chore(master): release 1.4.2 (#48)

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

.release-please-manifest.json

@@ -1,3 +1,3 @@
 {
-  ".": "1.3.9"
+  ".": "1.4.2"
 }

CHANGELOG.md

@@ -1,5 +1,47 @@
 # Changelog
 
+## [1.4.2](https://github.com/TPTBusiness/Predix/compare/v1.4.1...v1.4.2) (2026-05-03)
+
+
+### Bug Fixes
+
+* add missing sys import and fix undefined acc_rate in factor eval ([c45f990](https://github.com/TPTBusiness/Predix/commit/c45f9908ee321400f0a19c57f1482e4cd1394a50))
+
+## [1.4.1](https://github.com/TPTBusiness/Predix/compare/v1.4.0...v1.4.1) (2026-05-03)
+
+
+### Bug Fixes
+
+* 15 bug fixes across orchestrator, runner, backtest, and infrastructure ([163687d](https://github.com/TPTBusiness/Predix/commit/163687d7e1c278a085d7052a3f958a3edb501e77))
+* also catch ValueError in mean_variance for dimension mismatch ([ed73b72](https://github.com/TPTBusiness/Predix/commit/ed73b7253f7dc6459ee30dd81a1ce1194e46e9af))
+* close log file handle, fix FTMO equity double-count, remove bare except ([76219a5](https://github.com/TPTBusiness/Predix/commit/76219a53efddaafc2b8bd48a0f76c1d4325e6ea5))
+* correct project root paths and subprocess handling in parallel runner and CLI ([9735e3a](https://github.com/TPTBusiness/Predix/commit/9735e3a4d8f01e7b16fb9b185a002396a915cea4))
+* filter NaN in max(), remove redundant ternary, handle non-finite vbt results ([f89fbb3](https://github.com/TPTBusiness/Predix/commit/f89fbb3421faf6ccdc8e68a911fd9db2c166120f))
+* fix type annotation, remove unused parameter, improve import_class errors ([8b6ab73](https://github.com/TPTBusiness/Predix/commit/8b6ab735c05629bf6b76ddc2fd8b15617600cad7))
+* resolve dead code, shell injection risk, mutable defaults, and other bugs ([afff262](https://github.com/TPTBusiness/Predix/commit/afff26287f7c4df7ddfde4e816d280fe845e11eb))
+* resolve unbound variable, logger shadowing, withdraw_loop edge case, and other bugs in main scripts ([748cf9b](https://github.com/TPTBusiness/Predix/commit/748cf9b214a3e8447f1289fc4cf1e92ad6cc2f1a))
+
+## [1.4.0](https://github.com/TPTBusiness/Predix/compare/v1.3.11...v1.4.0) (2026-05-01)
+
+
+### Features
+
+* **optimizer:** add max_positions parameter to Optuna search space ([fdb4be3](https://github.com/TPTBusiness/Predix/commit/fdb4be3b3ebd93325e7821f4251148424184a40d))
+
+## [1.3.11](https://github.com/TPTBusiness/Predix/compare/v1.3.10...v1.3.11) (2026-05-01)
+
+
+### Bug Fixes
+
+* **ci:** lazy import logger in predix.py and cli.py to avoid ImportError in test env ([60763e8](https://github.com/TPTBusiness/Predix/commit/60763e8eae34f41865ba8e5e65bdfde13b564b4b))
+
+## [1.3.10](https://github.com/TPTBusiness/Predix/compare/v1.3.9...v1.3.10) (2026-05-01)
+
+
+### Bug Fixes
+
+* **security:** replace remaining assert statements with proper error handling ([928533d](https://github.com/TPTBusiness/Predix/commit/928533d9a81bd5062f07458fbf94d3c7fe347775))
+
 ## [1.3.9](https://github.com/TPTBusiness/Predix/compare/v1.3.8...v1.3.9) (2026-05-01)
 
 

predix.py

@@ -13,12 +13,18 @@ import sys
 from pathlib import Path
 
 from dotenv import load_dotenv
+
 load_dotenv(Path(__file__).parent / ".env")
 
 import typer
 from rich.console import Console
 
-from rdagent.utils.env import logger
+try:
+    from rdagent.utils.env import logger
+except ImportError:
+    import logging
+
+    logger = logging.getLogger(__name__)
 
 app = typer.Typer(help="Predix - AI Quantitative Trading Agent")
 console = Console()
@@ -109,7 +115,7 @@ def _ensure_kronos_factor_in_pool(con) -> None:
         color = "green" if abs(ic) > 0.01 else "yellow"
         con.print(
             f"  [bold {color}]Kronos Factor ready:[/bold {color}] IC={ic:.4f}, "
-            f"Hit-Rate={hit_rate:.1%} — added to strategy pool"
+            f"Hit-Rate={hit_rate:.1%} — added to strategy pool",
         )
 
     except Exception as e:
@@ -196,9 +202,9 @@ def quant(
         predix health - Check system health and configuration
     """
     import subprocess
+    import sys
     import threading
     import time
-    import sys
 
     # ---- Parallel Run Isolation ----
     # When run_id > 0, isolate all outputs (logs, results, workspace)
@@ -221,10 +227,9 @@ def quant(
         console.print(f"   [dim]Log: {log_file}[/dim]")
         console.print(f"   [dim]Results: results/runs/run{run_id}/[/dim]")
         console.print(f"   [dim]Workspace: {workspace_dir.name}/[/dim]")
-    else:
-        # Single run mode: default log file
-        if log_file is None:
-            log_file = "fin_quant.log"
+    # Single run mode: default log file
+    elif log_file is None:
+        log_file = "fin_quant.log"
 
     # ---- Log File Setup (daily-rotated) ----
     from datetime import datetime as _dt
@@ -232,10 +237,14 @@ def quant(
     _daily_dir = Path(__file__).parent / "logs" / _today
     _daily_dir.mkdir(parents=True, exist_ok=True)
 
+    _log_f = None
+    _orig_stdout = sys.stdout
+    _orig_stderr = sys.stderr
+
     if log_file.lower() != "none":
         log_path = _daily_dir / log_file
         # Open log file for appending (raw stdout/stderr capture)
-        log_f = open(log_path, "a", encoding="utf-8")
+        _log_f = open(log_path, "a", encoding="utf-8")
 
         # Redirect stdout and stderr to both console and log file
         class TeeWriter:
@@ -247,18 +256,18 @@ def quant(
                     try:
                         s.write(data)
                         s.flush()
-                    except:
+                    except Exception:
                         pass
 
             def flush(self):
                 for s in self._streams:
                     try:
                         s.flush()
-                    except:
+                    except Exception:
                         pass
 
-        sys.stdout = TeeWriter(sys.__stdout__, log_f)
-        sys.stderr = TeeWriter(sys.__stderr__, log_f)
+        sys.stdout = TeeWriter(_orig_stdout, _log_f)
+        sys.stderr = TeeWriter(_orig_stderr, _log_f)
 
         console.print(f"\n[dim]📝 Logging to: logs/{_today}/{log_file}[/dim]")
     else:
@@ -271,7 +280,7 @@ def quant(
         if not api_key:
             console.print("\n[bold red]❌ OPENROUTER_API_KEY not set in .env[/bold red]")
             console.print("[yellow]Add your API key to .env:[/yellow]")
-            console.print('  OPENROUTER_API_KEY=sk-or-your-key-here')
+            console.print("  OPENROUTER_API_KEY=sk-or-your-key-here")
             raise typer.Exit(code=1)
 
         # Setup both API keys for load balancing
@@ -284,7 +293,7 @@ def quant(
             os.environ["LITELLM_PARALLEL_CALLS"] = "2"
             console.print(f"\n[bold blue]🌐 Using OpenRouter (2 API Keys):[/bold blue] [cyan]{os.environ['CHAT_MODEL']}[/cyan]")
             console.print(f"   [dim]Keys: {api_key[:15]}*** + {api_key_2[:15]}***[/dim]")
-            console.print(f"   [dim]Parallel: 2 concurrent requests[/dim]")
+            console.print("   [dim]Parallel: 2 concurrent requests[/dim]")
         else:
             os.environ["OPENAI_API_KEY"] = api_key
             console.print(f"\n[bold blue]🌐 Using OpenRouter:[/bold blue] [cyan]{os.environ['CHAT_MODEL']}[/cyan]")
@@ -302,7 +311,7 @@ def quant(
     # ---- Dashboards ----
     if dashboard:
         def start_web_dashboard():
-            console.print(f"\n[bold green]🚀 Web Dashboard: http://localhost:5000[/bold green]")
+            console.print("\n[bold green]🚀 Web Dashboard: http://localhost:5000[/bold green]")
             subprocess.run(
                 ["python", "web/dashboard_api.py"],
                 cwd=str(Path(__file__).parent),
@@ -327,7 +336,7 @@ def quant(
     from rdagent.app.qlib_rd_loop.quant import main as fin_quant
     from rdagent.log.daily_log import session as _daily_session
 
-    console.print(f"\n[bold cyan]📊 Starting EURUSD Trading Loop...[/bold cyan]\n")
+    console.print("\n[bold cyan]📊 Starting EURUSD Trading Loop...[/bold cyan]\n")
 
     _ctx = {"model": model}
     if run_id:
@@ -337,11 +346,17 @@ def quant(
     if step_n:
         _ctx["steps"] = step_n
 
-    with _daily_session("fin_quant", **_ctx):
-        fin_quant(
-            step_n=step_n,
-            loop_n=loop_n,
-        )
+    try:
+        with _daily_session("fin_quant", **_ctx):
+            fin_quant(
+                step_n=step_n,
+                loop_n=loop_n,
+            )
+    finally:
+        if _log_f is not None:
+            sys.stdout = _orig_stdout
+            sys.stderr = _orig_stderr
+            _log_f.close()
 
 
 @app.command()
@@ -410,8 +425,8 @@ def evaluate(
         predix portfolio - Select a diversified portfolio of uncorrelated factors
         predix quant - Generate new factors via LLM trading loop
     """
-    from rich.panel import Panel
     from rdagent.log.daily_log import session as _daily_session
+    from rich.panel import Panel
 
     console.print(Panel(
         "[bold cyan]📊 Predix Factor Evaluator[/bold cyan]\n"
@@ -491,11 +506,12 @@ def top(
         predix portfolio - Select diversified portfolio from top factors
         predix build-strategies - Combine factors into trading strategies
     """
-    import json
     import glob as glob_module
+    import json
+
     import numpy as np
-    from rich.table import Table
     from rich.panel import Panel
+    from rich.table import Table
 
     factors_dir = Path(__file__).parent / "results" / "factors"
     if not factors_dir.exists():
@@ -560,9 +576,11 @@ def top(
 
     console.print(table)
 
-    # Summary
-    valid_ic = [r.get("ic") for r in results if r.get("ic") is not None]
-    valid_sharpe = [r.get("sharpe") for r in results if r.get("sharpe") is not None]
+    # Summary — filter None, NaN, and non-numeric values
+    valid_ic = [v for v in (r.get("ic") for r in results)
+                if isinstance(v, (int, float)) and v is not None and not np.isnan(v)]
+    valid_sharpe = [v for v in (r.get("sharpe") for r in results)
+                    if isinstance(v, (int, float)) and v is not None and not np.isnan(v)]
     # Filter extreme outliers for average
     valid_sharpe_filtered = [s for s in valid_sharpe if abs(s or 0) < 1e6]
 
@@ -637,16 +655,16 @@ def portfolio(
         predix top - View top factors before portfolio selection
         predix build-strategies - Build strategies from selected factors
     """
-    import json
     import glob as glob_module
+    import json
+    import shutil
     import subprocess
     import tempfile
-    import shutil
-    import numpy as np
+
     import pandas as pd
-    from rich.table import Table
     from rich.panel import Panel
-    from rich.progress import Progress, SpinnerColumn, TextColumn, BarColumn, TaskProgressColumn, TimeElapsedColumn
+    from rich.progress import BarColumn, Progress, SpinnerColumn, TaskProgressColumn, TextColumn, TimeElapsedColumn
+    from rich.table import Table
 
     factors_dir = Path(__file__).parent / "results" / "factors"
     if not factors_dir.exists():
@@ -678,12 +696,12 @@ def portfolio(
     # 2. Evaluate candidates to get time-series values for correlation
     # We need to run the factor code to get the series of values.
     # We do this sequentially to avoid OOM.
-    
+
     # Locate data file
     data_file = Path(__file__).parent / "git_ignore_folder" / "factor_implementation_source_data" / "intraday_pv.h5"
     if not data_file.exists():
         data_file = Path(__file__).parent / "git_ignore_folder" / "factor_implementation_source_data_debug" / "intraday_pv.h5"
-    
+
     if not data_file.exists():
         console.print("[red]Source data file (intraday_pv.h5) not found.[/red]")
         return
@@ -700,11 +718,11 @@ def portfolio(
         console=console,
     ) as progress:
         task = progress.add_task(f"Computing values for {len(candidates)} factors...", total=len(candidates))
-        
+
         for cand in candidates:
             fname = cand.get("factor_name", "unknown")
             fcode = cand.get("factor_code", "")
-            
+
             if not fcode:
                 errors.append((fname, "No code in JSON"))
                 progress.advance(task)
@@ -720,10 +738,10 @@ def portfolio(
                     # If symlink fails, copy the file
                     import shutil
                     shutil.copy(str(data_file), str(tmp_path / "intraday_pv.h5"))
-                
+
                 # Write code
                 (tmp_path / "factor.py").write_text(fcode)
-                
+
                 try:
                     # Run factor
                     result = subprocess.run(
@@ -731,16 +749,16 @@ def portfolio(
                         cwd=tmp_path,
                         capture_output=True,
                         text=True,
-                        timeout=120 # 2 min timeout per factor
+                        timeout=120, # 2 min timeout per factor
                     )
-                    
+
                     # Read result
                     res_file = tmp_path / "result.h5"
                     if res_file.exists():
                         df = pd.read_hdf(str(res_file), key="data")
                         # Get the series (first column)
                         series = df.iloc[:, 0]
-                        
+
                         # Count non-NaN values
                         non_nan = series.count()
                         if non_nan < 1000:
@@ -760,7 +778,7 @@ def portfolio(
                 except Exception as e:
                     errors.append((fname, str(e)[:100]))
                     progress.update(task, description=f"{fname} ❌ (Error)")
-            
+
             progress.advance(task)
 
