex
+            if order == "instrument_date":
+                repl = (
+                    f"{var}.groupby([{var}.index.get_level_values(1), "
+                    f"{var}.index.get_level_values(0).normalize()])"
+                )
+            else:  # date_instrument
+                repl = (
+                    f"{var}.groupby([{var}.index.get_level_values(0).normalize(), "
+                    f"{var}.index.get_level_values(1)])"
+                )
+            self.fixes_applied.append(f"multiindex_groupby: {m.group(0)[:60]} → two-level")
+            return repl
+
+        # groupby(['instrument', 'date']) — capture variable name before .groupby
+        fixed_code = re.sub(
+            r'(\w+)\.groupby\(\[\'instrument\',\s*\'date\'\]\)',
+            lambda m: _replace_two_col_groupby(m, "instrument_date"),
+            fixed_code,
+        )
+        # groupby(['date', 'instrument'])
+        fixed_code = re.sub(
+            r'(\w+)\.groupby\(\[\'date\',\s*\'instrument\'\]\)',
+            lambda m: _replace_two_col_groupby(m, "date_instrument"),
+            fixed_code,
+        )
+        # single: groupby(['instrument']) → groupby(level=1), but not on reset_index vars
+        def _replace_single_instrument_groupby(m: re.Match) -> str:
+            # Look backwards to find the variable name
+            prefix = fixed_code[: m.start()]
+            var_match = re.search(r'(\w+)\s*$', prefix)
+            var = var_match.group(1) if var_match else ''
+            if var in reset_vars:
+                return m.group(0)
+            self.fixes_applied.append("multiindex_groupby: groupby(['instrument']) → groupby(level=1)")
+            return ".groupby(level=1)"
+
+        if re.search(r"\.groupby\(\['instrument'\]\)", fixed_code):
+            fixed_code = re.sub(r"\.groupby\(\['instrument'\]\)", _replace_single_instrument_groupby, fixed_code)
+
+        # groupby(level=['instrument', 'date']) — uses level= keyword with string names.
+        # 'date' is NOT a valid level name in our (datetime, instrument) MultiIndex;
+        # replace with get_level_values to normalize datetime to daily timestamps.
+        fixed_code = re.sub(
+            r"(\w+)\.groupby\(level=\['instrument',\s*'date'\]\)",
+            lambda m: (
+                self.fixes_applied.append(
+                    f"multiindex_groupby: {m.group(0)[:60]} → two-level get_level_values"
+                )
+                or f"{m.group(1)}.groupby([{m.group(1)}.index.get_level_values(1), "
+                f"{m.group(1)}.index.get_level_values(0).normalize()])"
+            ),
+            fixed_code,
+        )
+        # groupby(level=['date', 'instrument'])
+        fixed_code = re.sub(
+            r"(\w+)\.groupby\(level=\['date',\s*'instrument'\]\)",
+            lambda m: (
+                self.fixes_applied.append(
+                    f"multiindex_groupby: {m.group(0)[:60]} → two-level get_level_values"
+                )
+                or f"{m.group(1)}.groupby([{m.group(1)}.index.get_level_values(0).normalize(), "
+                f"{m.group(1)}.index.get_level_values(1)])"
+            ),
+            fixed_code,
+        )
+        # single: groupby(level=['instrument']) → groupby(level=1)
+        fixed_code = re.sub(
+            r"\.groupby\(level=\['instrument'\]\)",
+            lambda m: (self.fixes_applied.append("multiindex_groupby: groupby(level=['instrument']) → level=1") or ".groupby(level=1)"),
+            fixed_code,
+        )
+
+        return fixed_code
+
+    def _fix_chained_groupby(self, code: str) -> str:
+        """
+        Fix two broken patterns the LLM generates when trying to group by (instrument, date):
+
+        Pattern A — chained groupby (runtime AttributeError):
+            var.groupby(level=1).groupby('date')
+            → var.groupby([var.index.get_level_values(1),
+                           var.index.get_level_values(0).normalize()])
+
+        Pattern B — keyword arg inside list (SyntaxError):
+            var.groupby([level=1, 'date'])
+            → same two-level replacement
+        """
+        fixed_code = code
+
+        def _two_level(var: str, tag: str) -> str:
+            self.fixes_applied.append(f"chained_groupby: {tag} → two-level")
+            return (
+                f"{var}.groupby([{var}.index.get_level_values(1), "
+                f"{var}.index.get_level_values(0).normalize()])"
+            )
+
+        # Pattern A: var.groupby(level=N).groupby('date')
+        fixed_code = re.sub(
+            r'(\w+)\.groupby\(level=\d+\)\.groupby\(["\']date["\']\)',
+            lambda m: _two_level(m.group(1), m.group(0)[:60]),
+            fixed_code,
+        )
+
+        # Pattern B: .groupby([level=N, 'date']) — SyntaxError in Python.
+        # The variable before .groupby may be complex (e.g. df[mask]) so we don't
+        # try to capture it; we use df as the index reference (always correct since
+        # all filtered frames share df's MultiIndex structure).
+        def _two_level_df(tag: str) -> str:
+            self.fixes_applied.append(f"chained_groupby: {tag} → two-level")
+            return ".groupby([df.index.get_level_values(1), df.index.get_level_values(0).normalize()])"
+
+        fixed_code = re.sub(
+            r'\.groupby\(\[\s*level\s*=\s*\d+\s*,\s*["\']?date["\']?\s*\]\)',
+            lambda m: _two_level_df(m.group(0)[:60]),
+            fixed_code,
+        )
+        # Also handle reversed order: ['date', level=N]
+        fixed_code = re.sub(
+            r'\.groupby\(\[\s*["\']?date["\']?\s*,\s*level\s*=\s*\d+\s*\]\)',
+            lambda m: _two_level_df(m.group(0)[:60]),
+            fixed_code,
+        )
+
+        return fixed_code
+
+    def _fix_rolling_ddof(self, code: str) -> str:
+        """
+        Fix: pandas rolling() does not accept a ddof kwarg — raises TypeError.
+        Remove ddof from both rolling(..., ddof=N) and rolling(...).std(ddof=N).
+        """
+        fixed_code = code
+
+        # Form 1: ddof inside rolling() — .rolling(window=N, min_periods=M, ddof=K)
+        def _strip_ddof_from_rolling(m):
+            inner = re.sub(r',?\s*ddof\s*=\s*\d+', '', m.group(1))
+            inner = inner.strip(', ')
+            self.fixes_applied.append("rolling_ddof: removed ddof from rolling()")
+            return f'.rolling({inner})'
+
+        fixed_code = re.sub(r'\.rolling\(([^)]*ddof\s*=\s*\d+[^)]*)\)', _strip_ddof_from_rolling, fixed_code)
+
+        # Form 2: ddof inside .std() / .var() — .std(ddof=N)
+        if re.search(r'\.(std|var)\([^)]*ddof\s*=\s*\d+', fixed_code):
+            fixed_code = re.sub(r'\.(std|var)\([^)]*ddof\s*=\s*\d+[^)]*\)', r'.\1()', fixed_code)
+            self.fixes_applied.append("rolling_ddof: removed ddof from std()/var()")
+
+        return fixed_code
+
     def _fix_min_periods(self, code: str) -> str:
         """
         Fix: Ensure min_periods matches window size in rolling calculations.
@@ -325,6 +681,45 @@ class FactorAutoFixer:
             fixed_code = fixed_code.replace(old_code, new_code)
             self.fixes_applied.append(f"groupby: fixed rolling correlation (window={window}) with reset_index")
 
+        # === GENERAL FIX: DF.groupby(level=N)['col'].apply(lambda x: EXPR) ===
+        # apply() on a grouped Series returns a MultiIndex result (extra level prepended),
+        # causing index shape mismatch when assigned back to df['col'].
+        # Replace with transform() which preserves the original index.
+        col_apply_pattern = re.compile(
+            r"(\w+)\.groupby\(level=(\d+)\)\['([^']+)'\]\.apply\((\s*lambda\s+\w+\s*:.*?)\)",
+            re.DOTALL,
+        )
+        for m in list(col_apply_pattern.finditer(fixed_code)):
+            full = m.group(0)
+            df_var = m.group(1)
+            level = m.group(2)
+            col = m.group(3)
+            lam = m.group(4).strip()
+            new_expr = f"{df_var}.groupby(level={level})['{col}'].transform({lam})"
+            fixed_code = fixed_code.replace(full, new_expr, 1)
+            self.fixes_applied.append(
+                f"groupby: {df_var}.groupby(level={level})['{col}'].apply() → transform()"
+            )
+
+        # === FIX: .transform(...).reset_index(level=N, drop=True) ===
+        # transform() already returns the same index as the input — adding reset_index()
+        # after it drops an index level and causes ValueError on assignment back to df['col'].
+        # Detected line-by-line: if a line contains both .transform( and .reset_index(level=
+        reset_suffix = re.compile(r'\s*\.reset_index\s*\(\s*level\s*=[^,)]+,\s*drop\s*=\s*True\s*\)\s*$')
+        new_lines = []
+        changed = False
+        for line in fixed_code.splitlines():
+            if '.transform(' in line and '.reset_index(' in line:
+                cleaned = reset_suffix.sub('', line)
+                if cleaned != line:
+                    new_lines.append(cleaned)
+                    changed = True
+                    continue
+            new_lines.append(line)
+        if changed:
+            fixed_code = '\n'.join(new_lines)
+            self.fixes_applied.append("groupby: removed spurious .reset_index() after .transform()")
+
         # Pattern: Simple groupby().apply() with rolling().method()
         # df.groupby(level=N).apply(lambda x: x['col'].rolling(...).method())
         apply_pattern = r"df\.groupby\(level=(\d+)\)\.apply\(\s*lambda\s+x:\s+x\['([^']+)'\]\.rolling\([^)]+\)\.(\w+)\([^)]*\)\s*\)"

rdagent/components/coder/factor_coder/factor.py

@@ -161,8 +161,7 @@ class FactorFBWorkspace(FBWorkspace):
 
             try:
                 subprocess.check_output(
-                    f"{FACTOR_COSTEER_SETTINGS.python_bin} {execution_code_path}",
-                    shell=True,
+                    [FACTOR_COSTEER_SETTINGS.python_bin, str(execution_code_path)],
                     cwd=self.workspace_path,
                     stderr=subprocess.STDOUT,
                     timeout=FACTOR_COSTEER_SETTINGS.file_based_execution_timeout,

rdagent/components/coder/factor_coder/prompts.yaml

@@ -53,7 +53,7 @@ evolving_strategy_factor_implementation_v1_system: |-
   - ALWAYS use `min_periods=N` where N equals the window size in rolling calculations (e.g., `.rolling(20, min_periods=20)`)
   - ALWAYS handle infinite values after division: `.replace([np.inf, -np.inf], np.nan)` before saving results
   - ALWAYS use `groupby(level=1)` or `groupby('instrument')` before rolling operations on MultiIndex dataframes
-  - Process the COMPLETE date range (2020-2026), do NOT filter by date
+  - Process the COMPLETE date range available in the HDF5 file (do NOT filter by date — the file may contain 2024 debug data or full 2020-2026 data)
   - Use `groupby().transform()` instead of `groupby().apply()` for single-column assignments
 
   Notice that you should not add any other text before or after the json format.

rdagent/components/coder/model_coder/benchmark/gt_code/A-DGN.py

@@ -123,8 +123,8 @@ model_cls = AntiSymmetricConv
 
 
 if __name__ == "__main__":
-    node_features = torch.load("node_features.pt")
-    edge_index = torch.load("edge_index.pt")
+    node_features = torch.load("node_features.pt", weights_only=True)
+    edge_index = torch.load("edge_index.pt", weights_only=True)
 
     # Model instantiation and forward pass
     model = AntiSymmetricConv(in_channels=node_features.size(-1))

rdagent/components/coder/model_coder/benchmark/gt_code/dirgnn.py

@@ -78,8 +78,8 @@ model_cls = DirGNNConv
 
 
 if __name__ == "__main__":
-    node_features = torch.load("node_features.pt")
-    edge_index = torch.load("edge_index.pt")
+    node_features = torch.load("node_features.pt", weights_only=True)
+    edge_index = torch.load("edge_index.pt", weights_only=True)
 
     # Model instantiation and forward pass
     model = DirGNNConv(MessagePassing())

rdagent/components/coder/model_coder/benchmark/gt_code/gpsconv.py

@@ -187,8 +187,8 @@ model_cls = GPSConv
 
 
 if __name__ == "__main__":
-    node_features = torch.load("node_features.pt")
-    edge_index = torch.load("edge_index.pt")
+    node_features = torch.load("node_features.pt", weights_only=True)
+    edge_index = torch.load("edge_index.pt", weights_only=True)
 