     # Show summary of errors
@@ -774,7 +792,7 @@ def portfolio(
     if len(factor_series) < 3:
         console.print("[red]Not enough valid factor series to build portfolio (need at least 3).[/red]")
         console.print("[yellow]Tip: Factors might be producing mostly NaN values or failing execution.[/yellow]")
-        
+
         # Fallback: Show top factors by IC without diversification
         console.print("\n[dim]Showing top factors by IC instead:[/dim]")
         table = Table(
@@ -792,50 +810,49 @@ def portfolio(
                 str(i),
                 cand.get("factor_name", "unknown")[:38],
                 f"{cand.get('ic', 0):.6f}",
-                f"{cand.get('sharpe', 0):.4f}" if cand.get('sharpe') else "N/A",
+                f"{cand.get('sharpe', 0):.4f}" if cand.get("sharpe") else "N/A",
             )
-        
+
         console.print(table)
         return
 
     # 3. Build Correlation Matrix
     console.print(f"\n[dim]Building correlation matrix from {len(factor_series)} factors...[/dim]")
-    
+
     # Align indices and drop NaN
     combined = pd.DataFrame(factor_series).dropna()
-    
+
     if combined.empty or len(combined) < 100:
         console.print("[red]Not enough valid overlapping data to compute correlation.[/red]")
         console.print("[dim]This means the factors produce values at different times or have too many NaN values.[/dim]")
         return
 
     corr_matrix = combined.corr().fillna(0)
-    ic_map = {cand['factor_name']: cand.get('ic', 0) for cand in candidates}
+    ic_map = {cand["factor_name"]: cand.get("ic", 0) for cand in candidates}
 
     # 4. Greedy Selection
     selected = []
     remaining = list(corr_matrix.columns)
-    
+
     # Sort remaining by IC to prioritize high IC factors
     remaining.sort(key=lambda x: abs(ic_map.get(x, 0)), reverse=True)
 
     for factor in remaining:
         if len(selected) >= target:
             break
-        
+
         # If it's the first one, just take it
         if not selected:
             selected.append(factor)
             continue
-        
+
         # Check correlation with already selected
         # We want max(|corr|) < max_corr
         max_c = 0
         for sel in selected:
             c = abs(corr_matrix.loc[factor, sel])
-            if c > max_c:
-                max_c = c
-        
+            max_c = max(max_c, c)
+
         if max_c < max_corr:
             selected.append(factor)
 
@@ -853,23 +870,23 @@ def portfolio(
 
     for i, fname in enumerate(selected, 1):
         # Find original data for display
-        data = next((c for c in candidates if c['factor_name'] == fname), {})
-        ic = data.get('ic')
-        sharpe = data.get('sharpe')
-        
+        data = next((c for c in candidates if c["factor_name"] == fname), {})
+        ic = data.get("ic")
+        sharpe = data.get("sharpe")
+
         # Calculate max corr with other selected factors
         max_c_val = 0
         for s in selected:
             if s != fname:
                 val = abs(corr_matrix.loc[fname, s])
-                if val > max_c_val: max_c_val = val
+                max_c_val = max(max_c_val, val)
 
         table.add_row(
             str(i),
             fname[:38],
             f"{ic:.6f}" if ic is not None else "N/A",
             f"{sharpe:.4f}" if sharpe is not None else "N/A",
-            f"{max_c_val:.4f}" if max_c_val > 0 else "-"
+            f"{max_c_val:.4f}" if max_c_val > 0 else "-",
         )
 
     console.print(table)
@@ -879,20 +896,20 @@ def portfolio(
         "selected_factors": selected,
         "max_correlation": max_corr,
         "pool_size": top,
-        "timestamp": pd.Timestamp.now().isoformat()
+        "timestamp": pd.Timestamp.now().isoformat(),
     }
-    
+
     out_dir = Path(__file__).parent / "results" / "portfolio"
     out_dir.mkdir(parents=True, exist_ok=True)
     out_file = out_dir / "selected_factors.json"
-    
+
     with open(out_file, "w") as f:
         json.dump(portfolio_data, f, indent=2)
-    
+
     console.print(Panel(
         f"[bold]Portfolio saved to results/portfolio/selected_factors.json[/bold]\n"
         f"Selected {len(selected)} unique factors from {top} candidates.",
-        border_style="green"
+        border_style="green",
     ))
 
 
@@ -938,13 +955,12 @@ def portfolio_simple(
         predix top - View top factors before portfolio selection
         predix build-strategies - Build strategies from selected factors
     """
-    import json
     import glob as glob_module
-    import re
-    import numpy as np
+    import json
+
     import pandas as pd
-    from rich.table import Table
     from rich.panel import Panel
+    from rich.table import Table
 
     factors_dir = Path(__file__).parent / "results" / "factors"
     if not factors_dir.exists():
@@ -988,14 +1004,14 @@ def portfolio_simple(
     for cand in candidates:
         fname = cand.get("factor_name", "").lower()
         assigned = False
-        
+
         # Check each category's keywords
         for cat, keywords in categories.items():
             if any(kw in fname for kw in keywords):
                 categorized[cat].append(cand)
                 assigned = True
                 break
-        
+
         if not assigned:
             categorized["other"].append(cand)
 
@@ -1006,7 +1022,7 @@ def portfolio_simple(
             best = categorized[cat][0]  # Already sorted by IC
             selected.append({
                 "factor": best,
-                "category": cat.capitalize() if cat != "other" else "Other"
+                "category": cat.capitalize() if cat != "other" else "Other",
             })
 
     # 5. Display Results
@@ -1029,7 +1045,7 @@ def portfolio_simple(
             cand.get("factor_name", "unknown")[:38],
             cat,
             f"{cand.get('ic', 0):.6f}",
-            f"{cand.get('sharpe', 0):.4f}" if cand.get('sharpe') else "N/A",
+            f"{cand.get('sharpe', 0):.4f}" if cand.get("sharpe") else "N/A",
         )
 
     console.print(table)
@@ -1039,7 +1055,7 @@ def portfolio_simple(
         "selected_factors": [item["factor"]["factor_name"] for item in selected],
         "categories": {item["category"]: item["factor"]["factor_name"] for item in selected},
         "method": "simple_keyword_categorization",
-        "timestamp": str(pd.Timestamp.now().isoformat())
+        "timestamp": str(pd.Timestamp.now().isoformat()),
     }
 
     out_dir = Path(__file__).parent / "results" / "portfolio"
@@ -1052,7 +1068,7 @@ def portfolio_simple(
     console.print(Panel(
         f"[bold]Simple Portfolio saved to results/portfolio/portfolio_simple.json[/bold]\n"
         f"Selected {len(selected)} factors across {len([c for c in categorized if categorized[c]])} categories.",
-        border_style="green"
+        border_style="green",
     ))
 
 
@@ -1115,12 +1131,10 @@ def build_strategies(
         predix portfolio - Select diversified factors before combining
         predix top - View top factors before building strategies
     """
-    import pandas as pd
     import numpy as np
-    from rich.table import Table
-    from rich.panel import Panel
-
     from rdagent.scenarios.qlib.developer.strategy_builder import StrategyBuilder
+    from rich.panel import Panel
+    from rich.table import Table
 
     console.print(Panel(
         "[bold cyan]🏗️  Predix Strategy Builder[/bold cyan]\n"
@@ -1276,9 +1290,10 @@ def build_strategies_ai(
         predix quant - Generate new alpha factors via LLM trading loop
         predix evaluate - Evaluate factors before strategy building
     """
-    from rich.panel import Panel
     from pathlib import Path
 
+    from rich.panel import Panel
+
     console.print(Panel(
         "[bold cyan]🧠 StrategyCoSTEER - AI Strategy Builder[/bold cyan]\n"
         "Generating trading strategies from existing factors\n"
@@ -1304,7 +1319,7 @@ def build_strategies_ai(
     # Setup LLM environment (same as quant command)
     api_key = os.getenv("OPENROUTER_API_KEY") or os.getenv("OPENAI_API_KEY", "")
     api_key_2 = os.getenv("OPENROUTER_API_KEY_2", "")
-    
+
     if api_key and not api_key.startswith("sk-or-"):
         # OPENROUTER_API_KEY not set, try to use what we have
         api_key = os.getenv("OPENROUTER_API_KEY", api_key)
@@ -1331,8 +1346,8 @@ def build_strategies_ai(
         return
 
     # Load evaluated factors
-    import json
     import glob as glob_module
+    import json
 
     factors = []
     for f in glob_module.glob(str(factors_dir / "*.json")):
@@ -1416,15 +1431,15 @@ def build_strategies_ai(
 
             for i, r in enumerate(results, 1):
                 # Monthly return: use real backtest if available, else estimate
-                rb = r.get('real_backtest', {})
-                if isinstance(rb, dict) and rb.get('status') == 'success':
-                    monthly_pct = rb.get('monthly_return_pct', r.get('monthly_return_pct', 0))
-                    n_trades = rb.get('n_trades', '-')
-                    real_ic = rb.get('ic', 0)
+                rb = r.get("real_backtest", {})
+                if isinstance(rb, dict) and rb.get("status") == "success":
+                    monthly_pct = rb.get("monthly_return_pct", r.get("monthly_return_pct", 0))
+                    n_trades = rb.get("n_trades", "-")
+                    real_ic = rb.get("ic", 0)
                 else:
-                    monthly_pct = r.get('monthly_return_pct', r.get('real_monthly_return', 0))
-                    n_trades = '-'
-                    real_ic = rb.get('ic', 0) if isinstance(rb, dict) else 0
+                    monthly_pct = r.get("monthly_return_pct", r.get("real_monthly_return", 0))
+                    n_trades = "-"
+                    real_ic = rb.get("ic", 0) if isinstance(rb, dict) else 0
 
                 table.add_row(
                     str(i),
@@ -1517,7 +1532,7 @@ def status():
     # Process check
     result = subprocess.run(
         ["pgrep", "-f", "fin_quant"],
-        capture_output=True, text=True
+        capture_output=True, text=True,
     )
     if result.returncode == 0:
         console.print("[bold green]✅ Trading Loop: RUNNING[/bold green]")
@@ -1535,7 +1550,7 @@ def status():
         factors = c.fetchone()[0]
         conn.close()
 
-        console.print(f"\n📊 Results:")
+        console.print("\n📊 Results:")
         console.print(f"  Backtest runs: {runs}")
         console.print(f"  Factors: {factors}")
 
@@ -1612,6 +1627,7 @@ def best(
         $ predix best -n 50 --export /tmp/top.json
     """
     import json
+
     from rich.table import Table
 
     items = _load_strategies()
@@ -1728,7 +1744,7 @@ def kronos_factor(
         console.print("Run data conversion first — see README Data Setup section.")
         raise typer.Exit(1)
 
-    console.print(f"[bold]Kronos Factor Generator[/bold]")
+    console.print("[bold]Kronos Factor Generator[/bold]")
     console.print(f"  Context: [cyan]{context}[/cyan] bars  |  Pred: [cyan]{pred}[/cyan] bars  |  Device: [cyan]{_device}[/cyan]")
 
     from rdagent.components.coder.kronos_adapter import build_kronos_factor
@@ -1811,7 +1827,7 @@ def kronos_eval(
         console.print(f"[red]ERROR: Data not found at {data_path}[/red]")
         raise typer.Exit(1)
 
-    console.print(f"[bold]Kronos Model Evaluator[/bold] (alongside LightGBM)")
+    console.print("[bold]Kronos Model Evaluator[/bold] (alongside LightGBM)")
     console.print(f"  Context: [cyan]{context}[/cyan] bars  |  Pred: [cyan]{pred}[/cyan] bars  |  Device: [cyan]{_device}[/cyan]")
     console.print("  Running evaluation...")
 
@@ -1826,12 +1842,12 @@ def kronos_eval(
         batch_size=batch_size,
     )
 
-    console.print(f"\n[bold]Kronos-mini Results[/bold]")
+    console.print("\n[bold]Kronos-mini Results[/bold]")
     console.print(f"  Predictions: [cyan]{metrics['n_predictions']}[/cyan]")
     console.print(f"  IC (mean):   [{'green' if metrics['IC_mean'] > 0.02 else 'yellow'}]{metrics['IC_mean']:.4f}[/]")
     console.print(f"  IC IR:       [{'green' if metrics['IC_IR'] > 0.5 else 'yellow'}]{metrics['IC_IR']:.4f}[/]  (>0.5 = strong signal)")
     console.print(f"  Hit Rate:    [{'green' if metrics['hit_rate'] > 0.52 else 'yellow'}]{metrics['hit_rate']:.2%}[/]  (>50% = directionally useful)")
-    console.print(f"\n[dim]Reference: LightGBM baseline IC typically 0.01–0.05 on 1-min EUR/USD[/dim]")
+    console.print("\n[dim]Reference: LightGBM baseline IC typically 0.01–0.05 on 1-min EUR/USD[/dim]")
 
     import json as _json
     out_dir = Path("results/kronos")

rdagent/app/cli.py

@@ -21,13 +21,17 @@ load_dotenv(".env")
 
 import subprocess
 from importlib.resources import path as rpath
-from typing import Dict, Optional
+from typing import Annotated
 
 import typer
 from rich.console import Console
-from typing_extensions import Annotated
 
-from rdagent.utils.env import logger
+try:
+    from rdagent.utils.env import logger
+except ImportError:
+    import logging
+
+    logger = logging.getLogger(__name__)
 
 from rdagent.app.data_science.loop import main as data_science
 from rdagent.app.finetune.llm.loop import main as llm_finetune
@@ -141,10 +145,10 @@ def ds_user_interact(port=19900):
 
 @app.command(name="fin_factor")
 def fin_factor_cli(
-    path: Optional[str] = None,
-    step_n: Optional[int] = None,
-    loop_n: Optional[int] = None,
-    all_duration: Optional[str] = None,
+    path: str | None = None,
+    step_n: int | None = None,
+    loop_n: int | None = None,
+    all_duration: str | None = None,
     checkout: CheckoutOption = True,
 ):
     fin_factor(path=path, step_n=step_n, loop_n=loop_n, all_duration=all_duration, checkout=checkout)
@@ -152,10 +156,10 @@ def fin_factor_cli(
 
 @app.command(name="fin_model")
 def fin_model_cli(
-    path: Optional[str] = None,
-    step_n: Optional[int] = None,
-    loop_n: Optional[int] = None,
-    all_duration: Optional[str] = None,
+    path: str | None = None,
+    step_n: int | None = None,
+    loop_n: int | None = None,
+    all_duration: str | None = None,
     checkout: CheckoutOption = True,
 ):
     fin_model(path=path, step_n=step_n, loop_n=loop_n, all_duration=all_duration, checkout=checkout)
@@ -163,10 +167,10 @@ def fin_model_cli(
 
 @app.command(name="fin_quant")
 def fin_quant_cli(
-    path: Optional[str] = None,
-    step_n: Optional[int] = None,
-    loop_n: Optional[int] = None,
-    all_duration: Optional[str] = None,
+    path: str | None = None,
+    step_n: int | None = None,
+    loop_n: int | None = None,
+    all_duration: str | None = None,
     checkout: CheckoutOption = True,
     with_dashboard: bool = typer.Option(False, "--with-dashboard/-d", help="Start web dashboard automatically"),
     with_cli_dashboard: bool = typer.Option(False, "--cli-dashboard/-c", help="Show beautiful CLI dashboard"),
@@ -226,7 +230,7 @@ def fin_quant_cli(
         if not api_key:
             console.print("\n[bold red]❌ OPENROUTER_API_KEY not set in .env[/bold red]")
             console.print("[yellow]Add your API key to .env and retry:[/yellow]")
-            console.print('  OPENROUTER_API_KEY=sk-or-your-key-here')
+            console.print("  OPENROUTER_API_KEY=sk-or-your-key-here")
             raise typer.Exit(code=1)
 
         os.environ["OPENAI_API_KEY"] = api_key
@@ -245,8 +249,8 @@ def fin_quant_cli(
         console.print(f"   [dim]Base URL: {os.environ['OPENAI_API_BASE']}[/dim]")
 
         # Wait until the llama.cpp server is fully loaded before starting the pipeline
-        import urllib.request
         import urllib.error
+        import urllib.request
 
         base_url = os.environ["OPENAI_API_BASE"].removesuffix("/v1").rstrip("/")
         health_url = f"{base_url}/health"
@@ -280,7 +284,7 @@ def fin_quant_cli(
             subprocess.run(
                 ["python", "web/dashboard_api.py"],
                 cwd=str(Path(__file__).parent.parent.parent),
-                env={**os.environ, "FLASK_ENV": "development"}
+                env={**os.environ, "FLASK_ENV": "development"},
             )
 
         dashboard_thread = threading.Thread(target=start_web_dashboard, daemon=True)
@@ -322,9 +326,9 @@ def fin_quant_cli(
 
 @app.command(name="fin_factor_report")
 def fin_factor_report_cli(
-    report_folder: Optional[str] = None,
-    path: Optional[str] = None,
-    all_duration: Optional[str] = None,
+    report_folder: str | None = None,
+    path: str | None = None,
+    all_duration: str | None = None,
     checkout: CheckoutOption = True,
 ):
     fin_factor_report(report_folder=report_folder, path=path, all_duration=all_duration, checkout=checkout)
@@ -337,12 +341,12 @@ def general_model_cli(report_file_path: str):
 
 @app.command(name="data_science")
 def data_science_cli(
-    path: Optional[str] = None,
+    path: str | None = None,
     checkout: CheckoutOption = True,
-    step_n: Optional[int] = None,
-    loop_n: Optional[int] = None,
-    timeout: Optional[str] = None,
-    competition: Optional[str] = None,
+    step_n: int | None = None,
+    loop_n: int | None = None,
+    timeout: str | None = None,
+    competition: str | None = None,
 ):
     data_science(
         path=path,
@@ -356,16 +360,16 @@ def data_science_cli(
 
 @app.command(name="llm_finetune")
 def llm_finetune_cli(
-    path: Optional[str] = None,
+    path: str | None = None,
     checkout: CheckoutOption = True,
-    benchmark: Optional[str] = None,
-    benchmark_description: Optional[str] = None,
-    dataset: Optional[str] = None,
-    base_model: Optional[str] = None,
-    upper_data_size_limit: Optional[int] = None,
-    step_n: Optional[int] = None,
-    loop_n: Optional[int] = None,
-    timeout: Optional[str] = None,
+    benchmark: str | None = None,
+    benchmark_description: str | None = None,
+    dataset: str | None = None,
+    base_model: str | None = None,
+    upper_data_size_limit: int | None = None,
+    step_n: int | None = None,
+    loop_n: int | None = None,
+    timeout: str | None = None,
 ):
     llm_finetune(
         path=path,
@@ -431,6 +435,7 @@ def rl_trading_cli(
         rdagent rl_trading --mode backtest --no-with-protections
     """
     from pathlib import Path
+
     import yaml
 
     console = Console()
@@ -442,18 +447,18 @@ def rl_trading_cli(
         with open(config_path) as f:
             config = yaml.safe_load(f) or {}
 
-    console.print(f"\n[bold blue]🤖 RL Trading Agent[/bold blue]")
+    console.print("\n[bold blue]🤖 RL Trading Agent[/bold blue]")
     console.print(f"Mode: [cyan]{mode}[/cyan]")
     console.print(f"Algorithm: [cyan]{algorithm.upper()}[/cyan]")
     console.print(f"Protections: {'[green]Enabled[/green]' if with_protections else '[red]Disabled[/red]'}")
 
     try:
-        from rdagent.components.coder.rl import RLTradingAgent, RLCosteer, TradingEnv
+        from rdagent.components.coder.rl import RLCosteer, RLTradingAgent, TradingEnv
     except ImportError as e:
-        console.print(f"[bold red]Error: RL components not available.[/bold red]")
+        console.print("[bold red]Error: RL components not available.[/bold red]")
         console.print(f"Details: {e}")
-        console.print(f"\n[yellow]Install RL dependencies:[/yellow]")
-        console.print(f"  pip install stable-baselines3 gymnasium")
+        console.print("\n[yellow]Install RL dependencies:[/yellow]")
+        console.print("  pip install stable-baselines3 gymnasium")
         raise typer.Exit(code=1)
 
     if mode == "train":
@@ -469,8 +474,8 @@ def rl_trading_cli(
             console.print("[dim]Loading market data...[/dim]")
             # TODO: Load actual data from config
             # For now, create mock environment
-            import numpy as np
             import gymnasium as gym
+            import numpy as np
 
             # Create simple mock environment for demonstration
             class MockTradingEnv(gym.Env):
@@ -506,7 +511,7 @@ def rl_trading_cli(
             model_path_out.parent.mkdir(parents=True, exist_ok=True)
             agent.save(model_path_out)
 
-            console.print(f"\n[bold green]✅ Training complete![/bold green]")
+            console.print("\n[bold green]✅ Training complete![/bold green]")
             console.print(f"Model saved to: [cyan]{model_path_out}[/cyan]")
             console.print(f"Algorithm: {result['algorithm']}")
             console.print(f"Timesteps: {result['total_timesteps']:,}")
@@ -532,9 +537,9 @@ def rl_trading_cli(
                 agent = RLTradingAgent(algorithm=algorithm.upper())
 
             # Run backtest
-            from rdagent.components.backtesting import FactorBacktester
-            import pandas as pd
             import numpy as np
+            import pandas as pd
+            from rdagent.components.backtesting import FactorBacktester
 
             backtester = FactorBacktester()
 
@@ -543,8 +548,8 @@ def rl_trading_cli(
             n_steps = 500
             mock_prices = pd.Series(100 + np.cumsum(np.random.randn(n_steps) * 0.5))
             mock_indicators = pd.DataFrame({
-                'rsi': np.random.uniform(30, 70, n_steps),
-                'macd': np.random.randn(n_steps) * 0.1,
+                "rsi": np.random.uniform(30, 70, n_steps),
+                "macd": np.random.randn(n_steps) * 0.1,
             })
 
             console.print("[yellow]Running backtest...[/yellow]")
@@ -555,7 +560,7 @@ def rl_trading_cli(
                 enable_protections=with_protections,
             )
 
-            console.print(f"\n[bold green]✅ Backtest complete![/bold green]")
+            console.print("\n[bold green]✅ Backtest complete![/bold green]")
             console.print(f"  Final Equity: [green]${metrics.get('final_equity', 0):,.2f}[/green]")
             console.print(f"  Sharpe Ratio: {metrics.get('sharpe_ratio', 0):.3f}")
             console.print(f"  Max Drawdown: {metrics.get('max_drawdown', 0):.2%}")
@@ -629,7 +634,7 @@ def generate_strategies_cli(
         rdagent generate_strategies -n 3 -i 10 --optuna-trials 50  # Deep optimization
     """
     from rich.console import Console
-    from rich.progress import Progress, SpinnerColumn, TextColumn, BarColumn, TimeRemainingColumn
+    from rich.progress import BarColumn, Progress, SpinnerColumn, TextColumn, TimeRemainingColumn
     from rich.table import Table
 
     console = Console()
@@ -648,7 +653,7 @@ def generate_strategies_cli(
         raise typer.Exit(code=1)
 
     console.print(f"\n[bold blue]{'='*60}[/bold blue]")
-    console.print(f"[bold blue]  PREDIX Strategy Generator[/bold blue]")
+    console.print("[bold blue]  PREDIX Strategy Generator[/bold blue]")
     console.print(f"[bold blue]{'='*60}[/bold blue]")
     console.print(f"  Strategies:  [cyan]{count}[/cyan]")
     console.print(f"  Workers:     [cyan]{workers}[/cyan]")
@@ -675,12 +680,12 @@ def generate_strategies_cli(
     _slog = _dlog.setup("strategies", **_strat_ctx)
 
     try:
-        from rdagent.components.coder.strategy_orchestrator import StrategyOrchestrator
         import pandas as pd
+        from rdagent.components.coder.strategy_orchestrator import StrategyOrchestrator
 
         all_results = []
         best_strategy = None
-        best_sharpe = float('-inf')
+        best_sharpe = float("-inf")
 
         # CONTINUOUS OPTIMIZATION LOOP
         for iteration in range(1, max_iterations + 1):
@@ -729,7 +734,7 @@ def generate_strategies_cli(
 
             # Track best strategy
             for r in results:
-                sharpe = r.get("sharpe_ratio", float('-inf'))
+                sharpe = r.get("sharpe_ratio", float("-inf"))
                 if sharpe > best_sharpe:
                     best_sharpe = sharpe
                     best_strategy = r
@@ -749,7 +754,7 @@ def generate_strategies_cli(
         rejected = [r for r in results if r.get("status") == "rejected"]
 
         console.print(f"\n[bold green]{'='*60}[/bold green]")
-        console.print(f"[bold green]  Strategy Generation Summary[/bold green]")
+        console.print("[bold green]  Strategy Generation Summary[/bold green]")
         console.print(f"[bold green]{'='*60}[/bold green]")
 
         table = Table(show_header=True, header_style="bold magenta", show_lines=True)
@@ -778,7 +783,7 @@ def generate_strategies_cli(
         # Show best strategy details
         if best_strategy:
             console.print(f"\n[bold gold1]{'='*60}[/bold gold1]")
-            console.print(f"[bold gold1]  BEST STRATEGY[/bold gold1]")
+            console.print("[bold gold1]  BEST STRATEGY[/bold gold1]")
             console.print(f"[bold gold1]{'='*60}[/bold gold1]")
             console.print(f"  Name:        [cyan]{best_strategy.get('strategy_name', 'Unknown')}[/cyan]")
             console.print(f"  Sharpe:      [green]{best_strategy.get('sharpe_ratio', 0):.4f}[/green]")
@@ -786,13 +791,13 @@ def generate_strategies_cli(
             console.print(f"  Max DD:      [yellow]{best_strategy.get('max_drawdown', 0):.2%}[/yellow]")
             console.print(f"  Win Rate:    [cyan]{best_strategy.get('win_rate', 0):.2%}[/cyan]")
             if best_strategy.get("best_params"):
-                console.print(f"\n  [bold]Optimized Parameters:[/bold]")
+                console.print("\n  [bold]Optimized Parameters:[/bold]")
                 for param, val in best_strategy["best_params"].items():
                     console.print(f"    {param}: [cyan]{val}[/cyan]")
             console.print(f"[bold gold1]{'='*60}[/bold gold1]")
 
         if accepted:
-            console.print(f"\n[bold]Accepted Strategies:[/bold]")
+            console.print("\n[bold]Accepted Strategies:[/bold]")
             acc_table = Table(show_header=True, header_style="bold cyan")
             acc_table.add_column("#", width=4)
             acc_table.add_column("Strategy", width=30)
@@ -815,13 +820,13 @@ def generate_strategies_cli(
                 )
             console.print(acc_table)
 
-        console.print(f"\n[bold green]Strategies saved to:[/bold green] [cyan]results/strategies_new/[/cyan]")
+        console.print("\n[bold green]Strategies saved to:[/bold green] [cyan]results/strategies_new/[/cyan]")
         console.print(f"[bold blue]{'='*60}[/bold blue]\n")
         _slog.success(f"Generated {len(all_results)} strategies ({len([r for r in all_results if r.get('status')=='accepted'])} accepted)")
 
     except ImportError as e:
         _slog.error(f"Strategy components not available: {e}")
-        console.print(f"[bold red]Error: Strategy components not available.[/bold red]")
+        console.print("[bold red]Error: Strategy components not available.[/bold red]")
         console.print(f"Details: {e}")
         raise typer.Exit(code=1)
     except Exception as e:
@@ -857,17 +862,18 @@ def optimize_portfolio_cli(
         raise typer.Exit(code=1)
 
     console.print(f"\n[bold blue]{'='*60}[/bold blue]")
-    console.print(f"[bold blue]  PREDIX Portfolio Optimizer[/bold blue]")
+    console.print("[bold blue]  PREDIX Portfolio Optimizer[/bold blue]")
     console.print(f"[bold blue]{'='*60}[/bold blue]")
     console.print(f"  Top N:    [cyan]{top_n}[/cyan]")
     console.print(f"  Method:   [cyan]{method}[/cyan]")
     console.print(f"[bold blue]{'='*60}[/bold blue]\n")
 
     try:
-        from rdagent.components.backtesting.risk_management import PortfolioOptimizer
         import json
         from pathlib import Path
 