     # Model instantiation and forward pass
     model = GPSConv(channels=node_features.size(-1), conv=MessagePassing())

rdagent/components/coder/model_coder/benchmark/gt_code/linkx.py

@@ -170,8 +170,8 @@ class LINKX(torch.nn.Module):
 model_cls = LINKX
 
 if __name__ == "__main__":
-    node_features = torch.load("node_features.pt")
-    edge_index = torch.load("edge_index.pt")
+    node_features = torch.load("node_features.pt", weights_only=True)
+    edge_index = torch.load("edge_index.pt", weights_only=True)
 
     # Model instantiation and forward pass
     model = LINKX(

rdagent/components/coder/model_coder/benchmark/gt_code/pmlp.py

@@ -102,8 +102,8 @@ class PMLP(torch.nn.Module):
 model_cls = PMLP
 
 if __name__ == "__main__":
-    node_features = torch.load("node_features.pt")
-    edge_index = torch.load("edge_index.pt")
+    node_features = torch.load("node_features.pt", weights_only=True)
+    edge_index = torch.load("edge_index.pt", weights_only=True)
 
     # Model instantiation and forward pass
     model = PMLP(

rdagent/components/coder/model_coder/benchmark/gt_code/visnet.py

@@ -1180,8 +1180,8 @@ model_cls = ViSNet
 
 
 if __name__ == "__main__":
-    node_features = torch.load("node_features.pt")
-    edge_index = torch.load("edge_index.pt")
+    node_features = torch.load("node_features.pt", weights_only=True)
+    edge_index = torch.load("edge_index.pt", weights_only=True)
 
     # Model instantiation and forward pass
     model = ViSNet()

rdagent/components/coder/model_coder/gt_code.py

@@ -125,8 +125,8 @@ class AntiSymmetricConv(torch.nn.Module):
 
 
 if __name__ == "__main__":
-    node_features = torch.load("node_features.pt")
-    edge_index = torch.load("edge_index.pt")
+    node_features = torch.load("node_features.pt", weights_only=True)
+    edge_index = torch.load("edge_index.pt", weights_only=True)
 
     # Model instantiation and forward pass
     model = AntiSymmetricConv(in_channels=node_features.size(-1))

rdagent/log/daily_log.py

@@ -27,6 +27,7 @@ Usage:
 from __future__ import annotations
 
 import json as _json
+import logging
 import sys
 import threading
 from contextlib import contextmanager
@@ -36,21 +37,24 @@ from typing import Any
 
 from loguru import logger as _root
 
-# ── paths ─────────────────────────────────────────────────────────────────────
+# ── paths ─────────────────────────────────────────────────────────────────────────────────
 LOGS_ROOT: Path = Path(__file__).parent.parent.parent / "logs"
 
-# ── format ────────────────────────────────────────────────────────────────────
+# ── format ────────────────────────────────────────────────────────────────────────────────
 _FILE_FMT = (
     "{time:YYYY-MM-DD HH:mm:ss.SSS} | {level: <8} | {extra[cmd]: <18} | {message}"
 )
 
-# ── internal state ─────────────────────────────────────────────────────────────
+# ── internal state ─────────────────────────────────────────────────────────────────────────────
 _registered: set[str] = set()   # command keys that already have a file sink
 _all_added: bool = False         # whether the combined all.log sink is active
 _llm_log_lock = threading.Lock()  # guards concurrent writes to llm_calls.jsonl
 
+# Maximum characters stored per field in llm_calls.jsonl to prevent GB-scale files.
+_LLM_CALL_MAX_CHARS = 500
 
-# ── helpers ───────────────────────────────────────────────────────────────────
+
+# ── helpers ────────────────────────────────────────────────────────────────────────────────
 
 def _today_dir() -> Path:
     d = LOGS_ROOT / datetime.now().strftime("%Y-%m-%d")
@@ -79,7 +83,7 @@ def _banner(log, title: str, meta: dict[str, Any]) -> None:
     log.info(sep)
 
 
-# ── public API ────────────────────────────────────────────────────────────────
+# ── public API ──────────────────────────────────────────────────────────────────────────────
 
 def log_llm_call(
     system: str | None,
@@ -88,16 +92,19 @@ def log_llm_call(
     start_time: Any = None,
     end_time: Any = None,
 ) -> None:
-    """Append one complete LLM call to logs/YYYY-MM-DD/llm_calls.jsonl.
+    """Append one LLM call summary to logs/YYYY-MM-DD/llm_calls.jsonl.
+
+    Prompt/response content is capped at _LLM_CALL_MAX_CHARS to prevent
+    GB-scale log files from long-running loops.
 
     Each line is a self-contained JSON object so the file is grep/jq-friendly:
       jq 'select(.duration_ms > 5000)' logs/2026-04-17/llm_calls.jsonl
     """
     entry: dict[str, Any] = {
         "ts": datetime.now().isoformat(timespec="milliseconds"),
-        "system": system or "",
-        "user": user,
-        "response": response,
+        "system": (system or "")[:_LLM_CALL_MAX_CHARS],
+        "user": user[:_LLM_CALL_MAX_CHARS],
+        "response": response[:_LLM_CALL_MAX_CHARS],
     }
     if start_time is not None and end_time is not None:
         try:
@@ -130,13 +137,13 @@ def setup(command: str, **context: Any):
     key = command.lower()
 
     if key not in _registered:
-        # Per-command rotating file
         _root.add(
             str(log_dir / f"{key}.log"),
             format=_FILE_FMT,
             filter=lambda r, k=key: r["extra"].get("cmd", "").lower() == k,
-            rotation="00:00",      # new file at midnight
-            retention="30 days",
+            rotation="50 MB",
+            compression="gz",
+            retention="7 days",
             encoding="utf-8",
             enqueue=True,
             backtrace=False,
@@ -145,13 +152,13 @@ def setup(command: str, **context: Any):
         _registered.add(key)
 
     if not _all_added:
-        # Combined log — all commands
         _root.add(
             str(log_dir / "all.log"),
             format=_FILE_FMT,
             filter=lambda r: "cmd" in r["extra"],
-            rotation="00:00",
-            retention="60 days",
+            rotation="100 MB",
+            compression="gz",
+            retention="7 days",
             encoding="utf-8",
             enqueue=True,
             backtrace=False,

rdagent/log/ui/llm_st.py

@@ -0,0 +1,306 @@
+import argparse
+import json
+import pickle  # nosec
+import re
+import time
+from pathlib import Path
+
+import streamlit as st
+from streamlit import session_state
+
+from rdagent.log.ui.conf import UI_SETTING
+from rdagent.log.utils import extract_evoid, extract_loopid_func_name
+
+st.set_page_config(layout="wide", page_title="debug_llm", page_icon="🎓", initial_sidebar_state="expanded")
+
+# 获取 log_path 参数
+parser = argparse.ArgumentParser(description="RD-Agent Streamlit App")
+parser.add_argument("--log_dir", type=str, help="Path to the log directory")
+args = parser.parse_args()
+
+
+def get_folders_sorted(log_path):
+    """缓存并返回排序后的文件夹列表，并加入进度打印"""
+    with st.spinner("正在加载文件夹列表..."):
+        folders = sorted(
+            (folder for folder in log_path.iterdir() if folder.is_dir() and list(folder.iterdir())),
+            key=lambda folder: folder.stat().st_mtime,
+            reverse=True,
+        )
+        st.write(f"找到 {len(folders)} 个文件夹")
+    return [folder.name for folder in folders]
+
+
+if UI_SETTING.enable_cache:
+    get_folders_sorted = st.cache_data(get_folders_sorted)
+
+
+# 设置主日志路径
+main_log_path = Path(args.log_dir) if args.log_dir else Path("./log")
+if not main_log_path.exists():
+    st.error(f"Log dir {main_log_path} does not exist!")
+    st.stop()
+
+if "data" not in session_state:
+    session_state.data = []
+if "log_path" not in session_state:
+    session_state.log_path = None
+
+tlist = []
+
+
+def load_data():
+    """加载数据到 session_state 并显示进度"""
+    log_file = main_log_path / session_state.log_path / "debug_llm.pkl"
+    try:
+        with st.spinner(f"正在加载数据文件 {log_file}..."):
+            start_time = time.time()
+            with open(log_file, "rb") as f:
+                session_state.data = pickle.load(f, encoding="utf-8")  # nosec
+            st.success(f"数据加载完成！耗时 {time.time() - start_time:.2f} 秒")
+            st.session_state["current_loop"] = 1
+    except Exception as e:
+        session_state.data = [{"error": str(e)}]
+        st.error(f"加载数据失败: {e}")
+
+
+# UI - Sidebar
+with st.sidebar:
+    st.markdown(":blue[**Log Path**]")
+    manually = st.toggle("Manual Input")
+    if manually:
+        st.text_input("log path", key="log_path", label_visibility="collapsed")
+    else:
+        folders = get_folders_sorted(main_log_path)
+        st.selectbox(f"**Select from {main_log_path.absolute()}**", folders, key="log_path")  # nosec B608 — not SQL, Bandit false positive on "Select" in UI label
+
+    if st.button("Refresh Data"):
+        load_data()
+        st.rerun()
+
+
+# Helper functions
+def show_text(text, lang=None):
+    """显示文本代码块"""
+    if lang:
+        st.code(text, language=lang, wrap_lines=True)
+    elif "\n" in text:
+        st.code(text, language="python", wrap_lines=True)
+    else:
+        st.code(text, language="html", wrap_lines=True)
+
+
+def highlight_prompts_uri(uri):
+    """高亮 URI 的格式"""
+    parts = uri.split(":")
+    return f"**{parts[0]}:**:green[**{parts[1]}**]"
+
+
+# Display Data
+progress_text = st.empty()
+progress_bar = st.progress(0)
+
+# 每页展示一个 Loop
+LOOPS_PER_PAGE = 1
+
+# 获取所有的 Loop ID
+loop_groups = {}
+for i, d in enumerate(session_state.data):
+    tag = d["tag"]
+    loop_id, _ = extract_loopid_func_name(tag)
+    if loop_id:
+        if loop_id not in loop_groups:
+            loop_groups[loop_id] = []
+        loop_groups[loop_id].append(d)
+
+# 按 Loop ID 排序
+sorted_loop_ids = sorted(loop_groups.keys(), key=int)  # 假设 Loop ID 是数字
+total_loops = len(sorted_loop_ids)
+total_pages = total_loops  # 每页展示一个 Loop
+
+
+# simple display
+# FIXME: Delete this simple UI if trace have tag(evo_id & loop_id)
+# with st.sidebar:
+#     start = int(st.text_input("start", 0))
+#     end = int(st.text_input("end", 100))
+# for m in session_state.data[start:end]:
+#     if "tpl" in m["tag"]:
+#         obj = m["obj"]
+#         uri = obj["uri"]
+#         tpl = obj["template"]
+#         cxt = obj["context"]
+#         rd = obj["rendered"]
+#         with st.expander(highlight_prompts_uri(uri), expanded=False, icon="⚙️"):
+#             t1, t2, t3 = st.tabs([":green[**Rendered**]", ":blue[**Template**]", ":orange[**Context**]"])
+#             with t1:
+#                 show_text(rd)
+#             with t2:
+#                 show_text(tpl, lang="django")
+#             with t3:
+#                 st.json(cxt)
+#     if "llm" in m["tag"]:
+#         obj = m["obj"]
+#         system = obj.get("system", None)
+#         user = obj["user"]
+#         resp = obj["resp"]
+#         with st.expander(f"**LLM**", expanded=False, icon="🤖"):
+#             t1, t2, t3 = st.tabs([":green[**Response**]", ":blue[**User**]", ":orange[**System**]"])
+#             with t1:
+#                 try:
+#                     rdict = json.loads(resp)
+#                     if "code" in rdict:
+#                         code = rdict["code"]
+#                         st.markdown(":red[**Code in response dict:**]")
+#                         st.code(code, language="python", wrap_lines=True, line_numbers=True)
+#                         rdict.pop("code")
+#                     elif "spec" in rdict:
+#                         spec = rdict["spec"]
+#                         st.markdown(":red[**Spec in response dict:**]")
+#                         st.markdown(spec)
+#                         rdict.pop("spec")
+#                     else:
+#                         # show model codes
+#                         showed_keys = []
+#                         for k, v in rdict.items():
+#                             if k.startswith("model_") and k.endswith(".py"):
+#                                 st.markdown(f":red[**{k}**]")
+#                                 st.code(v, language="python", wrap_lines=True, line_numbers=True)
+#                                 showed_keys.append(k)
+#                         for k in showed_keys:
+#                             rdict.pop(k)
+#                     st.write(":red[**Other parts (except for the code or spec) in response dict:**]")
+#                     st.json(rdict)
+#                 except:
+#                     st.json(resp)
+#             with t2:
+#                 show_text(user)
+#             with t3:
+#                 show_text(system or "No system prompt available")
+
+
+if total_pages:
+    # 初始化 current_loop
+    if "current_loop" not in st.session_state:
+        st.session_state["current_loop"] = 1
+
+    # Loop 导航按钮
+    col1, col2, col3, col4, col5 = st.sidebar.columns([1.2, 1, 2, 1, 1.2])
+
+    with col1:
+        if st.button("|<"):  # 首页
+            st.session_state["current_loop"] = 1
+    with col2:
+        if st.button("<") and st.session_state["current_loop"] > 1:  # 上一页
+            st.session_state["current_loop"] -= 1
+    with col3:
+        # 下拉列表显示所有 Loop
+        st.session_state["current_loop"] = st.selectbox(
+            "选择 Loop",
+            options=list(range(1, total_loops + 1)),
+            index=st.session_state["current_loop"] - 1,  # 默认选中当前 Loop
+            label_visibility="collapsed",  # 隐藏标签
+        )
+    with col4:
+        if st.button("\>") and st.session_state["current_loop"] < total_loops:  # 下一页
+            st.session_state["current_loop"] += 1
+    with col5:
+        if st.button("\>|"):  # 最后一页
+            st.session_state["current_loop"] = total_loops
+
+    # 获取当前 Loop
+    current_loop = st.session_state["current_loop"]
+
+    # 渲染当前 Loop 数据
+    loop_id = sorted_loop_ids[current_loop - 1]
+    progress_text = st.empty()
+    progress_text.text(f"正在处理 Loop {loop_id}...")
+    progress_bar.progress(current_loop / total_loops, text=f"Loop :green[**{current_loop}**] / {total_loops}")
+
+    # 渲染 Loop Header
+    loop_anchor = f"Loop_{loop_id}"
+    if loop_anchor not in tlist:
+        tlist.append(loop_anchor)
+        st.header(loop_anchor, anchor=loop_anchor, divider="blue")
+
+    # 渲染当前 Loop 的所有数据
+    loop_data = loop_groups[loop_id]
+    for d in loop_data:
+        tag = d["tag"]
+        obj = d["obj"]
+        _, func_name = extract_loopid_func_name(tag)
+        evo_id = extract_evoid(tag)
+
+        func_anchor = f"loop_{loop_id}.{func_name}"
+        if func_anchor not in tlist:
+            tlist.append(func_anchor)
+            st.header(f"in *{func_name}*", anchor=func_anchor, divider="green")
+
+        evo_anchor = f"loop_{loop_id}.evo_step_{evo_id}"
+        if evo_id and evo_anchor not in tlist:
+            tlist.append(evo_anchor)
+            st.subheader(f"evo_step_{evo_id}", anchor=evo_anchor, divider="orange")
+
+        # 根据 tag 渲染内容
+        if "debug_exp_gen" in tag:
+            with st.expander(
+                f"Exp in :violet[**{obj.experiment_workspace.workspace_path}**]", expanded=False, icon="🧩"
+            ):
+                st.write(obj)
+        elif "debug_tpl" in tag:
+            uri = obj["uri"]
+            tpl = obj["template"]
+            cxt = obj["context"]
+            rd = obj["rendered"]
+            with st.expander(highlight_prompts_uri(uri), expanded=False, icon="⚙️"):
+                t1, t2, t3 = st.tabs([":green[**Rendered**]", ":blue[**Template**]", ":orange[**Context**]"])
+                with t1:
+                    show_text(rd)
+                with t2:
+                    show_text(tpl, lang="django")
+                with t3:
+                    st.json(cxt)
+        elif "debug_llm" in tag:
+            system = obj.get("system", None)
+            user = obj["user"]
+            resp = obj["resp"]
+            with st.expander(f"**LLM**", expanded=False, icon="🤖"):
+                t1, t2, t3 = st.tabs([":green[**Response**]", ":blue[**User**]", ":orange[**System**]"])
+                with t1:
+                    try:
+                        rdict = json.loads(resp)
+                        if "code" in rdict:
+                            code = rdict["code"]
+                            st.markdown(":red[**Code in response dict:**]")
+                            st.code(code, language="python", wrap_lines=True, line_numbers=True)
+                            rdict.pop("code")
+                        elif "spec" in rdict:
+                            spec = rdict["spec"]
+                            st.markdown(":red[**Spec in response dict:**]")
+                            st.markdown(spec)
+                            rdict.pop("spec")
+                        else:
+                            # show model codes
+                            showed_keys = []
+                            for k, v in rdict.items():
+                                if k.startswith("model_") and k.endswith(".py"):
+                                    st.markdown(f":red[**{k}**]")
+                                    st.code(v, language="python", wrap_lines=True, line_numbers=True)
+                                    showed_keys.append(k)
+                            for k in showed_keys:
+                                rdict.pop(k)
+                        st.write(":red[**Other parts (except for the code or spec) in response dict:**]")
+                        st.json(rdict)
+                    except:
+                        st.json(resp)
+                with t2:
+                    show_text(user)
+                with t3:
+                    show_text(system or "No system prompt available")
+
+    progress_text.text("当前 Loop 数据处理完成！")
+
+    # Sidebar TOC
+    with st.sidebar:
+        toc = "\n".join([f"- [{t}](#{t})" if t.startswith("L") else f"  - [{t.split('.')[1]}](#{t})" for t in tlist])
+        st.markdown(toc, unsafe_allow_html=True)