+        from rdagent.components.backtesting.risk_management import PortfolioOptimizer
+
         project_root = Path(__file__).parent.parent.parent
         strategies_dir = project_root / "results" / "strategies_new"
 
@@ -1004,14 +1010,15 @@ def strategies_report_cli(
         rdagent strategies_report -s path/to/strategy.json  # Single strategy
         rdagent strategies_report -o custom/reports/      # Custom output dir
     """
+    from pathlib import Path
+
     from rich.console import Console
     from rich.progress import Progress, SpinnerColumn, TextColumn
-    from pathlib import Path
 
     console = Console()
 
     console.print(f"\n[bold blue]{'='*60}[/bold blue]")
-    console.print(f"[bold blue]  PREDIX Strategy Report Generator[/bold blue]")
+    console.print("[bold blue]  PREDIX Strategy Report Generator[/bold blue]")
     console.print(f"[bold blue]{'='*60}[/bold blue]\n")
 
     project_root = Path(__file__).parent.parent.parent
@@ -1063,20 +1070,20 @@ def strategies_report_cli(
                 progress.update(task, completed=1)
 
     console.print(f"\n[bold green]{'='*60}[/bold green]")
-    console.print(f"[bold green]  Report Generation Complete[/bold green]")
+    console.print("[bold green]  Report Generation Complete[/bold green]")
     console.print(f"[bold green]{'='*60}[/bold green]")
     console.print(f"  Reports generated: [cyan]{reports_generated}/{len(strategy_files)}[/cyan]")
     console.print(f"  Output directory:  [cyan]{output_dir_path}[/cyan]")
     console.print(f"[bold green]{'='*60}[/bold green]\n")
 
 
-def _generate_single_strategy_report(strategy_file: Path, output_dir: Path) -> Dict:
+def _generate_single_strategy_report(strategy_file: Path, output_dir: Path) -> dict:
     """Generate a report for a single strategy."""
     import json
+
     import matplotlib
     matplotlib.use("Agg")  # Non-interactive backend
     import matplotlib.pyplot as plt
-    import seaborn as sns
 
     with open(strategy_file, encoding="utf-8") as f:
         strategy = json.load(f)
@@ -1151,7 +1158,7 @@ if __name__ == "__main__":
 @app.command(name="start_llama")
 def start_llama_cli(
     model: str = typer.Option(
-        None, "--model", "-m", help="Path to model file"
+        None, "--model", "-m", help="Path to model file",
     ),
     port: int = typer.Option(8081, "--port", "-p", help="Server port"),
     gpu_layers: int = typer.Option(30, "--gpu-layers", "-g", help="GPU layers"),
@@ -1173,8 +1180,6 @@ def start_llama_cli(
       rdagent start_llama --gpu-layers 40 --ctx-size 4096
       rdagent start_llama --reasoning
     """
-    import subprocess
-    import sys
     import os
 
     model_path = model or os.getenv(
@@ -1211,7 +1216,7 @@ def start_llama_cli(
     if not reasoning:
         cmd.extend(["--reasoning", "off"])
 
-    print(f"🚀 Starting llama.cpp server...")
+    print("🚀 Starting llama.cpp server...")
     print(f"   Model: {Path(model_path).name}")
     print(f"   Port: {port}")
     print(f"   GPU Layers: {gpu_layers}")
@@ -1244,15 +1249,14 @@ def start_loop_cli(
       rdagent start_loop
       rdagent start_loop --target 5 --max-wait 3600
     """
-    import subprocess
-    import signal
-    import sys
     import os
-    from datetime import datetime
+    import signal
+    import subprocess
     import time
+    from datetime import datetime
 
-    script_dir = str(Path(__file__).parent.parent.parent.parent)
-    generator = f"python {script_dir}/scripts/predix_smart_strategy_gen.py"
+    script_dir = str(Path(__file__).parent.parent.parent)
+    generator = [sys.executable, f"{script_dir}/scripts/predix_smart_strategy_gen.py"]
     logfile = f"{script_dir}/results/logs/generator_loop.log"
     pidfile = "/tmp/predix_loop.pid"  # nosec B108 — administrative PID file, single-process daemon
 
@@ -1265,12 +1269,19 @@ def start_loop_cli(
         with open(logfile, "a") as f:
             f.write(line + "\n")
 
+    child_proc = None  # track current child PID for targeted cleanup
+
     def cleanup(signum=None, frame=None):
         log("Received termination signal. Cleaning up...")
-        try:
-            subprocess.run(["pkill", "-f", "predix_smart_strategy_gen.py"], capture_output=True)
-        except Exception:
-            pass
+        if child_proc is not None:
+            try:
+                child_proc.terminate()
+                child_proc.wait(timeout=10)
+            except Exception:
+                try:
+                    child_proc.kill()
+                except Exception:
+                    pass
         try:
             os.remove(pidfile)
         except FileNotFoundError:
@@ -1316,26 +1327,32 @@ def start_loop_cli(
         strat_count = len(list(strat_dir.glob("*.json"))) if strat_dir.exists() else 0
         log(f"📁 Existing strategies: {strat_count}")
 
-        # Kill stale processes
-        try:
-            subprocess.run(["pkill", "-9", "-f", "predix_smart_strategy_gen.py"], capture_output=True)
-        except Exception:
-            pass
-        time.sleep(2)
+        # Kill stale child from previous iteration
+        if child_proc is not None:
+            try:
+                child_proc.terminate()
+                child_proc.wait(timeout=10)
+            except subprocess.TimeoutExpired:
+                child_proc.kill()
+                child_proc.wait()
+            except Exception:
+                pass
+            child_proc = None
+            time.sleep(2)
 
         # Start generator
         log("🤖 Starting generator...")
-        proc = subprocess.Popen(
-            generator.split(),
+        child_proc = subprocess.Popen(
+            generator,
             cwd=script_dir,
             stdout=subprocess.DEVNULL,
             stderr=subprocess.DEVNULL,
         )
-        log(f"   PID: {proc.pid}")
+        log(f"   PID: {child_proc.pid}")
 
         # Monitor progress
         elapsed = 0
-        while proc.poll() is None:
+        while child_proc.poll() is None:
             time.sleep(30)
             elapsed += 30
 
@@ -1344,11 +1361,12 @@ def start_loop_cli(
 
             if elapsed >= max_wait:
                 log(f"   ⏰ Timeout after {elapsed}s. Killing...")
-                proc.kill()
+                child_proc.kill()
                 break
 
         # Check results
-        exit_code = proc.wait()
+        exit_code = child_proc.wait()
+        child_proc = None
         if exit_code == 0:
             log("✅ Generator completed successfully")
         elif exit_code == -9:
@@ -1389,24 +1407,24 @@ def parallel_cli(
       rdagent parallel -n 10 -k 2
     """
     import subprocess
-    import sys
     from pathlib import Path
+
     from rdagent.log import daily_log as _dlog
 
-    project_root = Path(__file__).parent.parent.parent.parent
+    project_root = Path(__file__).parent.parent.parent
     script = project_root / "scripts" / "predix_parallel.py"
 
     if not script.exists():
         typer.echo(f"❌ Script not found: {script}")
         raise typer.Exit(code=1)
 
-    cmd = [sys.executable, str(script), "--runs", str(runs), "--api-keys", str(api_keys), "-m", "local"]
+    cmd = [sys.executable, str(script), "--runs", str(runs), "--api-keys", str(api_keys)]
 
     _plog = _dlog.setup("parallel", runs=runs, api_keys=api_keys, model="local")
     typer.echo(f"🚀 Starting {runs} parallel runs...")
     typer.echo(f"   Script: {script}")
     typer.echo(f"   API Keys: {api_keys}")
-    typer.echo(f"   Model: local (llama.cpp)")
+    typer.echo("   Model: local (llama.cpp)")
 
     try:
         result = subprocess.run(cmd, cwd=str(project_root))
@@ -1440,11 +1458,11 @@ def eval_all_cli(
       rdagent eval_all -n 500 -p 8
     """
     import subprocess
-    import sys
     from pathlib import Path
+
     from rdagent.log import daily_log as _dlog
 
-    project_root = Path(__file__).parent.parent.parent.parent
+    project_root = Path(__file__).parent.parent.parent
     script = project_root / "scripts" / "predix_full_eval.py"
 
     if not script.exists():
@@ -1495,10 +1513,9 @@ def batch_backtest_cli(
       rdagent batch_backtest --all
     """
     import subprocess
-    import sys
     from pathlib import Path
 
-    project_root = Path(__file__).parent.parent.parent.parent
+    project_root = Path(__file__).parent.parent.parent
     script = project_root / "scripts" / "predix_batch_backtest.py"
 
     if not script.exists():
@@ -1548,10 +1565,9 @@ def simple_eval_cli(
       rdagent simple_eval --all
     """
     import subprocess
-    import sys
     from pathlib import Path
 
-    project_root = Path(__file__).parent.parent.parent.parent
+    project_root = Path(__file__).parent.parent.parent
     script = project_root / "scripts" / "predix_simple_eval.py"
 
     if not script.exists():
@@ -1581,7 +1597,7 @@ def simple_eval_cli(
 @app.command(name="rebacktest")
 def rebacktest_cli(
     strategies_dir: str = typer.Option(
-        None, "--strategies-dir", "-d", help="Directory containing strategy JSON files"
+        None, "--strategies-dir", "-d", help="Directory containing strategy JSON files",
     ),
 ):
     """
@@ -1595,10 +1611,9 @@ def rebacktest_cli(
       rdagent rebacktest -d results/strategies_new/
     """
     import subprocess
-    import sys
     from pathlib import Path
 
-    project_root = Path(__file__).parent.parent.parent.parent
+    project_root = Path(__file__).parent.parent.parent
     script = project_root / "scripts" / "predix_rebacktest_strategies.py"
 
     if not script.exists():
@@ -1623,10 +1638,10 @@ def rebacktest_cli(
 @app.command(name="report")
 def report_cli(
     strategy_path: str = typer.Option(
-        None, "--strategy", "-s", help="Path to single strategy JSON (default: all strategies)"
+        None, "--strategy", "-s", help="Path to single strategy JSON (default: all strategies)",
     ),
     output: str = typer.Option(
-        None, "--output", "-o", help="Output directory (default: results/strategy_reports/)"
+        None, "--output", "-o", help="Output directory (default: results/strategy_reports/)",
     ),
 ):
     """
@@ -1649,10 +1664,9 @@ def report_cli(
       rdagent report -o custom/reports/
     """
     import subprocess
-    import sys
     from pathlib import Path
 
-    project_root = Path(__file__).parent.parent.parent.parent
+    project_root = Path(__file__).parent.parent.parent
     script = project_root / "scripts" / "predix_strategy_report.py"
 
     if not script.exists():

rdagent/app/data_science/conf.py

@@ -201,6 +201,5 @@ class DataScienceBasePropSetting(KaggleBasePropSetting):
 DS_RD_SETTING = DataScienceBasePropSetting()
 
 # enable_cross_trace_diversity and llm_select_hypothesis should not be true at the same time
-assert not (
-    DS_RD_SETTING.enable_cross_trace_diversity and DS_RD_SETTING.llm_select_hypothesis
-), "enable_cross_trace_diversity and llm_select_hypothesis cannot be true at the same time"
+if DS_RD_SETTING.enable_cross_trace_diversity and DS_RD_SETTING.llm_select_hypothesis:
+    raise ValueError("enable_cross_trace_diversity and llm_select_hypothesis cannot be true at the same time")

rdagent/app/finetune/llm/loop.py

@@ -58,18 +58,18 @@ def main(
 
     if user_target_scenario:
         FT_RD_SETTING.user_target_scenario = user_target_scenario
-    assert (
-        FT_RD_SETTING.user_target_scenario is None
-    ), "user_target_scenario is not yet supported, please specify via benchmark and benchmark_description"
+    if FT_RD_SETTING.user_target_scenario is not None:
+        raise ValueError("user_target_scenario is not yet supported, please specify via benchmark and benchmark_description")
     if upper_data_size_limit:
         FT_RD_SETTING.upper_data_size_limit = upper_data_size_limit
         logger.info(f"Set upper_data_size_limit to {FT_RD_SETTING.upper_data_size_limit}")
     if benchmark and benchmark_description:
         FT_RD_SETTING.target_benchmark = benchmark
         FT_RD_SETTING.benchmark_description = benchmark_description
-    assert FT_RD_SETTING.user_target_scenario or (
-        FT_RD_SETTING.target_benchmark and FT_RD_SETTING.benchmark_description
-    ), "Either user_target_scenario or target_benchmark must be specified for LLM fine-tuning."
+    if not (
+        FT_RD_SETTING.user_target_scenario or (FT_RD_SETTING.target_benchmark and FT_RD_SETTING.benchmark_description)
+    ):
+        raise ValueError("Either user_target_scenario or target_benchmark must be specified for LLM fine-tuning.")
 
     # Update configuration with provided parameters
     if dataset:
@@ -82,9 +82,8 @@ def main(
     model_target = FT_RD_SETTING.base_model if FT_RD_SETTING.base_model else "auto selected model"
 
     # Temporary assertion until auto-selection is implemented
-    assert (
-        FT_RD_SETTING.base_model is not None
-    ), "Base model auto selection not yet supported, please specify via --base-model"
+    if FT_RD_SETTING.base_model is None:
+        raise ValueError("Base model auto selection not yet supported, please specify via --base-model")
 
     logger.info(f"Starting LLM fine-tuning on dataset='{data_set_target}' with model='{model_target}'")
 

rdagent/app/qlib_rd_loop/factor.py

@@ -4,10 +4,9 @@ Factor workflow with session control
 
 import asyncio
 from pathlib import Path
-from typing import Any, Optional
+from typing import Any
 
 import fire
-
 from rdagent.app.qlib_rd_loop.conf import FACTOR_PROP_SETTING
 from rdagent.components.workflow.rd_loop import RDLoop
 from rdagent.core.exception import CoderError, FactorEmptyError
@@ -21,20 +20,20 @@ class FactorRDLoop(RDLoop):
     def running(self, prev_out: dict[str, Any]):
         exp = self.runner.develop(prev_out["coding"])
         if exp is None:
-            logger.error(f"Factor extraction failed.")
+            logger.error("Factor extraction failed.")
             raise FactorEmptyError("Factor extraction failed.")
         logger.log_object(exp, tag="runner result")
         return exp
 
 
 def main(
-    path: Optional[str] = None,
-    step_n: Optional[int] = None,
-    loop_n: Optional[int] = None,
+    path: str | None = None,
+    step_n: int | None = None,
+    loop_n: int | None = None,
     all_duration: str | None = None,
     checkout: bool = True,
-    checkout_path: Optional[str] = None,
-    base_features_path: Optional[str] = None,
+    checkout_path: str | None = None,
+    base_features_path: str | None = None,
     **kwargs,
 ):
     """
@@ -47,7 +46,7 @@ def main(
         dotenv run -- python rdagent/app/qlib_rd_loop/factor.py $LOG_PATH/__session__/1/0_propose  --step_n 1   # `step_n` is a optional paramter
 
     """
-    if not checkout_path is None:
+    if checkout_path is not None:
         checkout = Path(checkout_path)
 
     if path is None:

rdagent/app/qlib_rd_loop/factor_from_report.py

@@ -1,10 +1,9 @@
 import asyncio
 import json
 from pathlib import Path
-from typing import Any, Dict, Tuple
+from typing import Any
 
 import fire
-
 from rdagent.app.qlib_rd_loop.conf import FACTOR_FROM_REPORT_PROP_SETTING
 from rdagent.app.qlib_rd_loop.factor import FactorRDLoop
 from rdagent.components.document_reader.document_reader import (
@@ -12,7 +11,7 @@ from rdagent.components.document_reader.document_reader import (
     load_and_process_pdfs_by_langchain,
 )
 from rdagent.core.conf import RD_AGENT_SETTINGS
-from rdagent.core.proposal import Hypothesis, HypothesisFeedback
+from rdagent.core.proposal import Hypothesis
 from rdagent.log import rdagent_logger as logger
 from rdagent.oai.llm_utils import APIBackend
 from rdagent.scenarios.qlib.experiment.factor_experiment import QlibFactorExperiment
@@ -36,14 +35,14 @@ def generate_hypothesis(factor_result: dict, report_content: str) -> str:
     """
     system_prompt = T(".prompts:hypothesis_generation.system").r()
     user_prompt = T(".prompts:hypothesis_generation.user").r(
-        factor_descriptions=json.dumps(factor_result), report_content=report_content
+        factor_descriptions=json.dumps(factor_result), report_content=report_content,
     )
 
     response = APIBackend().build_messages_and_create_chat_completion(
         user_prompt=user_prompt,
         system_prompt=system_prompt,
         json_mode=True,
-        json_target_type=Dict[str, str],
+        json_target_type=dict[str, str],
     )
 
     response_json = json.loads(response)
@@ -99,7 +98,7 @@ class FactorReportLoop(FactorRDLoop, metaclass=LoopMeta):
         super().__init__(PROP_SETTING=FACTOR_FROM_REPORT_PROP_SETTING)
         if report_folder is None:
             self.judge_pdf_data_items = json.load(
-                open(FACTOR_FROM_REPORT_PROP_SETTING.report_result_json_file_path, "r")
+                open(FACTOR_FROM_REPORT_PROP_SETTING.report_result_json_file_path),
             )
         else:
             self.judge_pdf_data_items = [i for i in Path(report_folder).rglob("*.pdf")]
@@ -118,7 +117,7 @@ class FactorReportLoop(FactorRDLoop, metaclass=LoopMeta):
                 if exp is None:
                     self.shift_report += 1
                     self.loop_n -= 1
-                    if self.loop_n < 0:  # NOTE: on every step, we self.loop_n -= 1 at first.
+                    if self.loop_n < 0:  # loop_n is decremented above when reports are empty; prevents infinite skipping
                         raise self.LoopTerminationError("Reach stop criterion and stop loop")
                     continue
                 exp.based_experiments = [QlibFactorExperiment(sub_tasks=[], hypothesis=exp.hypothesis)] + [

rdagent/app/qlib_rd_loop/quant.py

@@ -8,7 +8,6 @@ from pathlib import Path
 from typing import Any
 
 import fire
-
 from rdagent.app.qlib_rd_loop.conf import QUANT_PROP_SETTING
 from rdagent.components.workflow.conf import BasePropSetting
 from rdagent.components.workflow.rd_loop import RDLoop
@@ -44,11 +43,11 @@ class QuantRDLoop(RDLoop):
         logger.log_object(self.hypothesis_gen, tag="quant hypothesis generator")
 
         self.factor_hypothesis2experiment: Hypothesis2Experiment = import_class(
-            PROP_SETTING.factor_hypothesis2experiment
+            PROP_SETTING.factor_hypothesis2experiment,
         )()
         logger.log_object(self.factor_hypothesis2experiment, tag="factor hypothesis2experiment")
         self.model_hypothesis2experiment: Hypothesis2Experiment = import_class(
-            PROP_SETTING.model_hypothesis2experiment
+            PROP_SETTING.model_hypothesis2experiment,
         )()
         logger.log_object(self.model_hypothesis2experiment, tag="model hypothesis2experiment")
 
@@ -78,7 +77,8 @@ class QuantRDLoop(RDLoop):
         while True:
             if self.get_unfinished_loop_cnt(self.loop_idx) < RD_AGENT_SETTINGS.get_max_parallel():
                 hypo = self._propose()
-                assert hypo.action in ["factor", "model"]
+                if hypo.action not in ["factor", "model"]:
+                    raise ValueError(f"hypo.action must be 'factor' or 'model', got {hypo.action!r}")
                 if hypo.action == "factor":
                     exp = self.factor_hypothesis2experiment.convert(hypo, self.trace)
                 else:
@@ -132,7 +132,6 @@ class QuantRDLoop(RDLoop):
         """
         import json
         from datetime import datetime
-        from pathlib import Path
 
         try:
             project_root = Path(__file__).parent.parent.parent.parent
@@ -195,11 +194,11 @@ class QuantRDLoop(RDLoop):
         if prev_out["direct_exp_gen"]["propose"].action == "factor":
             exp = self.factor_runner.develop(prev_out["coding"])
             if exp is None:
-                logger.error(f"Factor extraction failed.")
+                logger.error("Factor extraction failed.")
                 raise FactorEmptyError("Factor extraction failed.")
 
             # Increment factor count for tracking
-            if hasattr(self, 'trace') and hasattr(self.trace, 'increment_factor_count'):
+            if hasattr(self, "trace") and hasattr(self.trace, "increment_factor_count"):
                 self.trace.increment_factor_count()
 
             # Handle failed experiments gracefully (don't break the loop)
@@ -210,7 +209,7 @@ class QuantRDLoop(RDLoop):
                     factor_name = getattr(exp.hypothesis, "hypothesis", "unknown")
                 logger.warning(
                     f"Factor '{factor_name}' failed evaluation: {reason}. "
-                    f"Continuing with next factor."
+                    f"Continuing with next factor.",
                 )
                 # Return exp anyway - loop will continue
         elif prev_out["direct_exp_gen"]["propose"].action == "model":
@@ -219,7 +218,7 @@ class QuantRDLoop(RDLoop):
         return exp
 
     def feedback(self, prev_out: dict[str, Any]):
-        e = prev_out.get(self.EXCEPTION_KEY, None)
+        e = prev_out.get(self.EXCEPTION_KEY)
         if e is not None:
             feedback = HypothesisFeedback(
                 observations=str(e),
@@ -245,11 +244,10 @@ class QuantRDLoop(RDLoop):
                     reason=reason,
                     decision=False,
                 )
-            else:
-                if prev_out["direct_exp_gen"]["propose"].action == "factor":
-                    feedback = self.factor_summarizer.generate_feedback(prev_out["running"], self.trace)
-                elif prev_out["direct_exp_gen"]["propose"].action == "model":
-                    feedback = self.model_summarizer.generate_feedback(prev_out["running"], self.trace)
+            elif prev_out["direct_exp_gen"]["propose"].action == "factor":
+                feedback = self.factor_summarizer.generate_feedback(prev_out["running"], self.trace)
+            elif prev_out["direct_exp_gen"]["propose"].action == "model":
+                feedback = self.model_summarizer.generate_feedback(prev_out["running"], self.trace)
 
             # NOTE: DB save is handled by factor_runner.py _save_result_to_database()
             # which runs immediately after Docker execution. No duplicate save needed here.
@@ -258,20 +256,20 @@ class QuantRDLoop(RDLoop):
         factor_count = self.trace.get_factor_count()
 
         # Check for auto-strategies trigger
-        auto_strategies = getattr(self, '_auto_strategies', False)
-        auto_threshold = getattr(self, '_auto_strategies_threshold', 500)
+        auto_strategies = getattr(self, "_auto_strategies", False)
+        auto_threshold = getattr(self, "_auto_strategies_threshold", 500)
 
         if auto_strategies and factor_count > 0 and factor_count % auto_threshold == 0:
             logger.info(
                 f"Auto-strategy trigger: {factor_count} factors evaluated. "
-                f"Suggesting strategy generation now..."
+                f"Suggesting strategy generation now...",
             )
             self._build_strategies_with_ai()
         elif factor_count > 0 and factor_count % 50 == 0 and not auto_strategies:
             # Standard periodic suggestion (every 50 factors)
             logger.info(
                 f"Periodic check: {factor_count} factors evaluated. "
-                f"Consider running 'rdagent generate_strategies' for AI strategy generation."
+                f"Consider running 'rdagent generate_strategies' for AI strategy generation.",
             )
 
         feedback = self._interact_feedback(feedback)
@@ -292,10 +290,11 @@ class QuantRDLoop(RDLoop):
         - Optuna hyperparameter optimization
         """
         try:
-            from rdagent.components.coder.strategy_orchestrator import StrategyOrchestrator
             from pathlib import Path
+
             import yaml
-            
+            from rdagent.components.coder.strategy_orchestrator import StrategyOrchestrator
+
             # Load improved prompt
             project_root = Path(__file__).parent.parent.parent.parent
             prompt_path = project_root / "prompts" / "strategy_generation_v2.yaml"
@@ -335,44 +334,47 @@ class QuantRDLoop(RDLoop):
 
             logger.info(f"StrategyOrchestrator: Building strategies from {len(top_factors)} top factors...")
             logger.info(f"  - Using improved prompt: {improved_prompt is not None}")
-            logger.info(f"  - Optuna optimization: enabled (20 trials)")
-            logger.info(f"  - Real OHLCV backtest: enabled")
+            logger.info("  - Optuna optimization: enabled (20 trials)")
+            logger.info("  - Real OHLCV backtest: enabled")
 
             # Initialize orchestrator with Optuna
             orchestrator = StrategyOrchestrator(
                 top_factors=20,
-                trading_style='swing',
+                trading_style="swing",
                 min_sharpe=0.5,
                 max_drawdown=-0.20,
                 min_win_rate=0.40,
                 use_optuna=True,
                 optuna_trials=20,
             )
-            
+
             # Override with improved prompt if available
             if improved_prompt:
-                orchestrator.strategy_prompt = improved_prompt.get('strategy_generation', {})
+                orchestrator.strategy_prompt = improved_prompt.get("strategy_generation", {})
 
             # Generate 3 strategies per cycle
             n_strategies = 3
             logger.info(f"Generating {n_strategies} strategies...")
-            
+
             # Load top factors for generation
             orch_factors = orchestrator.load_top_factors()
-            
+            if len(orch_factors) < 2:
+                logger.warning(f"Not enough factors for strategy generation (need >= 2, got {len(orch_factors)}). Skipping.")
+                return
+
             for i in range(n_strategies):
+                strategy_name = f"auto_gen_v{i+1}"
                 try:
                     # Select random factor combination
                     import random
                     n_factors = random.randint(2, min(5, len(orch_factors)))
                     factor_subset = random.sample(orch_factors, n_factors)
-                    
-                    strategy_name = f"auto_gen_v{i+1}"
+
                     code = orchestrator.generate_strategy_code(factor_subset, strategy_name)
-                    
+
                     if code:
                         result = orchestrator.evaluate_strategy(code, strategy_name, factor_subset)
-                        
+
                         if result.get("status") == "accepted":
                             logger.info(f"✅ Strategy {strategy_name} accepted!")
                             logger.info(f"   Sharpe: {result.get('sharpe_ratio', 0):.2f}")
@@ -430,7 +432,7 @@ def main(
         quant_loop._auto_strategies = True
         quant_loop._auto_strategies_threshold = auto_strategies_threshold
         logger.info(
-            f"Auto-strategies enabled. Will trigger after {auto_strategies_threshold} factors."
+            f"Auto-strategies enabled. Will trigger after {auto_strategies_threshold} factors.",
         )
     else:
         quant_loop._auto_strategies = False

rdagent/components/backtesting/backtest_engine.py

@@ -72,7 +72,7 @@ class BacktestMetrics:
 class FactorBacktester:
     def __init__(self):
         self.metrics = BacktestMetrics()
-        self.results_path = Path(__file__).parent.parent.parent / "results" / "backtests"
+        self.results_path = Path(__file__).parent.parent.parent.parent / "results" / "backtests"
         self.results_path.mkdir(parents=True, exist_ok=True)
 
     def run_backtest(
@@ -222,7 +222,7 @@ class FactorBacktester:
 