rdagent/oai/backend/base.py

@@ -720,7 +720,13 @@ class APIBackend(ABC):
 
             if finish_reason is None or finish_reason != "length":
                 break  # we get a full response now.
-            new_messages.append({"role": "assistant", "content": response})
+            # Merge into the previous assistant message if there already is one at the end.
+            # Appending a second consecutive assistant message causes llama-server to return 400
+            # ("Cannot have 2 or more assistant messages at the end of the list").
+            if new_messages and new_messages[-1]["role"] == "assistant":
+                new_messages[-1]["content"] += response
+            else:
+                new_messages.append({"role": "assistant", "content": response})
         else:
             raise RuntimeError(f"Failed to continue the conversation after {try_n} retries.")
 

rdagent/scenarios/data_science/proposal/exp_gen/merge.py

@@ -182,7 +182,7 @@ class ExpGen2Hypothesis(DSProposalV2ExpGen):
 
         success_fb_list = list(set(trace_fbs))
         logger.info(
-            f"Merge Hypothesis: select {len(success_fb_list)} from {len(trace_fbs)} SOTA experiments found in {len(leaves)} traces"
+            f"Merge Hypothesis: select {len(success_fb_list)} from {len(trace_fbs)} SOTA experiments found in {len(leaves)} traces"  # nosec B608 — not SQL, Bandit false positive on "select" in log message
         )
 
         if len(success_fb_list) > 0:

rdagent/scenarios/data_science/proposal/exp_gen/select/submit.py

@@ -1,3 +1,4 @@
+import ast
 import json
 import os
 import pickle
@@ -587,8 +588,8 @@ def _parsing_score(grade_stdout: str) -> Optional[float]:
         except:
             pass
         try:
-            # Priority 2: Eval dict
-            return float(eval(json_str)["score"])
+            # Priority 2: safe literal eval for Python-style dicts
+            return float(ast.literal_eval(json_str)["score"])
         except:
             pass
         try:

rdagent/scenarios/kaggle/developer/coder.py

@@ -38,7 +38,7 @@ class KGModelFeatureSelectionCoder(Developer[KGModelExperiment]):
         assert target_model_type in KG_SELECT_MAPPING
         if len(exp.experiment_workspace.data_description) == 1:
             code = (
-                Environment(undefined=StrictUndefined)
+                Environment(undefined=StrictUndefined)  # nosec B701 — renders Python code templates, not HTML; autoescape would corrupt code
                 .from_string(DEFAULT_SELECTION_CODE)
                 .render(feature_index_list=None)
             )
@@ -62,7 +62,7 @@ class KGModelFeatureSelectionCoder(Developer[KGModelExperiment]):
             chosen_index_to_list_index = [i - 1 for i in chosen_index]
 
             code = (
-                Environment(undefined=StrictUndefined)
+                Environment(undefined=StrictUndefined)  # nosec B701 — renders Python code templates, not HTML; autoescape would corrupt code
                 .from_string(DEFAULT_SELECTION_CODE)
                 .render(feature_index_list=chosen_index_to_list_index)
             )

rdagent/scenarios/kaggle/experiment/spaceship-titanic_template/fea_share_preprocess.py

@@ -79,12 +79,12 @@ def preprocess_script():
     This method applies the preprocessing steps to the training, validation, and test datasets.
     """
     if os.path.exists("/kaggle/input/X_train.pkl"):
-        X_train = pd.read_pickle("/kaggle/input/X_train.pkl")
-        X_valid = pd.read_pickle("/kaggle/input/X_valid.pkl")
-        y_train = pd.read_pickle("/kaggle/input/y_train.pkl")
-        y_valid = pd.read_pickle("/kaggle/input/y_valid.pkl")
-        X_test = pd.read_pickle("/kaggle/input/X_test.pkl")
-        others = pd.read_pickle("/kaggle/input/others.pkl")
+        X_train = pd.read_pickle("/kaggle/input/X_train.pkl")  # nosec B301 — trusted Kaggle input
+        X_valid = pd.read_pickle("/kaggle/input/X_valid.pkl")  # nosec B301
+        y_train = pd.read_pickle("/kaggle/input/y_train.pkl")  # nosec B301
+        y_valid = pd.read_pickle("/kaggle/input/y_valid.pkl")  # nosec B301
+        X_test = pd.read_pickle("/kaggle/input/X_test.pkl")  # nosec B301
+        others = pd.read_pickle("/kaggle/input/others.pkl")  # nosec B301
         y_train = pd.Series(y_train).reset_index(drop=True)
         y_valid = pd.Series(y_valid).reset_index(drop=True)
 

rdagent/scenarios/kaggle/experiment/templates/meta_tpl_deprecated/fea_share_preprocess.py

@@ -85,12 +85,12 @@ def preprocess_script():
     This method applies the preprocessing steps to the training, validation, and test datasets.
     """
     if os.path.exists("/kaggle/input/X_train.pkl"):
-        X_train = pd.read_pickle("/kaggle/input/X_train.pkl")
-        X_valid = pd.read_pickle("/kaggle/input/X_valid.pkl")
-        y_train = pd.read_pickle("/kaggle/input/y_train.pkl")
-        y_valid = pd.read_pickle("/kaggle/input/y_valid.pkl")
-        X_test = pd.read_pickle("/kaggle/input/X_test.pkl")
-        others = pd.read_pickle("/kaggle/input/others.pkl")
+        X_train = pd.read_pickle("/kaggle/input/X_train.pkl")  # nosec B301
+        X_valid = pd.read_pickle("/kaggle/input/X_valid.pkl")  # nosec B301
+        y_train = pd.read_pickle("/kaggle/input/y_train.pkl")  # nosec B301
+        y_valid = pd.read_pickle("/kaggle/input/y_valid.pkl")  # nosec B301
+        X_test = pd.read_pickle("/kaggle/input/X_test.pkl")  # nosec B301
+        others = pd.read_pickle("/kaggle/input/others.pkl")  # nosec B301
 
         return X_train, X_valid, y_train, y_valid, X_test, *others
     X_train, X_valid, y_train, y_valid = prepreprocess()

rdagent/scenarios/kaggle/experiment/templates/playground-series-s3e26/fea_share_preprocess.py

@@ -82,11 +82,11 @@ def preprocess_script():
     This method applies the preprocessing steps to the training, validation, and test datasets.
     """
     if os.path.exists("X_train.pkl"):
-        X_train = pd.read_pickle("X_train.pkl")
-        X_valid = pd.read_pickle("X_valid.pkl")
-        y_train = pd.read_pickle("y_train.pkl")
-        y_valid = pd.read_pickle("y_valid.pkl")
-        X_test = pd.read_pickle("X_test.pkl")
+        X_train = pd.read_pickle("X_train.pkl")  # nosec B301
+        X_valid = pd.read_pickle("X_valid.pkl")  # nosec B301
+        y_train = pd.read_pickle("y_train.pkl")  # nosec B301
+        y_valid = pd.read_pickle("y_valid.pkl")  # nosec B301
+        X_test = pd.read_pickle("X_test.pkl")  # nosec B301
         return X_train, X_valid, y_train, y_valid, X_test
 
     X_train, X_valid, y_train, y_valid, test, status_encoder, test_ids = prepreprocess()

rdagent/scenarios/kaggle/experiment/templates/playground-series-s4e8/fea_share_preprocess.py

@@ -73,12 +73,12 @@ def preprocess_script():
     This method applies the preprocessing steps to the training, validation, and test datasets.
     """
     if os.path.exists("/kaggle/input/X_train.pkl"):
-        X_train = pd.read_pickle("/kaggle/input/X_train.pkl")
-        X_valid = pd.read_pickle("/kaggle/input/X_valid.pkl")
-        y_train = pd.read_pickle("/kaggle/input/y_train.pkl")
-        y_valid = pd.read_pickle("/kaggle/input/y_valid.pkl")
-        X_test = pd.read_pickle("/kaggle/input/X_test.pkl")
-        others = pd.read_pickle("/kaggle/input/others.pkl")
+        X_train = pd.read_pickle("/kaggle/input/X_train.pkl")  # nosec B301
+        X_valid = pd.read_pickle("/kaggle/input/X_valid.pkl")  # nosec B301
+        y_train = pd.read_pickle("/kaggle/input/y_train.pkl")  # nosec B301
+        y_valid = pd.read_pickle("/kaggle/input/y_valid.pkl")  # nosec B301
+        X_test = pd.read_pickle("/kaggle/input/X_test.pkl")  # nosec B301
+        others = pd.read_pickle("/kaggle/input/others.pkl")  # nosec B301
         y_train = pd.Series(y_train).reset_index(drop=True)
         y_valid = pd.Series(y_valid).reset_index(drop=True)
 

rdagent/scenarios/qlib/developer/factor_runner.py

@@ -1,4 +1,6 @@
+import sys
 import os
+import logging
 from pathlib import Path
 """
 Qlib Factor Runner - Executes factor backtests in Docker.
@@ -29,6 +31,83 @@ from rdagent.scenarios.qlib.experiment.model_experiment import QlibModelExperime
 DIRNAME = Path(__file__).absolute().resolve().parent
 DIRNAME_local = Path.cwd()
 
+
+def _shift_daily_constant_factor_if_needed(factor_col: "pd.Series", factor_name: str) -> "pd.Series":
+    """Detect and fix look-ahead bias in daily-constant factors.
+
+    A factor is "daily-constant" when every minute bar within the same calendar
+    day carries an identical value. This happens when LLM code computes a daily
+    aggregate (e.g. today's log return) and forward-fills it across all intraday
+    bars without shifting — meaning the end-of-day value is visible at 00:00.
+
+    Fix: shift by one trading day so that the value assigned to day T is the
+    aggregate computed from day T-1, eliminating the forward-looking information.
+    """
+    import numpy as np
+
+    try:
+        notnull = factor_col.dropna()
+        if len(notnull) < 200:
+            return factor_col
+
+        datetimes = notnull.index.get_level_values("datetime")
+        dates = datetimes.normalize()
+
+        # Sample up to 50 random days and check intra-day uniqueness
+        unique_dates = pd.Series(dates.unique())
+        sample_dates = unique_dates.sample(min(50, len(unique_dates)), random_state=42)
+
+        daily_unique_counts = []
+        for d in sample_dates:
+            mask = dates == d
+            vals = notnull.values[mask]
+            if len(vals) > 1:
+                daily_unique_counts.append(len(np.unique(vals[~np.isnan(vals)])))
+
+        if not daily_unique_counts:
+            return factor_col
+
+        # If >90% of sampled days have exactly 1 unique value → daily-constant
+        fraction_constant = sum(1 for c in daily_unique_counts if c == 1) / len(daily_unique_counts)
+        if fraction_constant < 0.90:
+            return factor_col  # Intraday factor — no shift needed
+
+        logger.warning(
+            f"[LookAheadFix] Factor '{factor_name}' is daily-constant "
+            f"({fraction_constant:.0%} of days). Applying 1-day shift to remove look-ahead bias."
+        )
+
+        # Shift: for each instrument, map daily values forward by 1 trading day
+        instruments = factor_col.index.get_level_values("instrument").unique()
+        shifted_parts = []
+        for inst in instruments:
+            inst_series = factor_col.xs(inst, level="instrument")
+            # Get one value per calendar day (the first non-null bar)
+            inst_dt = inst_series.index.normalize()
+            daily_vals = inst_series.groupby(inst_dt).first()
+            # Shift by 1 day
+            daily_vals_shifted = daily_vals.shift(1)
+            # Forward-fill back to minute bars
+            minute_idx = inst_series.index
+            minute_dates = minute_idx.normalize()
+            shifted_minute = minute_dates.map(daily_vals_shifted)
+            shifted_s = pd.Series(
+                shifted_minute.values,
+                index=pd.MultiIndex.from_arrays(
+                    [inst_series.index, [inst] * len(inst_series)],
+                    names=["datetime", "instrument"],
+                ),
+                name=factor_col.name,
+            )
+            shifted_parts.append(shifted_s)
+
+        return pd.concat(shifted_parts).sort_index()
+
+    except Exception as e:
+        logger.debug(f"[LookAheadFix] Could not apply daily shift for '{factor_name}': {e}")
+        return factor_col
+
+
 # TODO: supporting multiprocessing and keep previous results
 
 
@@ -391,8 +470,19 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
         import numpy as np
 
         try:
-            # Get workspace path
-            workspace_path = exp.experiment_workspace.workspace_path
+            # Get workspace path — factor code and result.h5 live in sub_workspace_list[0],
+            # not in experiment_workspace (which is the Qlib template workspace).
+            workspace_path = None
+            if exp.sub_workspace_list:
+                for ws in exp.sub_workspace_list:
+                    if ws is not None and hasattr(ws, 'workspace_path'):
+                        candidate = ws.workspace_path / "result.h5"
+                        if candidate.exists():
+                            workspace_path = ws.workspace_path
+                            break
+            if workspace_path is None:
+                # Fallback to experiment_workspace
+                workspace_path = exp.experiment_workspace.workspace_path
             if workspace_path is None:
                 return None
 
@@ -409,6 +499,12 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
             factor_col = factor_values.iloc[:, 0]
             factor_name = factor_values.columns[0]
 
+            # Detect and fix look-ahead bias in daily-constant factors.
+            # If a factor has the same value for all minute bars within each calendar day
+            # it was computed from same-day data (e.g. today's close return at 00:00).
+            # Fix: shift by 1 trading day so value at day T = aggregate of day T-1.
+            factor_col = _shift_daily_constant_factor_if_needed(factor_col, factor_name)
+
             # Load source data for forward returns
             data_path = (
                 Path(__file__).parent.parent.parent.parent.parent
@@ -587,10 +683,12 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
             from pathlib import Path
             from rdagent.components.backtesting import ResultsDatabase
 
-            # Get factor name from hypothesis
+            # Get factor name: prefer hypothesis, fallback to result Series 'factor_name' key
             factor_name = "unknown"
             if hasattr(exp, 'hypothesis') and exp.hypothesis is not None:
                 factor_name = getattr(exp.hypothesis, 'hypothesis', 'unknown')
+            if factor_name == 'unknown' and isinstance(result, pd.Series) and 'factor_name' in result.index:
+                factor_name = str(result['factor_name'])
 
             # Check if already rejected by protection
             if getattr(exp, 'rejected_by_protection', False):
@@ -824,41 +922,74 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
         """
         Save factor time-series values as parquet for strategy building.
 
-        This is essential for walk-forward validation and strategy combination.
-
-        Parameters
-        ----------
-        factor_name : str
-            Name of the factor
-        exp : QlibFactorExperiment
-            The experiment with factor values
+        Reruns the factor code on the FULL 6-year dataset so the parquet covers
+        the complete backtest range (not just the debug 2024 subset).
         """
         import os as _os
+        import subprocess
+        import shutil
+        import tempfile
 
         try:
-            # Get workspace path
-            workspace_path = exp.experiment_workspace.workspace_path
+            # factor.py lives in sub_workspace_list[0], not experiment_workspace
+            workspace_path = None
+            if exp.sub_workspace_list:
+                for ws in exp.sub_workspace_list:
+                    if ws is not None and hasattr(ws, 'workspace_path'):
+                        fp = ws.workspace_path / "factor.py"
+                        if fp.exists():
+                            workspace_path = ws.workspace_path
+                            break
+            if workspace_path is None:
+                workspace_path = exp.experiment_workspace.workspace_path
             if workspace_path is None:
                 return
 
-            result_h5 = workspace_path / "result.h5"
-            if not result_h5.exists():
+            factor_py = workspace_path / "factor.py"
+            if not factor_py.exists():
                 return
 
-            # Read factor values
+            project_root = Path(__file__).parent.parent.parent.parent.parent
+            full_data = (
+                project_root
+                / "git_ignore_folder"
+                / "factor_implementation_source_data"
+                / "intraday_pv.h5"
+            )
+            if not full_data.exists():
+                return
+
+            # Run factor code on full data in a temp workspace
             import pandas as pd
-            df = pd.read_hdf(str(result_h5), key="data")
+            with tempfile.TemporaryDirectory(prefix="predix_fullval_") as tmp_dir:
+                tmp = Path(tmp_dir)
+                shutil.copy(str(factor_py), str(tmp / "factor.py"))
+                shutil.copy(str(full_data), str(tmp / "intraday_pv.h5"))
+
+                ret = subprocess.run(
+                    ["sys.executable", "factor.py"],
+                    cwd=str(tmp),
+                    capture_output=True,
+                    timeout=300,
+                )
+                if ret.returncode != 0:
+                    # Fall back to debug-data result if full-data run fails
+                    result_h5 = workspace_path / "result.h5"
+                    if not result_h5.exists():
+                        return
+                    df = pd.read_hdf(str(result_h5), key="data")
+                else:
+                    result_h5_full = tmp / "result.h5"
+                    if not result_h5_full.exists():
+                        return
+                    df = pd.read_hdf(str(result_h5_full), key="data")
+
             if df is None or df.empty:
                 return
 
-            # Get the factor series (first column)
             series = df.iloc[:, 0]
             series.name = factor_name
 
-            # Save to results/factors/values/
-            project_root = Path(__file__).parent.parent.parent.parent.parent
-
-            # Parallel run isolation
             parallel_run_id = _os.getenv("PARALLEL_RUN_ID", "0")
             if parallel_run_id != "0":
                 values_dir = project_root / "results" / "runs" / f"run{parallel_run_id}" / "factors" / "values"
@@ -866,17 +997,12 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
                 values_dir = project_root / "results" / "factors" / "values"
 
             values_dir.mkdir(parents=True, exist_ok=True)
-
-            # Safe filename
             safe_name = factor_name.replace("/", "_").replace("\\", "_").replace(" ", "_")[:100]
             parquet_path = values_dir / f"{safe_name}.parquet"
+            series.to_frame().to_parquet(str(parquet_path))
 
-            # Save as parquet (with datetime index)
-            series.to_parquet(str(parquet_path))
-
-        except Exception as e:
-            # Don't let factor value saving break the main workflow
-            pass
+        except Exception:
+    logging.debug("Error in save_factor_values_to_parquet", exc_info=True)
 
     def _log_result_warnings(self, factor_name: str, result, metrics: dict) -> None:
         """

rdagent/scenarios/qlib/experiment/factor_data_template/README.md

@@ -23,14 +23,25 @@ $low: low price at 1-minute bar.
 $volume: volume at 1-minute bar (tick volume for FX).
 
 ## Important Notes for 1min Data
-- 96 bars = 1 trading day (24 hours for FX)
+- 1 bar = 1 minute (confirmed)
 - 16 bars = 16 minutes
-- 4 bars = 4 minutes
-- 1 bar = 1 minute
+- 60 bars = 1 hour
+- ~1440 bars = 1 full trading day (FX trades nearly 24h, Mon 00:00 - Fri 22:00 UTC approx.)
+- Typical bars per calendar day: ~1200-1440 (varies by weekday, holidays have fewer)
+- Do NOT assume 96 bars/day — the actual count depends on the date
 - Data range: 2020-01-01 to 2026-03-20
 - Instrument: EURUSD
 - Timezone: UTC
 
+## IMPORTANT: Bars per Day Correction
+The dataset has approximately 1440 bars per full trading day (1 bar = 1 minute, ~24h of FX trading).
+Some older documentation incorrectly stated "96 bars = 1 day" — this is WRONG. Always use:
+- 60 bars = 1 hour
+- 480 bars = 8 hours (London session 08:00-16:00 UTC)
+- 180 bars = 3 hours (London/NY overlap 13:00-16:00 UTC)
+Use datetime hour filtering (e.g., `df[df.index.get_level_values('datetime').hour.between(8, 15)]`)
+to select session bars — do NOT use bar-count offsets to define sessions.
+
 ## Session Times (UTC)
 - Asian: 00:00-08:00 UTC (low volatility)
 - London: 08:00-16:00 UTC (high volatility)

rdagent/scenarios/qlib/experiment/prompts.yaml

@@ -104,7 +104,7 @@ qlib_factor_strategy: |-
      result_df.columns = ['daily_volume_price_divergence']
      ```
 
-  4. **Process ALL data — do not filter dates**: The source HDF5 contains data from 2020-01-01 to 2026-03-20. Do NOT filter to a single year. If your output has only 314 entries (one year of daily data), the factor will be rejected. Expected output: ~1500+ daily entries for 2020-2026.
+  4. **Process ALL data — do not filter dates**: The source HDF5 contains data from 2020-01-01 to 2026-03-20 (development runs may use a 2024-only debug dataset with ~300 entries, which is acceptable). Do NOT filter to a single year in your code. Write your code to process whatever date range is available in the HDF5 file — do not hardcode date filters. Expected output for production data: ~1500+ daily entries for 2020-2026. Expected output for debug data: ~300 daily entries for 2024. Both are valid.
 