             # Calculate return for this step
             if step > 0:
-                prev_price = float(price_values[step - 1]) if step > 0 else current_price
+                prev_price = float(price_values[step - 1])
                 if prev_price > 0:
                     step_return = (current_price - prev_price) / prev_price * position
                     returns_history.append(step_return)

rdagent/components/backtesting/results_db.py

@@ -166,7 +166,7 @@ class ResultsDatabase:
         self.conn.commit()
         return c.lastrowid
     
-    def add_loop(self, loop_idx: int, success: int, fail: int, best_ic: float = None, status: str = "completed") -> int:
+    def add_loop(self, loop_idx: int, success: int, fail: int, best_ic: float | None = None, status: str = "completed") -> int:
         c = self.conn.cursor()
         rate = success / (success + fail) if (success + fail) > 0 else 0
         c.execute("""INSERT INTO loop_results (loop_index, factors_success, factors_fail, success_rate, best_ic, status)
@@ -330,13 +330,13 @@ class ResultsDatabase:
         worst_drawdown = all_results['max_drawdown'].min() if total_runs > 0 and all_results['max_drawdown'].notna().any() else None
 
         # Scan factors directory for JSON files
-        factors_dir = Path(__file__).parent.parent.parent / "results" / "factors"
+        factors_dir = Path(__file__).parent.parent.parent.parent / "results" / "factors"
         json_factor_files = 0
         if factors_dir.exists():
             json_factor_files = len(list(factors_dir.glob("*.json")))
 
         # Scan failed runs
-        failed_dir = Path(__file__).parent.parent.parent / "results" / "failed_runs"
+        failed_dir = Path(__file__).parent.parent.parent.parent / "results" / "failed_runs"
         failed_runs_file = failed_dir / "failed_runs.json"
         failed_runs_count = 0
         failed_runs_data = []

rdagent/components/backtesting/risk_management.py

@@ -1,21 +1,19 @@
 """
 Predix Risk Management - Korrelation, Portfolio-Optimierung
 """
+
 import numpy as np
 import pandas as pd
-from pathlib import Path
-from typing import Dict, List, Optional
-from datetime import datetime
-import json
+
 
 class CorrelationAnalyzer:
     def __init__(self, lookback: int = 60):
         self.lookback = lookback
-    
+
     def calculate_matrix(self, returns: pd.DataFrame) -> pd.DataFrame:
         return returns.dropna().corr()
-    
-    def find_uncorrelated(self, corr: pd.DataFrame, threshold: float = 0.3) -> List[str]:
+
+    def find_uncorrelated(self, corr: pd.DataFrame, threshold: float = 0.3) -> list[str]:
         result = []
         for f in corr.columns:
             others = [x for x in corr.columns if x != f]
@@ -28,9 +26,9 @@ class PortfolioOptimizer:
         try:
             w = np.linalg.inv(cov.values) @ exp_ret.values
             return w / np.sum(w)
-        except:
+        except (np.linalg.LinAlgError, ValueError):
             return np.ones(len(exp_ret)) / len(exp_ret)
-    
+
     def risk_parity(self, cov: pd.DataFrame, max_iter: int = 100) -> np.ndarray:
         n = cov.shape[0]
         w = np.ones(n) / n
@@ -53,36 +51,36 @@ class AdvancedRiskManager:
         self.max_dd = max_dd
         self.corr_analyzer = CorrelationAnalyzer()
         self.optimizer = PortfolioOptimizer()
-    
-    def check_limits(self, weights: np.ndarray, vol: float, dd: float) -> Dict[str, bool]:
+
+    def check_limits(self, weights: np.ndarray, vol: float, dd: float) -> dict[str, bool]:
         return {
-            'position_limit': np.max(np.abs(weights)) <= self.max_pos,
-            'leverage_limit': np.sum(np.abs(weights)) <= self.max_lev,
-            'drawdown_limit': abs(dd) <= self.max_dd,
+            "position_limit": np.max(np.abs(weights)) <= self.max_pos,
+            "leverage_limit": np.sum(np.abs(weights)) <= self.max_lev,
+            "drawdown_limit": abs(dd) <= self.max_dd,
         }
 
 if __name__ == "__main__":
     print("=== Risk Test ===")
     np.random.seed(42)
-    n, names = 252, ['Mom', 'MeanRev', 'Vol', 'Volu', 'ML']
+    n, names = 252, ["Mom", "MeanRev", "Vol", "Volu", "ML"]
     ret = pd.DataFrame(np.random.randn(n, 5), columns=names)
-    
+
     corr = CorrelationAnalyzer().calculate_matrix(ret)
     print("Korrelationsmatrix:")
     print(corr.round(2))
-    
+
     opt = PortfolioOptimizer()
     exp_ret = pd.Series([0.1, 0.08, 0.06, 0.07, 0.12], index=names)
     cov = ret.cov() * 252
-    
+
     mv = opt.mean_variance(exp_ret, cov)
     print("\nMean-Variance:")
     for n, w in zip(names, mv): print(f"  {n}: {w:.2%}")
-    
+
     rp = opt.risk_parity(cov)
     print("\nRisk Parity:")
     for n, w in zip(names, rp): print(f"  {n}: {w:.2%}")
-    
+
     rm = AdvancedRiskManager()
     checks = rm.check_limits(mv, 0.15, -0.08)
     print(f"\nLimits OK: {all(checks.values())}")

rdagent/components/backtesting/vbt_backtest.py

@@ -19,7 +19,7 @@ Design goals
 """
 from __future__ import annotations
 
-from typing import Any, Dict, Optional
+from typing import Any
 
 import numpy as np
 import pandas as pd
@@ -67,9 +67,8 @@ def _cross_check_with_vbt(
     close: pd.Series,
     position: pd.Series,
     txn_cost: float,
-    manual_total_return: float,
     freq: str,
-) -> Optional[float]:
+) -> float | None:
     """Run a vectorbt simulation and return its total_return for comparison."""
     if not VBT_AVAILABLE:
         return None
@@ -84,7 +83,8 @@ def _cross_check_with_vbt(
             init_cash=10_000.0,
             freq=freq,
         )
-        return float(pf.total_return())
+        tr = float(pf.total_return())
+        return tr if np.isfinite(tr) else None
     except Exception:
         return None
 
@@ -95,9 +95,9 @@ def backtest_signal(
     txn_cost_bps: float = DEFAULT_TXN_COST_BPS,
     freq: str = "1min",
     bars_per_year: int = DEFAULT_BARS_PER_YEAR,
-    forward_returns: Optional[pd.Series] = None,
+    forward_returns: pd.Series | None = None,
     cross_check: bool = False,
-) -> Dict[str, Any]:
+) -> dict[str, Any]:
     """
     Run a single-asset backtest from a position signal.
 
@@ -204,7 +204,7 @@ def backtest_signal(
     calmar = ann_return_arith / abs(max_dd) if max_dd < 0 else 0.0
 
     trade_pnl = _compute_trade_pnl(position, strategy_returns)
-    n_trades = int(len(trade_pnl))
+    n_trades = len(trade_pnl)
     n_position_changes = int((position.diff().fillna(0) != 0).sum())
 
     if n_trades > 0:
@@ -216,7 +216,7 @@ def backtest_signal(
         win_rate = 0.0
         profit_factor = 0.0
 
-    ic: Optional[float] = None
+    ic: float | None = None
     if forward_returns is not None:
         fwd = pd.to_numeric(forward_returns, errors="coerce")
         common = signal.index.intersection(fwd.dropna().index)
@@ -227,7 +227,7 @@ def backtest_signal(
                 ic_val = float(s.corr(f))
                 ic = ic_val if np.isfinite(ic_val) else None
 
-    result: Dict[str, Any] = {
+    result: dict[str, Any] = {
         "status": "success",
         "sharpe": sharpe,
         "sortino": sortino,
@@ -244,7 +244,7 @@ def backtest_signal(
         "volatility": volatility,
         "n_trades": n_trades,
         "n_position_changes": n_position_changes,
-        "n_bars": int(len(strategy_returns)),
+        "n_bars": len(strategy_returns),
         "n_months": float(n_months),
         "signal_long": int((signal > 0).sum()),
         "signal_short": int((signal < 0).sum()),
@@ -264,7 +264,6 @@ def backtest_signal(
             close=close,
             position=position,
             txn_cost=txn_cost,
-            manual_total_return=total_return,
             freq=freq,
         )
 
@@ -293,7 +292,7 @@ def _apply_ftmo_mask(
 
     daily_breaches = 0
     total_breached = False
-    total_breach_ts: Optional[pd.Timestamp] = None
+    total_breach_ts: pd.Timestamp | None = None
     current_day    = None
     day_start_eq   = FTMO_INITIAL_CAPITAL
 
@@ -308,11 +307,8 @@ def _apply_ftmo_mask(
         pos_i  = float(signal.at[ts]) * leverage
         ret_i  = float(bar_ret.get(ts, 0.0))
         cost_i = abs(pos_i - pos_prev) * txn_cost
-        ret_net = pos_prev * ret_i - cost_i
-        equity  = equity * (1.0 + ret_net / FTMO_INITIAL_CAPITAL * FTMO_INITIAL_CAPITAL / equity
-                            if equity > 0 else 1.0)
-        # Simpler: track as fraction
-        equity += FTMO_INITIAL_CAPITAL * ret_net
+        ret_frac = pos_prev * ret_i - cost_i
+        equity *= 1.0 + ret_frac if equity > 0 else 1.0
         pos_prev = pos_i
 
         if total_breached:
@@ -399,7 +395,7 @@ def walk_forward_rolling(
     is_years: int = WF_IS_YEARS,
     oos_years: int = WF_OOS_YEARS,
     step_years: int = WF_STEP_YEARS,
-) -> Dict[str, Any]:
+) -> dict[str, Any]:
     """
     Rolling walk-forward validation: multiple IS/OOS windows shifted by ``step_years``.
 
@@ -433,7 +429,7 @@ def walk_forward_rolling(
             yr += step_years
             continue
 
-        window: Dict[str, Any] = {
+        window: dict[str, Any] = {
             "is_start":  str(is_start.date()),
             "is_end":    str(is_end.date()),
             "oos_start": str(is_end.date()),
@@ -475,11 +471,11 @@ def backtest_signal_ftmo(
     stop_pips: float = FTMO_STOP_PIPS,
     max_leverage: float = FTMO_MAX_LEVERAGE,
     bars_per_year: int = DEFAULT_BARS_PER_YEAR,
-    forward_returns: Optional[pd.Series] = None,
-    oos_start: Optional[str] = OOS_START_DEFAULT,
+    forward_returns: pd.Series | None = None,
+    oos_start: str | None = OOS_START_DEFAULT,
     wf_rolling: bool = False,
     mc_n_permutations: int = 0,
-) -> Dict[str, Any]:
+) -> dict[str, Any]:
     """
     FTMO-compliant backtest of a strategy signal on EUR/USD.
 
@@ -547,7 +543,7 @@ def backtest_signal_ftmo(
         is_mask  = close.index < oos_ts
         oos_mask = close.index >= oos_ts
 
-        def _split_bt(mask: "pd.Series[bool]", prefix: str) -> None:
+        def _split_bt(mask: pd.Series[bool], prefix: str) -> None:
             if mask.sum() < 100:
                 return
             close_s  = close.loc[mask]
@@ -602,7 +598,7 @@ def backtest_from_forward_returns(
     forward_returns: pd.Series,
     txn_cost_bps: float = DEFAULT_TXN_COST_BPS,
     bars_per_year: int = DEFAULT_BARS_PER_YEAR,
-) -> Dict[str, Any]:
+) -> dict[str, Any]:
     """
     Backtest a factor using sign(factor) as signal against forward returns.
 
@@ -640,7 +636,7 @@ def backtest_from_forward_returns(
     ic = ic_val if np.isfinite(ic_val) else 0.0
 
     trade_pnl = _compute_trade_pnl(position, strategy_returns)
-    n_trades = int(len(trade_pnl))
+    n_trades = len(trade_pnl)
     win_rate = float((trade_pnl > 0).mean()) if n_trades > 0 else 0.0
 
     ann_return = float(strategy_returns.mean() * bars_per_year)
@@ -656,7 +652,7 @@ def backtest_from_forward_returns(
         "win_rate": win_rate,
         "n_trades": n_trades,
         "ic": ic,
-        "n_bars": int(len(strategy_returns)),
+        "n_bars": len(strategy_returns),
         "txn_cost_bps": txn_cost_bps,
         "bars_per_year": bars_per_year,
     }

rdagent/components/coder/CoSTEER/__init__.py

@@ -75,8 +75,10 @@ class CoSTEER(Developer[Experiment]):
 
     def _get_last_fb(self) -> CoSTEERMultiFeedback:
         fb = self.evolve_agent.evolving_trace[-1].feedback
-        assert fb is not None, "feedback is None"
-        assert isinstance(fb, CoSTEERMultiFeedback), "feedback must be of type CoSTEERMultiFeedback"
+        if fb is None:
+            raise AssertionError("feedback is None")
+        if not isinstance(fb, CoSTEERMultiFeedback):
+            raise TypeError("feedback must be of type CoSTEERMultiFeedback")
         return fb
 
     def should_use_new_evo(self, base_fb: CoSTEERMultiFeedback | None, new_fb: CoSTEERMultiFeedback) -> bool:
@@ -121,7 +123,8 @@ class CoSTEER(Developer[Experiment]):
 
         for evo_exp in self.evolve_agent.multistep_evolve(evo_exp, self.evaluator):
             iteration_count += 1
-            assert isinstance(evo_exp, Experiment)  # multiple inheritance
+            if not isinstance(evo_exp, Experiment):
+                raise TypeError("evo_exp must be an instance of Experiment")
             evo_fb = self._get_last_fb()
             update_fallback = self.should_use_new_evo(
                 base_fb=fallback_evo_fb,
@@ -154,7 +157,8 @@ class CoSTEER(Developer[Experiment]):
                 evo_exp = fallback_evo_exp
                 evo_exp.recover_ws_ckp()
                 evo_fb = fallback_evo_fb
-            assert evo_fb is not None  # multistep_evolve should run at least once
+            if evo_fb is None:
+                raise AssertionError("multistep_evolve should run at least once")
             evo_exp = self._exp_postprocess_by_feedback(evo_exp, evo_fb)
         except CoderError as e:
             e.caused_by_timeout = reached_max_seconds
@@ -264,9 +268,12 @@ class CoSTEER(Developer[Experiment]):
         - Raise Error if it failed to handle the develop task
         -
         """
-        assert isinstance(evo, Experiment)
-        assert isinstance(feedback, CoSTEERMultiFeedback)
-        assert len(evo.sub_workspace_list) == len(feedback)
+        if not isinstance(evo, Experiment):
+            raise TypeError("evo must be an instance of Experiment")
+        if not isinstance(feedback, CoSTEERMultiFeedback):
+            raise TypeError("feedback must be an instance of CoSTEERMultiFeedback")
+        if len(evo.sub_workspace_list) != len(feedback):
+            raise ValueError("Length of sub_workspace_list must match length of feedback")
 