   5. **Use `transform()` instead of `apply()` for per-group calculations**: `transform()` preserves the original index while `apply()` may reduce the number of rows unexpectedly:
      ```python
@@ -121,6 +121,35 @@ qlib_factor_strategy: |-
      assert result_df.index.names == ['datetime', 'instrument'], f"Index names must be ['datetime', 'instrument'], got {result_df.index.names}"
      ```
 
+  7. **NEVER use same-day aggregations as the factor value — always shift by 1 day**: If your factor computes a daily aggregate (e.g. daily close return, daily OHLC range, daily volume), that aggregate is only known at end-of-day. Using it at the start of the same day is look-ahead bias. You MUST shift the daily aggregate by 1 day before forward-filling to minute bars:
+     ```python
+     # WRONG: look-ahead bias! Today's close return is not known at 00:00
+     daily_ret = df['$close'].groupby(level='instrument').resample('1D', level='datetime').last().pct_change()
+     result_df['my_factor'] = daily_ret.groupby(level='instrument').transform(lambda x: x.reindex(df.index.get_level_values('datetime'), method='ffill'))
+
+     # CORRECT: shift by 1 trading day so factor value at day T = aggregate of day T-1
+     daily_close = df.groupby([df.index.get_level_values('datetime').normalize(), df.index.get_level_values('instrument')])['$close'].last()
+     daily_close.index.names = ['date', 'instrument']
+     daily_ret = daily_close.groupby(level='instrument').pct_change().shift(1)  # <-- shift(1) is MANDATORY
+     # then map back to minute bars via ffill
+     ```
+     This rule applies to ALL daily aggregations: returns, OHLC stats, volume, momentum, slopes, etc.
+     **Session-based aggregations (London, NY, Asian session returns) are also daily aggregations** — the London
+     session (08:00-16:00 UTC) ends at 16:00, so its return must be shifted by 1 day before use.
+     Intraday rolling factors (e.g. 30-min rolling std computed at bar t using only bars t-N..t-1) do NOT need this shift.
+
+  8. **PREFER pure intraday rolling factors**: Factors that use only a trailing window of recent bars (e.g.
+     rolling(30).mean() of returns, RSI(14), Bollinger Band z-score) have NO look-ahead risk and vary every
+     minute. These are the best candidates for short-horizon (60-180 bar) prediction. Examples:
+     - Rolling 15-min / 30-min / 60-min return momentum (15, 30, 60 bars respectively)
+     - Rolling volatility (std of returns over 20-60 bars)
+     - Distance of close from N-bar moving average (z-score)
+     - RSI or similar oscillators computed on 1-min bars
+     - VWAP deviation (requires volume — use $volume column)
+     Always use `.shift(1)` on the lagged window (e.g. `rolling(N).mean().shift(1)`) to avoid using the
+     current bar's own price in its own feature value.
+     NOTE: 1 bar = 1 minute. The data has ~1440 bars per full trading day. Do NOT use 96 as a day proxy.
+
 qlib_factor_output_format: |-
   Your output should be a pandas dataframe similar to the following example information:
   <class 'pandas.core.frame.DataFrame'>

rdagent/scenarios/qlib/experiment/utils.py

@@ -67,7 +67,7 @@ def get_file_desc(p: Path, variable_list=[]) -> str:
     """
     p = Path(p)
 
-    JJ_TPL = Environment(undefined=StrictUndefined).from_string("""
+    JJ_TPL = Environment(undefined=StrictUndefined).from_string("""  # nosec B701 — renders plain text description, not HTML; autoescape not applicable
 # {{file_name}}
 
 ## File Type

rdagent/scenarios/qlib/proposal/factor_proposal.py

@@ -1,4 +1,6 @@
+import logging
 import json
+import os
 from typing import List, Tuple
 
 from rdagent.components.coder.factor_coder.factor import FactorExperiment, FactorTask
@@ -9,6 +11,47 @@ from rdagent.scenarios.qlib.experiment.model_experiment import QlibModelExperime
 from rdagent.scenarios.qlib.experiment.quant_experiment import QlibQuantScenario
 from rdagent.utils.agent.tpl import T
 
+
+def _build_compressed_history(trace: Trace, max_history: int) -> str:
+    """Return hypothesis_and_feedback string with only `max_history` entries.
+
+    Older entries beyond the last 2 are compressed to one bullet line each.
+    """
+    if len(trace.hist) == 0:
+        return "No previous hypothesis and feedback available since it's the first round."
+
+    FULL_DETAIL = 2
+    old_hist = trace.hist[:-FULL_DETAIL] if len(trace.hist) > FULL_DETAIL else []
+    recent_hist = trace.hist[-FULL_DETAIL:] if len(trace.hist) > FULL_DETAIL else trace.hist
+
+    parts = []
+    if old_hist:
+        lines = ["## Earlier experiments (summarized):"]
+        for exp, fb in old_hist:
+            names = []
+            for task in exp.sub_tasks:
+                if task is not None and hasattr(task, "factor_name"):
+                    names.append(task.factor_name)
+                elif task is not None and hasattr(task, "model_type"):
+                    names.append(getattr(task, "model_type", "model"))
+            ic_str = ""
+            try:
+                if exp.result is not None and "IC" in exp.result.index:
+                    ic_str = f" IC={exp.result.loc['IC']:.4f}"
+            except Exception:
+                logging.debug("Exception caught", exc_info=True)
+            decision = "PASS" if fb.decision else "FAIL"
+            obs = (fb.observations or "")[:120].replace("\n", " ")
+            lines.append(f"- [{decision}]{ic_str} {', '.join(names) or 'unknown'}: {obs}")
+        parts.append("\n".join(lines))
+
+    if recent_hist:
+        rt = Trace(trace.scen)
+        rt.hist = recent_hist
+        parts.append(T("scenarios.qlib.prompts:hypothesis_and_feedback").r(trace=rt))
+
+    return "\n\n".join(parts)
+
 QlibFactorHypothesis = Hypothesis
 
 
@@ -17,13 +60,10 @@ class QlibFactorHypothesisGen(FactorHypothesisGen):
         super().__init__(scen)
 
     def prepare_context(self, trace: Trace) -> Tuple[dict, bool]:
-        hypothesis_and_feedback = (
-            T("scenarios.qlib.prompts:hypothesis_and_feedback").r(
-                trace=trace,
-            )
-            if len(trace.hist) > 0
-            else "No previous hypothesis and feedback available since it's the first round."
-        )
+        max_h = int(os.environ.get("QLIB_QUANT_MAX_FACTOR_HISTORY", "20"))
+        limited = Trace(trace.scen)
+        limited.hist = trace.hist[-max_h:] if len(trace.hist) > max_h else trace.hist
+        hypothesis_and_feedback = _build_compressed_history(limited, max_h)
         last_hypothesis_and_feedback = (
             T("scenarios.qlib.prompts:last_hypothesis_and_feedback").r(
                 experiment=trace.hist[-1][0], feedback=trace.hist[-1][1]
@@ -70,15 +110,15 @@ class QlibFactorHypothesis2Experiment(FactorHypothesis2Experiment):
         if len(trace.hist) == 0:
             hypothesis_and_feedback = "No previous hypothesis and feedback available since it's the first round."
         else:
+            max_h = int(os.environ.get("QLIB_QUANT_MAX_FACTOR_HISTORY", "20"))
+            factor_hist = [
+                e for e in trace.hist
+                if not hasattr(e[0].hypothesis, "action") or e[0].hypothesis.action == "factor"
+            ][-max_h:]
             specific_trace = Trace(trace.scen)
-            for i in range(len(trace.hist) - 1, -1, -1):
-                if not hasattr(trace.hist[i][0].hypothesis, "action") or trace.hist[i][0].hypothesis.action == "factor":
-                    specific_trace.hist.insert(0, trace.hist[i])
-            if len(specific_trace.hist) > 0:
-                specific_trace.hist.reverse()
-                hypothesis_and_feedback = T("scenarios.qlib.prompts:hypothesis_and_feedback").r(
-                    trace=specific_trace,
-                )
+            specific_trace.hist = factor_hist
+            if specific_trace.hist:
+                hypothesis_and_feedback = _build_compressed_history(specific_trace, max_h)
             else:
                 hypothesis_and_feedback = "No previous hypothesis and feedback available."
 

rdagent/scenarios/qlib/proposal/quant_proposal.py

@@ -1,3 +1,4 @@
+import logging
 import json
 import os
 import random
@@ -152,9 +153,41 @@ class QlibQuantHypothesisGen(FactorAndModelHypothesisGen):
                         factor_inserted = True
             if len(specific_trace.hist) > 0:
                 specific_trace.hist.reverse()
-                hypothesis_and_feedback = T("scenarios.qlib.prompts:hypothesis_and_feedback").r(
-                    trace=specific_trace,
-                )
+                # Keep only the 2 most recent experiments in full detail; compress older ones
+                # to brief bullet points to stay within the LLM context window.
+                FULL_DETAIL_COUNT = 2
+                old_hist = specific_trace.hist[:-FULL_DETAIL_COUNT] if len(specific_trace.hist) > FULL_DETAIL_COUNT else []
+                recent_hist = specific_trace.hist[-FULL_DETAIL_COUNT:] if len(specific_trace.hist) > FULL_DETAIL_COUNT else specific_trace.hist
+
+                parts = []
+                if old_hist:
+                    summary_lines = ["## Earlier experiments (summarized):"]
+                    for exp, fb in old_hist:
+                        factor_names = []
+                        for task in exp.sub_tasks:
+                            if task is not None and hasattr(task, "factor_name"):
+                                factor_names.append(task.factor_name)
+                            elif task is not None and hasattr(task, "model_type"):
+                                factor_names.append(getattr(task, "model_type", "model"))
+                        names_str = ", ".join(factor_names) if factor_names else "unknown"
+                        ic_str = ""
+                        try:
+                            if exp.result is not None:
+                                ic_val = exp.result.loc["IC"] if "IC" in exp.result.index else ""
+                                ic_str = f" IC={ic_val:.4f}" if ic_val != "" else ""
+                        except Exception:
+    logging.debug("Error getting IC", exc_info=True)
+                        decision_str = "PASS" if fb.decision else "FAIL"
+                        obs_short = (fb.observations or "")[:120].replace("\n", " ")
+                        summary_lines.append(f"- [{decision_str}]{ic_str} {names_str}: {obs_short}")
+                    parts.append("\n".join(summary_lines))
+
+                if recent_hist:
+                    recent_trace = Trace(specific_trace.scen)
+                    recent_trace.hist = recent_hist
+                    parts.append(T("scenarios.qlib.prompts:hypothesis_and_feedback").r(trace=recent_trace))
+
+                hypothesis_and_feedback = "\n\n".join(parts)
             else:
                 hypothesis_and_feedback = "No previous hypothesis and feedback available."
 

rdagent/scenarios/rl/autorl_bench/core/server.py

@@ -391,7 +391,7 @@ def set_baseline():
     return jsonify({"baseline_score": score, "status": "set"})
 
 
-def run_server(task: str, base_model: str, workspace: str, host: str = "0.0.0.0", port: int = 5000):
+def run_server(task: str, base_model: str, workspace: str, host: str = "127.0.0.1", port: int = 5000):
     """启动服务器"""
     init_server(task, base_model, workspace)
     logger.info(f"Grading Server | task={task} | {host}:{port}")
@@ -435,7 +435,7 @@ class LocalServerContext(GradingServerContext):
         logger.info(f"[Local Mode] Starting evaluation server on port {self.port}...")
         self.server = init_server(self.task, self.base_model, self.workspace)
 
-        self._http_server = make_server("0.0.0.0", self.port, app, threaded=True)
+        self._http_server = make_server("0.0.0.0", self.port, app, threaded=True)  # nosec B104 — intentional: Docker sandbox requires all-interface binding
         self._thread = threading.Thread(target=self._http_server.serve_forever, daemon=True)
         self._thread.start()
 
@@ -488,7 +488,7 @@ if __name__ == "__main__":
     parser.add_argument("--base-model", type=str, default="")
     parser.add_argument("--workspace", type=str, default=".")
     parser.add_argument("--port", type=int, default=5000)
-    parser.add_argument("--host", type=str, default="0.0.0.0")
+    parser.add_argument("--host", type=str, default="127.0.0.1")
     args = parser.parse_args()
 
     run_server(args.task, args.base_model, args.workspace, args.host, args.port)

rdagent/scenarios/rl/autorl_bench/requirements.txt

@@ -9,7 +9,7 @@ peft>=0.18.1
 
 # Evaluation
 opencompass==0.5.1
-setuptools<75  # uv venv doesn't include, opencompass depends on pkg_resources
+setuptools>=78.1.1  # Security fix: GHSA-8g6x-3r52-4m6c (path traversal in PackageIndex.download, arbitrary file write/RCE)
 