         # FIXME: when whould the feedback be None?
         failed_feedbacks = [

rdagent/components/coder/CoSTEER/evolving_strategy.py

@@ -122,7 +122,8 @@ class MultiProcessEvolvingStrategy(EvolvingStrategy):
         last_feedback = None
         if len(evolving_trace) > 0:
             last_feedback = evolving_trace[-1].feedback
-            assert isinstance(last_feedback, CoSTEERMultiFeedback)
+            if not isinstance(last_feedback, CoSTEERMultiFeedback):
+                raise TypeError("last_feedback must be of type CoSTEERMultiFeedback")
 
         # 1.找出需要evolve的task
         to_be_finished_task_index: list[int] = []

rdagent/components/coder/CoSTEER/knowledge_management.py

@@ -1028,7 +1028,8 @@ class CoSTEERKnowledgeBaseV2(EvolvingKnowledgeBase):
 
         """
         node_count = len(nodes)
-        assert node_count >= 2, "nodes length must >=2"
+        if node_count < 2:
+            raise ValueError("nodes length must >=2")
         intersection_node_list = []
         if output_intersection_origin:
             origin_list = []

rdagent/components/coder/factor_coder/eva_utils.py

@@ -328,12 +328,13 @@ class FactorEqualValueRatioEvaluator(FactorEvaluator):
                 "The source dataframe is None. Please check the implementation.",
                 -1,
             )
+        acc_rate = -1
         try:
             close_values = gen_df.sub(gt_df).abs().lt(1e-6)
             result_int = close_values.astype(int)
             pos_num = result_int.sum().sum()
             acc_rate = pos_num / close_values.size
-        except:
+        except Exception:
             close_values = gen_df
         if close_values.all().iloc[0]:
             return (

rdagent/components/coder/model_coder/eva_utils.py

@@ -58,10 +58,12 @@ class ModelCodeEvaluator(CoSTEEREvaluator):
         model_execution_feedback: str = "",
         model_value_feedback: str = "",
     ):
-        assert isinstance(target_task, ModelTask)
-        assert isinstance(implementation, ModelFBWorkspace)
-        if gt_implementation is not None:
-            assert isinstance(gt_implementation, ModelFBWorkspace)
+        if not isinstance(target_task, ModelTask):
+            raise TypeError("target_task must be of type ModelTask")
+        if not isinstance(implementation, ModelFBWorkspace):
+            raise TypeError("implementation must be of type ModelFBWorkspace")
+        if gt_implementation is not None and not isinstance(gt_implementation, ModelFBWorkspace):
+            raise TypeError("gt_implementation must be of type ModelFBWorkspace")
 
         model_task_information = target_task.get_task_information()
         code = implementation.all_codes
@@ -113,10 +115,12 @@ class ModelFinalEvaluator(CoSTEEREvaluator):
         model_value_feedback: str,
         model_code_feedback: str,
     ):
-        assert isinstance(target_task, ModelTask)
-        assert isinstance(implementation, ModelFBWorkspace)
-        if gt_implementation is not None:
-            assert isinstance(gt_implementation, ModelFBWorkspace)
+        if not isinstance(target_task, ModelTask):
+            raise TypeError("target_task must be of type ModelTask")
+        if not isinstance(implementation, ModelFBWorkspace):
+            raise TypeError("implementation must be of type ModelFBWorkspace")
+        if gt_implementation is not None and not isinstance(gt_implementation, ModelFBWorkspace):
+            raise TypeError("gt_implementation must be of type ModelFBWorkspace")
 
         system_prompt = T(".prompts:evaluator_final_feedback.system").r(
             scenario=(

rdagent/components/coder/optuna_optimizer.py

@@ -23,7 +23,7 @@ import pandas as pd
 
 from rdagent.log import rdagent_logger as logger
 
-logger = logging.getLogger(__name__)
+_optuna_logger = logging.getLogger(__name__)
 
 try:
     import optuna
@@ -292,6 +292,7 @@ class OptunaOptimizer:
             "volatility_lookback": trial.suggest_int("volatility_lookback", 5, 500, step=5),
             "signal_bias": trial.suggest_float("signal_bias", -1.0, 1.0, step=0.05),
             "max_hold_bars": trial.suggest_int("max_hold_bars", 5, 1000, step=5),
+            "max_positions": trial.suggest_int("max_positions", 1, 5, step=1),
         }
 
     # Parameters that are allowed to be negative (not clamped to 0).
@@ -308,6 +309,7 @@ class OptunaOptimizer:
         "volatility_lookback": 1.0,
         "signal_bias": -1.0,
         "max_hold_bars": 1.0,
+        "max_positions": 1.0,
     }
 
     def _suggest_bounded(
@@ -357,6 +359,7 @@ class OptunaOptimizer:
             "volatility_lookback": (center.get("volatility_lookback", 100), 30),
             "signal_bias": (center.get("signal_bias", 0.0), 0.2),
             "max_hold_bars": (center.get("max_hold_bars", 100), 50),
+            "max_positions": (center.get("max_positions", 1), 2),
         }
         return {key: self._suggest_bounded(trial, key, c, hw) for key, (c, hw) in ranges.items()}
 
@@ -388,6 +391,7 @@ class OptunaOptimizer:
             "volatility_lookback": (center.get("volatility_lookback", 100), 10),
             "signal_bias": (center.get("signal_bias", 0.0), 0.07),
             "max_hold_bars": (center.get("max_hold_bars", 100), 17),
+            "max_positions": (center.get("max_positions", 1), 1),
         }
         return {key: self._suggest_bounded(trial, key, c, hw) for key, (c, hw) in ranges.items()}
 
@@ -467,6 +471,9 @@ class OptunaOptimizer:
 
             # Max holding periods (in bars)
             "max_hold_bars": trial.suggest_int("max_hold_bars", 10, 500, step=10),
+
+            # Max concurrent positions (1 = no pyramiding, 2-5 = scale-in)
+            "max_positions": trial.suggest_int("max_positions", 1, 5, step=1),
         }
 
         return params
@@ -597,6 +604,13 @@ class OptunaOptimizer:
             if signal_bias != 0.0:
                 signal = (signal.astype(float) + signal_bias).round().astype(int).clip(-1, 1)
 
+            # Apply max_positions: scale signal by position_size_pct and cap exposure
+            max_positions   = int(params.get("max_positions", 1))
+            position_size_pct = float(params.get("position_size_pct", 1.0))
+            # Each "position" is position_size_pct of equity; total exposure capped at max_positions × size
+            effective_size  = min(position_size_pct * max_positions, 1.0)
+            signal = (signal.astype(float) * effective_size).clip(-1.0, 1.0)
+
             # Build a synthetic close from the factor-mean so we can route
             # through the same unified engine as every other backtest path.
             # Backtest formulas must match the orchestrator's real-OHLCV path.

rdagent/components/coder/strategy_orchestrator.py

@@ -26,22 +26,18 @@ import traceback
 from concurrent.futures import ThreadPoolExecutor, as_completed
 from datetime import datetime
 from pathlib import Path
-from typing import Any, Dict, List, Optional
+from typing import Any
 
 import numpy as np
 import pandas as pd
-import requests
-
-from rdagent.components.prompt_loader import load_prompt
 from rdagent.components.coder.optuna_optimizer import OptunaOptimizer
+from rdagent.components.prompt_loader import load_prompt
 
 # OHLCV data path
 OHLCV_PATH = Path(os.getenv(
-    'PREDIX_OHLCV_PATH',
-    '/home/nico/Predix/git_ignore_folder/factor_implementation_source_data/intraday_pv.h5'
+    "PREDIX_OHLCV_PATH",
+    "/home/nico/Predix/git_ignore_folder/factor_implementation_source_data/intraday_pv.h5",
 ))
-from rdagent.log import rdagent_logger as logger
-
 logger = logging.getLogger(__name__)
 
 
@@ -60,7 +56,7 @@ class StrategyOrchestrator:
         min_sharpe: float = 0.3,
         max_drawdown: float = -0.30,
         min_win_rate: float = 0.40,
-        results_dir: Optional[str] = None,
+        results_dir: str | None = None,
         use_optuna: bool = True,
         optuna_trials: int = 20,
         continuous_optimization: bool = True,
@@ -118,7 +114,7 @@ class StrategyOrchestrator:
 
         logger.info(
             f"StrategyOrchestrator initialized: style={self.trading_style}, "
-            f"top_factors={self.top_factors}, min_sharpe={self.min_sharpe}"
+            f"top_factors={self.top_factors}, min_sharpe={self.min_sharpe}",
         )
 
     def load_ohlcv_close(self) -> pd.Series:
@@ -126,31 +122,31 @@ class StrategyOrchestrator:
         if not OHLCV_PATH.exists():
             logger.warning(f"OHLCV data not found: {OHLCV_PATH}")
             return None
-        
+
         try:
-            ohlcv = pd.read_hdf(str(OHLCV_PATH), key='data')
-            if '$close' in ohlcv.columns:
-                close = ohlcv['$close'].dropna()
-            elif 'close' in ohlcv.columns:
-                close = ohlcv['close'].dropna()
+            ohlcv = pd.read_hdf(str(OHLCV_PATH), key="data")
+            if "$close" in ohlcv.columns:
+                close = ohlcv["$close"].dropna()
+            elif "close" in ohlcv.columns:
+                close = ohlcv["close"].dropna()
             else:
                 close = ohlcv.select_dtypes(include=[np.number]).iloc[:, 0].dropna()
-            
+
             # Handle MultiIndex
             if isinstance(close.index, pd.MultiIndex):
                 try:
-                    close = close.xs('EURUSD', level='instrument')
+                    close = close.xs("EURUSD", level="instrument")
                 except KeyError:
-                    idx = close.index.get_level_values('instrument') == 'EURUSD'
+                    idx = close.index.get_level_values("instrument") == "EURUSD"
                     close = close[idx]
-                    close.index = close.index.droplevel('instrument')
-            
+                    close.index = close.index.droplevel("instrument")
+
             return close
         except Exception as e:
             logger.warning(f"Failed to load OHLCV data: {e}")
             return None
 
-    def load_top_factors(self) -> List[Dict[str, Any]]:
+    def load_top_factors(self) -> list[dict[str, Any]]:
         """
         Load top evaluated factors from JSON files.
 
@@ -177,7 +173,7 @@ class StrategyOrchestrator:
 
         # Sort by absolute IC and take top N
         factors.sort(key=lambda x: abs(x.get("ic", 0) or 0), reverse=True)
-        
+
         # Filter to only include factors that have parquet files
         factors_with_files = []
         for f in factors:
@@ -188,16 +184,16 @@ class StrategyOrchestrator:
                 factors_with_files.append(f)
             else:
                 logger.debug(f"Skipping {fname} - no parquet file")
-        
+
         # Select diverse factor TYPES, not just top IC
         # This ensures we get momentum, volatility, session, volume, etc.
         type_keywords = {
             "momentum": [], "trend": [], "volatility": [], "volume": [],
             "session": [], "london": [], "range": [], "vwap": [],
             "return": [], "ofi": [], "spread": [], "close": [],
-            "divergence": [], "other": []
+            "divergence": [], "other": [],
         }
-        
+
         for f in factors_with_files:
             name = f.get("factor_name", "").lower()
             matched = False
@@ -208,18 +204,18 @@ class StrategyOrchestrator:
                     break
             if not matched:
                 type_keywords["other"].append(f)
-        
+
         # Select best from each type (ensures diversity)
         selected = []
         already_names = set()
-        
+
         # Priority order: momentum, divergence, volatility, session, volume, etc.
-        priority_types = ["momentum", "divergence", "volatility", "session", 
-                         "london", "range", "vwap", "volume", "ofi", "spread", 
+        priority_types = ["momentum", "divergence", "volatility", "session",
+                         "london", "range", "vwap", "volume", "ofi", "spread",
                          "return", "trend", "close", "other"]
-        
+
         per_type = max(2, self.top_factors // len(priority_types))
-        
+
         for kw in priority_types:
             for f in sorted(type_keywords[kw], key=lambda x: abs(x.get("ic", 0)), reverse=True):
                 if f["factor_name"] not in already_names:
@@ -227,13 +223,13 @@ class StrategyOrchestrator:
                     already_names.add(f["factor_name"])
                     if len([s for s in selected if s["factor_name"] in [x["factor_name"] for x in type_keywords[kw]]]) >= per_type:
                         break
-        
+
         # Fill remaining with highest IC not yet selected
         if len(selected) < self.top_factors:
             remaining = [f for f in factors_with_files if f["factor_name"] not in already_names]
             remaining.sort(key=lambda x: abs(x.get("ic", 0)), reverse=True)
             selected.extend(remaining[:self.top_factors - len(selected)])
-        
+
         # Log diversity
         type_counts = {}
         for f in selected:
@@ -246,12 +242,12 @@ class StrategyOrchestrator:
                     break
             if not matched:
                 type_counts["other"] = type_counts.get("other", 0) + 1
-        
+
         logger.info(f"Selected {len(selected)} diverse factors: {type_counts}")
-        
+
         return selected[:self.top_factors]
 
-    def load_factor_values(self, factor_name: str) -> Optional[pd.Series]:
+    def load_factor_values(self, factor_name: str) -> pd.Series | None:
         """
         Load factor time-series values from parquet file.
 