 # Inference acceleration (optional, TRL supports 0.10.2-0.12.0)
 # Security: Version >=0.14.0 fixes CVE-2026-22807 (RCE via auto_map dynamic module loading)

rdagent/utils/env.py

@@ -850,24 +850,22 @@ class QlibCondaEnv(LocalEnv[QlibCondaConf]):
     def prepare(self) -> None:
         """Prepare the conda environment if not already created."""
         try:
-            envs = subprocess.run("conda env list", capture_output=True, text=True, shell=True)
+            envs = subprocess.run(["conda", "env", "list"], capture_output=True, text=True)
             if self.conf.conda_env_name not in envs.stdout:
                 print(f"[yellow]Conda env '{self.conf.conda_env_name}' not found, creating...[/yellow]")
                 subprocess.check_call(
-                    f"conda create -y -n {self.conf.conda_env_name} python=3.10",
-                    shell=True,
+                    ["conda", "create", "-y", "-n", self.conf.conda_env_name, "python=3.10"],
                 )
                 subprocess.check_call(
-                    f"conda run -n {self.conf.conda_env_name} pip install --upgrade pip cython",
-                    shell=True,
+                    ["conda", "run", "-n", self.conf.conda_env_name, "pip", "install", "--upgrade", "pip", "cython"],
                 )
                 subprocess.check_call(
-                    f"conda run -n {self.conf.conda_env_name} pip install git+https://github.com/microsoft/qlib.git@2fb9380b342556ddb50a4b24e4fe8655d548b2b8",
-                    shell=True,
+                    ["conda", "run", "-n", self.conf.conda_env_name, "pip", "install",
+                     "git+https://github.com/microsoft/qlib.git@2fb9380b342556ddb50a4b24e4fe8655d548b2b8"],
                 )
                 subprocess.check_call(
-                    f"conda run -n {self.conf.conda_env_name} pip install catboost xgboost tables torch",
-                    shell=True,
+                    ["conda", "run", "-n", self.conf.conda_env_name, "pip", "install",
+                     "catboost", "xgboost", "tables", "torch"],
                 )
 
         except Exception as e:

rdagent/utils/workflow/loop.py

@@ -270,6 +270,11 @@ class LoopBase:
                         msg = "We have reset the loop instance, stop all the routines and resume."
                         raise self.LoopResumeError(msg) from e
                     else:
+                        # Do NOT advance step_idx for unhandled exceptions (e.g. LoopResumeError
+                        # propagating from _propose). Keeping step_idx at the current step lets
+                        # kickoff_loop retry step 0 on the next resume instead of permanently
+                        # corrupting the loop with a missing direct_exp_gen result.
+                        step_forward = False
                         raise  # re-raise unhandled exceptions
                 finally:
                     # No matter the execution succeed or not, we have to finish the following steps

requirements.txt

@@ -9,8 +9,8 @@ psutil
 fire
 fuzzywuzzy
 openai
-litellm>=1.73  # to support `from litellm import get_valid_models`
-aiohttp>=3.13.4  # CVE-2026-22815, CVE-2026-34515, CVE-2026-34516, CVE-2026-34525
+litellm>=1.83.14  # to support `from litellm import get_valid_models`
+aiohttp>=3.13.5  # CVE-2026-22815, CVE-2026-34515, CVE-2026-34516, CVE-2026-34525
 azure.identity
 pyarrow
 rich
@@ -37,7 +37,7 @@ tables
 tree-sitter-python
 tree-sitter
 
-python-dotenv
+python-dotenv>=1.2.2  # CVE: symlink following allows arbitrary file overwrite
 
 # infrastructure related.
 docker
@@ -98,8 +98,8 @@ optuna>=3.5.0
 beautifulsoup4>=4.12.0
 
 # ML Training Pipeline
-lightgbm>=3.3.0
-scipy>=1.9.0
+lightgbm>=3.3.5
+scipy>=1.15.3
 
 # RL Trading (optional - system works without these)
 # Install for full RL training: pip install stable-baselines3[extra] gymnasium

requirements/rl.txt

@@ -8,7 +8,7 @@
 # Only install if you want to use full PPO/A2C/SAC training.
 
 # Core RL library
-stable-baselines3[extra]>=2.0.0
+stable-baselines3[extra]>=2.8.0
 
 # Gymnasium environment (OpenAI Gym successor)
 gymnasium>=0.29.0

scripts/predix_full_eval.py

@@ -198,6 +198,55 @@ def scan_factors(workspace_dir: Path, skip_evaluated: bool = True) -> List[Facto
     return factors
 
 
+# ---------------------------------------------------------------------------
+# Look-ahead bias detection for daily-constant factors
+# ---------------------------------------------------------------------------
+def _shift_daily_constant_factor_if_needed(factor_col: "pd.Series", factor_name: str) -> "pd.Series":
+    """Detect daily-constant factors (look-ahead bias) and shift by 1 trading day."""
+    sample_days = factor_col.index.get_level_values("datetime").normalize().unique()
+    if len(sample_days) < 10:
+        return factor_col
+    rng = np.random.default_rng(42)
+    days_to_check = rng.choice(sample_days, size=min(50, len(sample_days)), replace=False)
+    constant_count = 0
+    for day in days_to_check:
+        day_mask = factor_col.index.get_level_values("datetime").normalize() == day
+        day_vals = factor_col[day_mask].dropna()
+        if len(day_vals) == 0:
+            continue
+        if day_vals.nunique() == 1:
+            constant_count += 1
+    fraction_constant = constant_count / len(days_to_check)
+    if fraction_constant < 0.90:
+        return factor_col
+    # Shift by 1 trading day per instrument
+    import logging
+    logging.getLogger(__name__).info(
+        "Factor '%s' is %.0f%% daily-constant — shifting 1 trading day to fix look-ahead bias",
+        factor_name, fraction_constant * 100,
+    )
+    instruments = factor_col.index.get_level_values("instrument").unique() if "instrument" in factor_col.index.names else [None]
+    shifted_parts = []
+    for instr in instruments:
+        if instr is not None:
+            mask = factor_col.index.get_level_values("instrument") == instr
+            col_instr = factor_col[mask]
+        else:
+            col_instr = factor_col
+        dates = col_instr.index.get_level_values("datetime").normalize()
+        trading_days = dates.unique().sort_values()
+        day_first = col_instr.groupby(dates).first()
+        day_first_shifted = day_first.shift(1)
+        day_first_shifted.index = pd.to_datetime(day_first_shifted.index)
+        day_map = day_first_shifted.reindex(pd.to_datetime(trading_days)).values
+        new_vals = pd.Series(
+            day_map[np.searchsorted(trading_days.values, dates.values)],
+            index=col_instr.index,
+        )
+        shifted_parts.append(new_vals)
+    return pd.concat(shifted_parts).sort_index()
+
+
 # ---------------------------------------------------------------------------
 # Factor evaluator
 # ---------------------------------------------------------------------------
@@ -263,6 +312,7 @@ def evaluate_factor_full(factor: FactorInfo, full_data: pd.DataFrame,
             result = pd.read_hdf(str(result_file), key="data")
             total_count = len(result)
             factor_val = result.iloc[:, 0]
+            factor_val = _shift_daily_constant_factor_if_needed(factor_val, factor.factor_name)
             non_null_count = factor_val.notna().sum()
 
             if non_null_count < 1000:

scripts/predix_gen_strategies_real_bt.py

@@ -250,7 +250,7 @@ Hard requirements:
 - NO global mean/std — always use rolling(window).mean() with shift(1) to avoid look-ahead bias"""
 
         else:
-            system_prompt = f"""You are a quantitative trading expert specializing in EUR/USD intraday strategies.
+            system_prompt = f"""You are a quantitative trading expert specializing in EUR/USD daily swing strategies.
 
 CRITICAL RULES for {STYLE_DESC} (forward horizon: {FORWARD_BARS} bars = ~{FORWARD_BARS/60:.1f} hours):
 1. ONLY use the factors listed below - no others!
@@ -258,15 +258,27 @@ CRITICAL RULES for {STYLE_DESC} (forward horizon: {FORWARD_BARS} bars = ~{FORWAR
 3. Create a pandas Series called 'signal' with values: 1 (long), -1 (short), 0 (neutral)
 4. signal.index MUST match close.index
 5. signal.name must be 'signal'
+6. IMPORTANT: factors are DAILY values broadcast to every 1-minute bar — they change once per day.
+   Use daily-level logic: compare today's factor value to a rolling daily mean (window 5-20 DAYS).
+   To get daily rolling mean: group by date, take first value per day, compute rolling, then reindex back.
+   Example: dates = factors[col].index.get_level_values('datetime').normalize()
+            daily_vals = factors[col].groupby(dates).first()
+            daily_mean = daily_vals.rolling(10).mean().shift(1)
+            daily_signal = (daily_vals > daily_mean).astype(int) * 2 - 1
+            signal = daily_signal.reindex(dates).values  (broadcast back to minute bars)
+7. The signal should change roughly once per day — this produces ~250-500 trades over 6 years.
+8. Keep conditions SIMPLE: one factor above/below its N-day rolling average. Avoid combining 3+ conditions.
 
 Output ONLY valid JSON with these fields:
 {{"strategy_name": "short_name", "factor_names": ["f1", "f2"], "description": "one sentence", "code": "python code"}}"""
 
-            user_prompt = f"""Create a EUR/USD trading strategy using these factors:
+            user_prompt = f"""Create a EUR/USD SWING trading strategy (hold ~{FORWARD_BARS/60:.0f} hours) using these factors:
 
 {factor_list}
 
-{f'Previous feedback: {feedback}' if feedback else 'First attempt - be creative!'}"""
+{f'Previous feedback: {feedback}' if feedback else 'First attempt - be creative!'}
+
+Use daily-level signal logic (factor above/below rolling daily mean). Signal changes once per day."""
         
         api = APIBackend()
         response = api.build_messages_and_create_chat_completion(
@@ -371,6 +383,8 @@ signal.fillna(0).to_pickle('signal.pkl')
         txn_cost_bps=TXN_COST_BPS,
         forward_returns=fwd_returns,
         oos_start=OOS_START_DEFAULT,
+        wf_rolling=False,   # too slow on 2M bars — run via rebacktest script instead
+        mc_n_permutations=50,
     )
 
 # ============================================================================
@@ -567,9 +581,18 @@ def main(target_count=10):
                     progress.update(task, advance=1)
                     continue
 
-                # Check acceptance criteria — OOS must be profitable (primary filter)
-                if (abs(ic) > MIN_IC and sharpe > MIN_SHARPE and trades > MIN_TRADES and dd > MAX_DRAWDOWN
-                        and oos_sharpe > 0.0 and oos_monthly > 0.0):
+                # Monte Carlo p-value (edge significance)
+                mc_pvalue = bt_result.get('mc_pvalue')
+
+                # Rolling walk-forward metrics
+                wf_consistency = bt_result.get('wf_oos_consistency')
+                wf_sharpe_mean = bt_result.get('wf_oos_sharpe_mean')
+
+                # Check acceptance criteria — OOS must be profitable + statistically significant
+                mc_ok = mc_pvalue is None or mc_pvalue < 0.20  # lenient: top 20% non-random
+                wf_ok = wf_consistency is None or wf_consistency >= 0.5  # ≥50% of WF windows profitable
+                if (abs(ic or 0) > MIN_IC and sharpe > MIN_SHARPE and trades > MIN_TRADES and dd > MAX_DRAWDOWN
+                        and oos_sharpe > 0.0 and oos_monthly > 0.0 and mc_ok and wf_ok):
                     # ACCEPT
                     strategy['real_backtest'] = bt_result
                     strategy['metrics'] = bt_result
@@ -583,7 +606,7 @@ def main(target_count=10):
                         'ohlcv_only': OHLCV_ONLY,
                         'engine': 'ftmo_v2',
                         'txn_cost_bps': TXN_COST_BPS,
-                        # Walk-forward OOS metrics
+                        # Walk-forward OOS split
                         'oos_sharpe': bt_result.get('oos_sharpe'),
                         'oos_monthly_return_pct': bt_result.get('oos_monthly_return_pct'),
                         'oos_max_drawdown': bt_result.get('oos_max_drawdown'),
@@ -592,6 +615,15 @@ def main(target_count=10):
                         'is_sharpe': bt_result.get('is_sharpe'),
                         'is_monthly_return_pct': bt_result.get('is_monthly_return_pct'),
                         'oos_start': bt_result.get('oos_start'),
+                        # Rolling walk-forward
+                        'wf_n_windows': bt_result.get('wf_n_windows'),
+                        'wf_oos_sharpe_mean': wf_sharpe_mean,
+                        'wf_oos_sharpe_std': bt_result.get('wf_oos_sharpe_std'),
+                        'wf_oos_monthly_return_mean': bt_result.get('wf_oos_monthly_return_mean'),
+                        'wf_oos_consistency': wf_consistency,
+                        # Monte Carlo significance
+                        'mc_pvalue': mc_pvalue,
+                        'mc_n_permutations': bt_result.get('mc_n_permutations'),
                     }
                     
                     fname = f"{int(time.time())}_{strategy['strategy_name']}.json"
@@ -614,12 +646,17 @@ def main(target_count=10):
                                           f"IC={ic:.4f}, Sharpe={sharpe:.3f}, Trades={trades}, DD={dd:.1%}")
                 else:
                     oos_info = f"OOS_Sharpe={oos_sharpe:+.2f} OOS_Mon={oos_monthly:+.2f}%" if oos_sharpe is not None else ""
-                    _log.info(f"REJECTED  IC={ic:.4f}  Sharpe={sharpe:.2f}  Trades={trades}  DD={dd:.1%}  {oos_info}")
+                    mc_info  = f" MC_p={mc_pvalue:.2f}" if mc_pvalue is not None else ""
+                    wf_info  = f" WF_consistency={wf_consistency:.0%}" if wf_consistency is not None else ""
+                    _ic = ic or 0; _sh = sharpe or 0; _dd = dd or 0
+                    _log.info(f"REJECTED  IC={_ic:.4f}  Sharpe={_sh:.2f}  Trades={trades}  DD={_dd:.1%}  {oos_info}{mc_info}{wf_info}")
                     feedback_history.append(
-                        f"Failed: IC={ic:.4f}, Sharpe={sharpe:.2f}, Trades={trades}, DD={dd:.1%}, "
-                        f"OOS_Sharpe={oos_sharpe:+.2f}, OOS_Monthly={oos_monthly:+.2f}%. "
-                        f"Need |IC|>{MIN_IC}, Sharpe>{MIN_SHARPE}, Trades>{MIN_TRADES}, "
-                        f"OOS_Sharpe>0 AND OOS_Monthly>0 — strategy must generalise to unseen data (2024+)."
+                        f"Failed: IC={_ic:.4f}, Sharpe={_sh:.2f}, Trades={trades}, DD={_dd:.1%}, "
+                        f"OOS_Sharpe={oos_sharpe:+.2f}, OOS_Monthly={oos_monthly:+.2f}%"
+                        + (f", MC_p={mc_pvalue:.2f}" if mc_pvalue is not None else "")
+                        + (f", WF_consistency={wf_consistency:.0%}" if wf_consistency is not None else "")
+                        + f". Need |IC|>{MIN_IC}, Sharpe>{MIN_SHARPE}, Trades>{MIN_TRADES}, "
+                        f"OOS_Sharpe>0, OOS_Monthly>0, MC_p<0.20, WF_consistency≥50%."
                     )
             
             progress.update(task, advance=1)

scripts/predix_rebacktest_unified.py

@@ -188,6 +188,8 @@ def rebacktest_one(
         close=close_a,
         signal=signal,
         txn_cost_bps=txn_cost_bps,
+        wf_rolling=True,
+        mc_n_permutations=200,
     )
     result["status_detail"] = result.pop("status")
     result["status"] = "ok"
@@ -264,6 +266,15 @@ def main() -> None:
                     "oos_win_rate": bt.get("oos_win_rate"),
                     "oos_n_trades": bt.get("oos_n_trades"),
                     "oos_start": bt.get("oos_start"),
+                    # Rolling walk-forward
+                    "wf_n_windows": bt.get("wf_n_windows"),
+                    "wf_oos_sharpe_mean": bt.get("wf_oos_sharpe_mean"),
+                    "wf_oos_sharpe_std": bt.get("wf_oos_sharpe_std"),
+                    "wf_oos_monthly_return_mean": bt.get("wf_oos_monthly_return_mean"),
+                    "wf_oos_consistency": bt.get("wf_oos_consistency"),
+                    # Monte Carlo significance
+                    "mc_pvalue": bt.get("mc_pvalue"),
+                    "mc_n_permutations": bt.get("mc_n_permutations"),
                 }
                 data["sharpe_ratio"] = bt.get("sharpe")
                 data["max_drawdown"] = bt.get("max_drawdown")
@@ -299,6 +310,12 @@ def main() -> None:
                 "oos_monthly_pct": bt.get("oos_monthly_return_pct"),
                 "oos_dd": bt.get("oos_max_drawdown"),
                 "oos_trades": bt.get("oos_n_trades"),
+                # Rolling walk-forward
+                "wf_n_windows": bt.get("wf_n_windows"),
+                "wf_oos_sharpe_mean": bt.get("wf_oos_sharpe_mean"),
+                "wf_oos_consistency": bt.get("wf_oos_consistency"),
+                # Monte Carlo
+                "mc_pvalue": bt.get("mc_pvalue"),
             }
             if "annualized_return" in bt:
                 row["new_annual_return_cagr"] = bt["annualized_return"]

test/backtesting/test_ftmo_oos.py

@@ -0,0 +1,240 @@
+"""
+Tests for backtest_signal_ftmo and walk-forward OOS validation.
+
+Covers:
+- FTMO daily/total loss limits
+- Risk-based leverage calculation
+- OOS split returns independent IS and OOS metrics
+- OOS uses fresh FTMO simulation (not contaminated by IS losses)
+- Monte Carlo permutation test helper
+"""
+from __future__ import annotations
+
+import numpy as np
+import pandas as pd
+import pytest
+
+from rdagent.components.backtesting.vbt_backtest import (
+    OOS_START_DEFAULT,
+    backtest_signal_ftmo,
+    FTMO_MAX_DAILY_LOSS,
+    FTMO_MAX_TOTAL_LOSS,
+    monte_carlo_trade_pvalue,
+    walk_forward_rolling,
+)
+
+
+# ---------------------------------------------------------------------------
+# Fixtures
+# ---------------------------------------------------------------------------
+@pytest.fixture
+def close_2yr() -> pd.Series:
+    """~3 months of synthetic 1-min EUR/USD (enough bars for all leverage/FTMO tests)."""
+    np.random.seed(42)
+    n = 90 * 1440  # 90 days × 1440 min
+    idx = pd.date_range("2022-01-01", periods=n, freq="1min")
+    price = 1.10 + np.cumsum(np.random.randn(n) * 0.00005)
+    return pd.Series(price, index=idx)
+
+
+@pytest.fixture
+def close_6yr() -> pd.Series:
+    """Synthetic data crossing the 2024-01-01 IS/OOS boundary.
+
+    120 days starting 2023-09-01 → ends ~2024-01-01, giving ~30 days of OOS data.
+    Small enough to keep tests fast.
+    """
+    np.random.seed(7)
+    n = 150 * 1440  # 2023-09-01 + 150d ≈ 2024-01-28 → ~28 days of OOS data
+    idx = pd.date_range("2023-09-01", periods=n, freq="1min")
+    price = 1.10 + np.cumsum(np.random.randn(n) * 0.00005)
+    return pd.Series(price, index=idx)
+
+
+def _random_signal(index: pd.Index, seed: int = 0) -> pd.Series:
+    np.random.seed(seed)
+    return pd.Series(np.random.choice([-1.0, 0.0, 1.0], size=len(index)), index=index)
+
+
+# ---------------------------------------------------------------------------
+# FTMO leverage tests
+# ---------------------------------------------------------------------------
+def test_ftmo_result_contains_leverage_fields(close_2yr):
+    signal = _random_signal(close_2yr.index)
+    r = backtest_signal_ftmo(close_2yr, signal, oos_start=None)
+    assert "ftmo_leverage" in r
+    assert "ftmo_risk_pct" in r
+    assert "ftmo_stop_pips" in r
+    assert r["ftmo_leverage"] > 0
+
+
+def test_ftmo_leverage_capped_at_max(close_2yr):
+    signal = _random_signal(close_2yr.index)
+    # With very tight stop (1 pip) risk_pct=0.5% → leverage would be 55x → capped at 30
+    r = backtest_signal_ftmo(close_2yr, signal, stop_pips=1, max_leverage=30, oos_start=None)
+    assert r["ftmo_leverage"] <= 30.0
+
+
+def test_ftmo_zero_signal_produces_no_trades(close_2yr):
+    signal = pd.Series(0.0, index=close_2yr.index)
+    r = backtest_signal_ftmo(close_2yr, signal, oos_start=None)
+    assert r["n_trades"] == 0
+    assert r["total_return"] == 0.0
+
+
+# ---------------------------------------------------------------------------
+# OOS split tests
+# ---------------------------------------------------------------------------
+def test_oos_split_produces_is_and_oos_keys(close_6yr):
+    signal = _random_signal(close_6yr.index)
+    r = backtest_signal_ftmo(close_6yr, signal, oos_start="2024-01-01")
+
+    assert "is_sharpe" in r
+    assert "oos_sharpe" in r
+    assert "is_monthly_return_pct" in r
+    assert "oos_monthly_return_pct" in r
+    assert "is_n_bars" in r
+    assert "oos_n_bars" in r
+    assert r["oos_start"] == "2024-01-01"
+
+
+def test_oos_split_bars_sum_to_total(close_6yr):
+    signal = _random_signal(close_6yr.index)
+    r = backtest_signal_ftmo(close_6yr, signal, oos_start="2024-01-01")
+    assert r["is_n_bars"] + r["oos_n_bars"] == len(close_6yr)
+
+
+def test_oos_none_disables_split(close_6yr):
+    signal = _random_signal(close_6yr.index)
+    r = backtest_signal_ftmo(close_6yr, signal, oos_start=None)
+    assert "is_sharpe" not in r
+    assert "oos_sharpe" not in r
+
+
+def test_oos_is_independent_of_is_losses(close_6yr):
+    """OOS must use a fresh FTMO simulation — IS blowup must not zero OOS trades."""
+    # Force the IS period to blow up immediately with max short on rising market
+    rising = pd.Series(
+        np.linspace(1.0, 2.0, len(close_6yr)),
+        index=close_6yr.index,
+    )
+    always_short = pd.Series(-1.0, index=close_6yr.index)
+
+    r = backtest_signal_ftmo(rising, always_short, oos_start="2024-01-01")
+
+    # IS should be wiped out (total loss limit hit), but OOS must still trade
+    assert r.get("oos_n_trades", 0) is not None
+    assert r.get("oos_n_bars", 0) > 0
+
+
+def test_oos_default_start_matches_constant(close_6yr):
+    signal = _random_signal(close_6yr.index)
+    r = backtest_signal_ftmo(close_6yr, signal)
+    assert r.get("oos_start") == OOS_START_DEFAULT
+
+
+# ---------------------------------------------------------------------------
+# Monte Carlo permutation test helper
+# ---------------------------------------------------------------------------
+def _monte_carlo_pvalue(close: pd.Series, signal: pd.Series, n_permutations: int = 200, seed: int = 0) -> float:
+    """
+    Estimate p-value: fraction of random permutations that beat the real Sharpe.
+    p < 0.05 → strategy has statistically significant edge.
+    """
+    real_r = backtest_signal_ftmo(close, signal, oos_start=None)
+    real_sharpe = real_r.get("sharpe", 0.0) or 0.0
+
+    rng = np.random.default_rng(seed)
+    beat = 0
+    signal_vals = signal.values.copy()
+    for _ in range(n_permutations):
+        perm = rng.permutation(signal_vals)
+        perm_signal = pd.Series(perm, index=signal.index)
+        perm_r = backtest_signal_ftmo(close, perm_signal, oos_start=None)
+        if (perm_r.get("sharpe") or 0.0) >= real_sharpe:
+            beat += 1
+    return beat / n_permutations
+
+
+@pytest.mark.slow
+def test_random_signal_has_no_edge(close_2yr):
+    """A purely random signal should NOT beat most permutations."""
+    signal = _random_signal(close_2yr.index, seed=42)
+    pval = _monte_carlo_pvalue(close_2yr, signal, n_permutations=50)
+    # Random vs random: p-value should be near 0.5 (not significant)
+    assert pval > 0.10, f"Random signal unexpectedly significant: p={pval:.2f}"
+
+
+@pytest.mark.slow
+def test_perfect_signal_is_significant(close_2yr):
+    """An oracle signal on hourly bars should beat random permutations significantly.
+
+    Per-minute oracle trading is unprofitable due to FTMO transaction costs, so we
+    use 60-bar held positions (≈1h) where each directional move is large enough to
+    cover the spread.
+    """
+    bar_ret = close_2yr.pct_change().fillna(0)
+    # Hourly oracle: sign of 60-bar future return, broadcast to all 60 minute bars
+    hourly_ret = bar_ret.rolling(60).sum().shift(-60).fillna(0)
+    perfect = pd.Series(np.sign(hourly_ret), index=close_2yr.index)
+    pval = _monte_carlo_pvalue(close_2yr, perfect, n_permutations=50)
+    assert pval < 0.30, f"Hourly oracle signal should beat random permutations: p={pval:.2f}"
+
+
+# ---------------------------------------------------------------------------
+# FTMO metrics in result dict
+# ---------------------------------------------------------------------------
+def test_ftmo_result_has_equity_and_profit(close_2yr):
+    signal = _random_signal(close_2yr.index)
+    r = backtest_signal_ftmo(close_2yr, signal, oos_start=None)
+    assert "ftmo_end_equity" in r
+    assert "ftmo_monthly_profit" in r
+    assert r["ftmo_end_equity"] > 0
+
+
+# ---------------------------------------------------------------------------
+# Monte Carlo trade permutation tests
+# ---------------------------------------------------------------------------
+def test_mc_pvalue_in_result(close_2yr):
+    signal = _random_signal(close_2yr.index)
+    r = backtest_signal_ftmo(close_2yr, signal, oos_start=None, mc_n_permutations=50)
+    assert "mc_pvalue" in r
+    assert 0.0 <= r["mc_pvalue"] <= 1.0
+    assert r["mc_n_permutations"] == 50
+
+
+def test_mc_pvalue_disabled_by_default(close_2yr):
+    signal = _random_signal(close_2yr.index)
+    r = backtest_signal_ftmo(close_2yr, signal, oos_start=None)
+    assert "mc_pvalue" not in r
+
+
+def test_mc_zero_trades_returns_one(close_2yr):
+    """Zero-signal → no trades → p-value must be 1.0 (no edge)."""
+    trade_pnl = pd.Series([], dtype=float)
+    assert monte_carlo_trade_pvalue(trade_pnl, n_permutations=10) == 1.0
+
+
+# ---------------------------------------------------------------------------
+# Rolling walk-forward tests
+# ---------------------------------------------------------------------------
+def test_wf_rolling_keys_in_result(close_6yr):
+    signal = _random_signal(close_6yr.index)
+    r = backtest_signal_ftmo(close_6yr, signal, oos_start="2024-01-01", wf_rolling=True)
+    # With only ~150 days of data, windows may be 0 — just check key presence
+    assert "wf_n_windows" in r
+
+
+def test_wf_rolling_disabled_by_default(close_6yr):
+    signal = _random_signal(close_6yr.index)
+    r = backtest_signal_ftmo(close_6yr, signal, oos_start="2024-01-01")
+    assert "wf_n_windows" not in r
+
+
+def test_wf_consistency_range(close_6yr):
+    """wf_oos_consistency must be in [0, 1] when windows exist."""
+    signal = _random_signal(close_6yr.index)
+    r = backtest_signal_ftmo(close_6yr, signal, oos_start="2024-01-01", wf_rolling=True)
+    c = r.get("wf_oos_consistency")
+    if c is not None:
+        assert 0.0 <= c <= 1.0

test/qlib/test_auto_fixer.py

@@ -0,0 +1,249 @@
+"""Tests for FactorAutoFixer — the pre-execution code patcher."""
+
+import pytest
+
+from rdagent.components.coder.factor_coder.auto_fixer import FactorAutoFixer
+
+
+@pytest.fixture()
+def fixer():
+    return FactorAutoFixer()
+
+
+class TestResetIndexGroupby:
+    def test_replaces_level_groupby_on_reset_var(self, fixer):
+        code = "df_r = df.reset_index()\ndf_r['x'] = df_r.groupby(level=1)['$close'].mean()"
+        result = fixer.fix(code)
+        assert "groupby('instrument')" in result
+
+    def test_does_not_touch_normal_multiindex_groupby(self, fixer):
+        code = "df['x'] = df.groupby(level=1)['$close'].mean()"
+        result = fixer.fix(code)
+        assert "groupby(level=1)" in result
+
+
+class TestGroupbyMixedLevels:
+    def test_strips_string_from_mixed_list(self, fixer):
+        result = fixer.fix("df.groupby(level=[1, 'date']).apply(fn)")
+        assert "groupby(level=1)" in result
+
+    def test_multiple_ints_kept(self, fixer):
+        result = fixer.fix("df.groupby(level=[0, 1, 'x']).apply(fn)")
+        assert "groupby(level=[0, 1])" in result
+
+
+class TestGroupbyColumnOnMultiindex:
+    def test_instrument_date_becomes_two_level(self, fixer):
+        code = "df['v'] = df.groupby(['instrument', 'date'])['$volume'].cumsum()"
+        result = fixer.fix(code)
+        assert "get_level_values(1)" in result
+        assert "normalize()" in result
+        assert "level=1)" not in result.split("get_level_values")[0]
+
+    def test_date_instrument_becomes_two_level(self, fixer):
+        code = "df['v'] = df.groupby(['date', 'instrument'])['$volume'].cumsum()"
+        result = fixer.fix(code)
+        assert "get_level_values(0).normalize()" in result
+        assert "get_level_values(1)" in result
+
+    def test_single_instrument_becomes_level1(self, fixer):
+        result = fixer.fix("df.groupby(['instrument'])['x'].mean()")
+        assert "groupby(level=1)" in result
+
+    def test_reset_index_not_double_fixed(self, fixer):
+        # After reset_index fix emits groupby('instrument'), this fixer must NOT
+        # convert that to groupby(level=1).
+        code = "df_r = df.reset_index()\ndf_r['x'] = df_r.groupby(level=1)['p'].mean()"
+        result = fixer.fix(code)
+        assert "groupby('instrument')" in result
+
+
+class TestChainedGroupby:
+    def test_chained_groupby_level_then_date(self, fixer):
+        code = "df.groupby(level=1).groupby('date')['price_volume'].transform('cumsum')"
+        result = fixer.fix(code)
+        assert "get_level_values(1)" in result
+        assert "get_level_values(0).normalize()" in result
+        assert ".groupby('date')" not in result
+
+    def test_chained_groupby_with_double_quotes(self, fixer):
+        code = 'df.groupby(level=0).groupby("date")["col"].sum()'
+        result = fixer.fix(code)
+        assert "get_level_values" in result
+        assert '.groupby("date")' not in result
+
+    def test_list_with_level_keyword_syntax_error(self, fixer):
+        # groupby([level=1, 'date']) is a SyntaxError — must be fixed before execution
+        code = "asian_vol = df[mask].groupby([level=1, 'date'])['log_return'].std()"
+        result = fixer.fix(code)
+        assert "get_level_values(1)" in result
+        assert "normalize()" in result
+        assert "level=1," not in result
+
+    def test_list_with_level_keyword_reversed(self, fixer):
+        code = "df.groupby(['date', level=1])['x'].mean()"
+        result = fixer.fix(code)
+        assert "get_level_values" in result
+        assert "level=1" not in result
+
+
+class TestMinPeriodsNotTouched:
+    def test_small_min_periods_preserved(self, fixer):
+        # _fix_min_periods is disabled — LLM-set min_periods must not be changed.
+        # window=60, min_periods=1 should stay as-is (was wrongly raised to 60 before).
+        result = fixer.fix("df.groupby(level=1)['x'].transform(lambda x: x.rolling(window=60, min_periods=1).mean())")
+        assert "min_periods=1" in result
+
+    def test_large_window_min_periods_preserved(self, fixer):
+        # window=240 > 96 bars/day: if min_periods were set to 240 the output would be
+        # all-NaN for intraday data. Verify we leave it untouched.
+        result = fixer.fix("df['x'] = df.groupby(level=1)['y'].transform(lambda x: x.rolling(240, min_periods=10).std())")
+        assert "min_periods=10" in result
+
+
+class TestInstrumentColumnAccess:
+    def test_instrument_column_replaced(self, fixer):
+        code = "df['group_key'] = df['instrument'] + '_' + df['day_id'].astype(str)"
+        result = fixer.fix(code)
+        assert "df.index.get_level_values(1)" in result
+        assert "df['instrument']" not in result
+
+    def test_reset_index_var_not_touched(self, fixer):
+        # After reset_index, 'instrument' IS a real column — must not be replaced
+        code = "df_r = df.reset_index()\nval = df_r['instrument'].unique()"
+        result = fixer.fix(code)
+        assert "df_r['instrument']" in result
+        assert "get_level_values" not in result
+
+    def test_groupby_after_instrument_fix(self, fixer):
+        # Combined: df['instrument'] in a groupby context
+        code = "df['key'] = df['instrument']\nout = df.groupby(df['key'])[['$close']].mean()"
+        result = fixer.fix(code)
+        assert "df['instrument']" not in result
+
+    def test_assignment_target_not_touched(self, fixer):
+        # df['instrument'] = <expr> is an assignment — must NOT be converted to
+        # df.index.get_level_values(1) = <expr> (SyntaxError)
+        code = "df['instrument'] = df.index.get_level_values('instrument')"
+        result = fixer.fix(code)
+        assert "df['instrument'] =" in result
+
+
+class TestInstrumentLocMultiindex:
+    def test_loc_replaced_with_xs(self, fixer):
+        code = (
+            "for instrument in df.index.get_level_values('instrument').unique():\n"
+            "    inst_df = df.loc[instrument].copy()\n"
+        )
+        result = fixer.fix(code)
+        assert "df.xs(instrument, level=1)" in result
+        assert "df.loc[instrument]" not in result
+
+    def test_loc_replaced_with_level1_int(self, fixer):
+        code = (
+            "for inst in df.index.get_level_values(1).unique():\n"
+            "    data = df.loc[inst]\n"
+        )
+        result = fixer.fix(code)
+        assert "df.xs(inst, level=1)" in result
+
+    def test_loc_assignment_not_touched(self, fixer):
+        # Write-back df.loc[instrument] = ... must not be changed
+        code = (
+            "for instrument in df.index.get_level_values('instrument').unique():\n"
+            "    df.loc[instrument] = modified\n"
+        )
+        result = fixer.fix(code)
+        assert "df.loc[instrument] = modified" in result
+
+    def test_non_instrument_loop_not_touched(self, fixer):
+        # for-loop not related to instrument levels must not be changed
+        code = "for date in dates:\n    sub = df.loc[date]\n"
+        result = fixer.fix(code)
+        assert "df.loc[date]" in result
+
+
+class TestGroupbyLevelStringNames:
+    def test_level_instrument_date_replaced(self, fixer):
+        code = "df.groupby(level=['instrument', 'date'])['col'].transform('sum')"
+        result = fixer.fix(code)
+        assert "get_level_values(1)" in result
+        assert "get_level_values(0).normalize()" in result
+        assert "level=['instrument', 'date']" not in result
+
+    def test_level_date_instrument_replaced(self, fixer):
+        code = "data.groupby(level=['date', 'instrument'])['x'].mean()"
+        result = fixer.fix(code)
+        assert "get_level_values(0).normalize()" in result
+        assert "get_level_values(1)" in result
+
+    def test_level_instrument_single_replaced(self, fixer):
+        code = "df.groupby(level=['instrument'])['vol'].sum()"
+        result = fixer.fix(code)
+        assert "groupby(level=1)" in result
+        assert "level=['instrument']" not in result
+
+
+class TestGroupbyApplyToTransform:
+    def test_col_apply_lambda_replaced(self, fixer):
+        code = "df_overlap.groupby(level=1)['$close'].apply(lambda x: np.log(x / x.shift(1)))"
+        result = fixer.fix(code)
+        assert ".transform(" in result
+        assert ".apply(" not in result
+
+    def test_col_apply_lambda_preserves_lambda_body(self, fixer):
+        code = "series.groupby(level=1)['ret'].apply(lambda x: x.cumsum())"
+        result = fixer.fix(code)
+        assert "lambda x: x.cumsum()" in result
+        assert ".transform(" in result
+
+    def test_transform_reset_index_stripped(self, fixer):
+        # .transform() already preserves index — .reset_index() after it is wrong
+        code = "df['v'] = df.groupby(level=1)['x'].transform(lambda x: x.rolling(20).mean()).reset_index(level=0, drop=True)"
+        result = fixer.fix(code)
+        assert ".reset_index(level=0, drop=True)" not in result
+        assert ".transform(" in result
+
+
+class TestZeroVolumeProxy:
+    def test_injects_proxy_when_volume_used(self, fixer):
+        code = (
+            "def calc():\n"
+            "    df = pd.read_hdf('data.h5', key='data')\n"
+            "    df['pv'] = df['$close'] * df['$volume']\n"
+            "    return df[['pv']]\n"
+        )
+        result = fixer.fix(code)
+        assert "volume proxy" in result
+        assert "df['$volume'] = df['$high'] - df['$low']" in result
+        # Proxy must come right after read_hdf line
+        lines = result.splitlines()
+        hdf_idx = next(i for i, l in enumerate(lines) if "read_hdf" in l)
+        assert "volume proxy" in lines[hdf_idx + 1]
+
+    def test_no_injection_when_volume_absent(self, fixer):
+        code = "df = pd.read_hdf('data.h5', key='data')\ndf['x'] = df['$close'].pct_change()\n"
+        result = fixer.fix(code)
+        assert "volume proxy" not in result
+
+    def test_no_double_injection(self, fixer):
+        code = (
+            "def calc():\n"
+            "    df = pd.read_hdf('data.h5', key='data')\n"
+            "    # volume proxy: $volume is always 0 in FX data — use price-range as proxy\n"
+            "    if (df['$volume'] == 0).all():\n"
+            "        df['$volume'] = df['$high'] - df['$low']\n"
+            "    df['pv'] = df['$close'] * df['$volume']\n"
+        )
+        result = fixer.fix(code)
+        assert result.count("volume proxy") == 1
+
+
+class TestRollingDdof:
+    def test_removes_ddof_from_rolling_args(self, fixer):
+        result = fixer.fix("df.rolling(20, min_periods=1, ddof=1).std()")
+        assert "ddof" not in result
+
+    def test_removes_ddof_from_std_args(self, fixer):
+        result = fixer.fix("df.rolling(20).std(ddof=1)")
+        assert "ddof" not in result