@@ -273,32 +269,32 @@ class StrategyOrchestrator:
 
         try:
             df = pd.read_parquet(str(parquet_path))
-            
+
             # Handle empty DataFrame
             if df.empty or len(df.columns) == 0:
                 logger.warning(f"Empty parquet file: {parquet_path}")
                 return None
-            
+
             # Handle MultiIndex (datetime, instrument)
             if isinstance(df.index, pd.MultiIndex):
                 # Get the factor column name (should be the only column)
                 factor_col = df.columns[0]
                 # Extract EURUSD series
                 try:
-                    series = df.xs('EURUSD', level='instrument')[factor_col]
+                    series = df.xs("EURUSD", level="instrument")[factor_col]
                 except KeyError:
                     # Try alternative extraction
                     df_reset = df.reset_index()
-                    if 'instrument' in df_reset.columns:
-                        df_eur = df_reset[df_reset['instrument'] == 'EURUSD'].set_index('datetime')
+                    if "instrument" in df_reset.columns:
+                        df_eur = df_reset[df_reset["instrument"] == "EURUSD"].set_index("datetime")
                         series = df_eur[factor_col] if factor_col in df_eur.columns else df_eur.iloc[:, -1]
                     else:
                         series = df.iloc[:, 0]
             else:
                 series = df.iloc[:, 0]
-            
+
             # Ensure numeric
-            series = pd.to_numeric(series, errors='coerce')
+            series = pd.to_numeric(series, errors="coerce")
             series.name = factor_name
             return series
         except Exception as e:
@@ -307,10 +303,10 @@ class StrategyOrchestrator:
 
     def generate_strategy_code(
         self,
-        factors: List[Dict[str, Any]],
+        factors: list[dict[str, Any]],
         strategy_name: str,
         max_retries: int = 3,
-    ) -> Optional[str]:
+    ) -> str | None:
         """
         Generate strategy code using LLM from factor combinations.
 
@@ -341,6 +337,12 @@ class StrategyOrchestrator:
             user_prompt = user_prompt.replace("{{ additional_context }}", f"Strategy name: {strategy_name}")
             user_prompt = user_prompt.replace("{{ trading_style }}", self.trading_style)
             user_prompt = user_prompt.replace("{{ min_sharpe }}", str(self.min_sharpe))
+
+            if "{{" in user_prompt:
+                unreplaced = [w for w in user_prompt.split() if "{{" in w]
+                logger.warning(
+                    f"Unreplaced template variables in prompt for '{strategy_name}': {unreplaced}"
+                )
             user_prompt = user_prompt.replace("{{ max_drawdown }}", str(self.max_drawdown))
             system_prompt = self.strategy_prompt.get("system", "")
         else:
@@ -371,12 +373,12 @@ class StrategyOrchestrator:
                     last_error = f"Attempt {attempt}: LLM returned empty or invalid code"
                     logger.warning(f"LLM attempt {attempt}/{max_retries} failed: {last_error}")
                 except Exception as e:
-                    last_error = f"Attempt {attempt}: {str(e)}"
+                    last_error = f"Attempt {attempt}: {e!s}"
                     logger.warning(f"LLM attempt {attempt}/{max_retries} failed with exception: {e}")
 
             logger.warning(
                 f"LLM strategy generation failed after {max_retries} attempts. "
-                f"Last error: {last_error}"
+                f"Last error: {last_error}",
             )
 
         # Fallback: generate template code programmatically
@@ -385,10 +387,10 @@ class StrategyOrchestrator:
 
     def _generate_with_llm(
         self,
-        context: Dict[str, Any],
+        context: dict[str, Any],
         attempt: int = 1,
-        feedback: Optional[str] = None,
-    ) -> Optional[str]:
+        feedback: str | None = None,
+    ) -> str | None:
         """
         Generate strategy code using LLM with APIBackend (same as Factor Coder).
 
@@ -406,7 +408,6 @@ class StrategyOrchestrator:
         str or None
             Validated Python strategy code, or None if invalid
         """
-        import json as json_module
 
         # Build user message with optional feedback
         user_content = context.get("user_prompt", "")
@@ -446,14 +447,13 @@ class StrategyOrchestrator:
                 if self._validate_python_code(code):
                     logger.info(f"[DEBUG] Valid Python code extracted ({len(code)} chars)")
                     return code
-                else:
-                    logger.warning(f"JSON 'code' field contains invalid Python (attempt {attempt}). Preview: {code[:200]}")
+                logger.warning(f"JSON 'code' field contains invalid Python (attempt {attempt}). Preview: {code[:200]}")
             else:
                 logger.warning(f"JSON parsed but no valid 'code' field found (attempt {attempt}). Keys: {list(json_data.keys())}")
 
         # === STEP 2: Fallback - Extract Python code block directly (like Factor Coder) ===
         import re
-        code_block_match = re.search(r'```python\s*\n(.*?)\n```', content, re.DOTALL)
+        code_block_match = re.search(r"```python\s*\n(.*?)\n```", content, re.DOTALL)
         if code_block_match:
             code = code_block_match.group(1).strip()
             if code and self._validate_python_code(code):
@@ -463,7 +463,7 @@ class StrategyOrchestrator:
         logger.warning(f"All extraction methods failed (attempt {attempt}). Response preview: {response[:200]}")
         return None
 
-    def _extract_json(self, content: str) -> Optional[Dict[str, Any]]:
+    def _extract_json(self, content: str) -> dict[str, Any] | None:
         """
         Extract JSON object from LLM response content.
 
@@ -491,7 +491,7 @@ class StrategyOrchestrator:
                 pass
 
         # Strategy 2: Find ```json ... ``` blocks
-        json_block_match = re.search(r'```json\s*\n(.*?)\n```', content, re.DOTALL)
+        json_block_match = re.search(r"```json\s*\n(.*?)\n```", content, re.DOTALL)
         if json_block_match:
             try:
                 return json_module.loads(json_block_match.group(1))
@@ -499,7 +499,7 @@ class StrategyOrchestrator:
                 pass
 
         # Strategy 3: Find ```python ... ``` blocks (Qwen often puts JSON in python blocks)
-        python_block_match = re.search(r'```python\s*\n(.*?)\n```', content, re.DOTALL)
+        python_block_match = re.search(r"```python\s*\n(.*?)\n```", content, re.DOTALL)
         if python_block_match:
             block = python_block_match.group(1).strip()
             if block.startswith("{") and block.endswith("}"):
@@ -519,13 +519,13 @@ class StrategyOrchestrator:
                 # Try to fix common JSON issues (trailing commas, unescaped newlines)
                 try:
                     # Remove trailing commas before } or ]
-                    json_str_fixed = re.sub(r',\s*([}\]])', r'\1', json_str)
+                    json_str_fixed = re.sub(r",\s*([}\]])", r"\1", json_str)
                     return json_module.loads(json_str_fixed)
                 except json_module.JSONDecodeError:
                     pass
 
         # Strategy 5: Find ``` ... ``` blocks (any language tag)
-        code_block_match = re.search(r'```\w*\s*\n(.*?)\n```', content, re.DOTALL)
+        code_block_match = re.search(r"```\w*\s*\n(.*?)\n```", content, re.DOTALL)
         if code_block_match:
             block = code_block_match.group(1).strip()
             if block.startswith("{") and block.endswith("}"):
@@ -536,7 +536,7 @@ class StrategyOrchestrator:
 
         return None
 
-    def _extract_code_from_json(self, json_data: Dict[str, Any]) -> Optional[str]:
+    def _extract_code_from_json(self, json_data: dict[str, Any]) -> str | None:
         """
         Extract Python code from parsed JSON data.
 
@@ -559,7 +559,7 @@ class StrategyOrchestrator:
 
         return None
 
-    def _extract_code_from_raw(self, content: str) -> Optional[str]:
+    def _extract_code_from_raw(self, content: str) -> str | None:
         """
         Extract Python code from raw (non-JSON) LLM response.
 
@@ -579,12 +579,12 @@ class StrategyOrchestrator:
         code = content.strip()
 
         # Try to find ```python blocks
-        python_match = re.search(r'```python\s*\n(.*?)\n```', code, re.DOTALL)
+        python_match = re.search(r"```python\s*\n(.*?)\n```", code, re.DOTALL)
         if python_match:
             code = python_match.group(1)
         else:
             # Try generic ``` blocks
-            block_match = re.search(r'```\s*\n(.*?)\n```', code, re.DOTALL)
+            block_match = re.search(r"```\s*\n(.*?)\n```", code, re.DOTALL)
             if block_match:
                 code = block_match.group(1)
 
@@ -636,7 +636,7 @@ class StrategyOrchestrator:
         # Remove non-ASCII characters (emojis, etc.)
         code = code.encode("ascii", "ignore").decode("ascii").strip()
 
-        return code if code else None
+        return code or None
 
     def _validate_python_code(self, code: str) -> bool:
         """
@@ -663,14 +663,14 @@ class StrategyOrchestrator:
             logger.debug(f"Python syntax error: {e}")
             return False
 
-    def _generate_fallback_code(self, context: Dict[str, Any]) -> str:
+    def _generate_fallback_code(self, context: dict[str, Any]) -> str:
         """Generate fallback strategy code programmatically."""
         factor_names = context["factor_names"]
         style_config = "daytrading" if context["trading_style"] == "daytrading" else "swing"
 
         # Build factor assignment code
         factor_assignments = "\n    ".join(
-            [f'"{name}": factors["{name}"]' for name in factor_names if name != "timestamp"]
+            [f'"{name}": factors["{name}"]' for name in factor_names if name != "timestamp"],
         )
 
         code = f'''"""
@@ -717,8 +717,8 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
         return code
 
     def evaluate_strategy(
-        self, strategy_code: str, strategy_name: str, factors: List[Dict[str, Any]]
-    ) -> Dict[str, Any]:
+        self, strategy_code: str, strategy_name: str, factors: list[dict[str, Any]],
+    ) -> dict[str, Any]:
         """
         Evaluate a strategy by executing its code and calculating metrics.
 
@@ -755,23 +755,20 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
                 }
 
             # Align factor values with common index
-            if not factor_values:
-                df_factors = pd.DataFrame()
-            else:
-                # Find common index across all series
-                common_idx = None
-                for name, s in factor_values.items():
-                    if common_idx is None:
-                        common_idx = s.index
-                    else:
-                        common_idx = common_idx.intersection(s.index)
-                
-                if common_idx is not None and len(common_idx) > 100:
-                    df_factors = pd.DataFrame({
-                        name: s.reindex(common_idx) for name, s in factor_values.items()
-                    }).dropna()
+            # Find common index across all series
+            common_idx = None
+            for name, s in factor_values.items():
+                if common_idx is None:
+                    common_idx = s.index
                 else:
-                    df_factors = pd.DataFrame()
+                    common_idx = common_idx.intersection(s.index)
+
+            if common_idx is not None and len(common_idx) > 100:
+                df_factors = pd.DataFrame({
+                    name: s.reindex(common_idx) for name, s in factor_values.items()
+                }).dropna()
+            else:
+                df_factors = pd.DataFrame()
 
             if len(df_factors) < 100:
                 return {
@@ -783,8 +780,8 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
 
             # Convert all factor columns to numeric
             for col in df_factors.columns:
-                df_factors[col] = pd.to_numeric(df_factors[col], errors='coerce')
-            
+                df_factors[col] = pd.to_numeric(df_factors[col], errors="coerce")
+
             # Forward-fill daily factors to match OHLCV 1-min index
             # Many factors are daily (1 value per day), need to ffill to 1-min
             # FIX 6: Track ffill ratio for data quality monitoring
@@ -799,11 +796,11 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
                     f"[DEBUG] {strategy_name}: data quality: "
                     f"original_rows={original_len}, "
                     f"ffill_rows={len(df_factors) - original_len}, "
-                    f"ffill_ratio={ffill_ratio:.2%}"
+                    f"ffill_ratio={ffill_ratio:.2%}",
                 )
 
             df_factors = df_factors.dropna()
-            
+
             if len(df_factors) < 1000:
                 return {
                     "strategy_name": strategy_name,
@@ -811,54 +808,58 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
                     "reason": f"Insufficient numeric data after conversion ({len(df_factors)} rows)",
                     "factors_used": factor_names,
                 }
-            
+
             # close is already loaded above for ffill, reuse it
             # Reindex close to match factor index
             if close is not None:
                 close = close.reindex(df_factors.index)
-            
+
             # Execute strategy code with factor data and close prices
             local_vars = {"factors": df_factors}
             if close is not None:
                 local_vars["close"] = close
-            
+
             try:
                 exec(strategy_code, {"np": np, "pd": pd, "numpy": np}, local_vars)
             except Exception as e:
+                import traceback
+                logger.error(
+                    f"Strategy code execution failed for '{strategy_name}': {e}\n"
+                    f"{traceback.format_exc()[-2000:]}"
+                )
                 return {
                     "strategy_name": strategy_name,
                     "status": "rejected",
-                    "reason": f"Code execution error: {str(e)}",
+                    "reason": f"Code execution error: {e!s}",
                     "factors_used": factor_names,
                 }
 
-            if "signal" not in local_vars:
+            signal = local_vars.get("signal")
+            if signal is None or (isinstance(signal, pd.Series) and signal.empty):
                 return {
                     "strategy_name": strategy_name,
                     "status": "rejected",
-                    "reason": "Strategy did not produce 'signal' variable",
+                    "reason": "Strategy did not produce valid 'signal' variable",
                     "factors_used": factor_names,
                 }
 
-            signal = local_vars["signal"]
-
             logger.info(
                 f"[DEBUG] {strategy_name}: signal stats: "
                 f"len={len(signal)}, "
                 f"long={int((signal > 0).sum())}, "
                 f"short={int((signal < 0).sum())}, "
                 f"flat={int((signal == 0).sum())}, "
-                f"unique={signal.nunique()}"
+                f"unique={signal.nunique()}",
             )
 
             # Delegate all metric computation to the single source of truth.
             # Same formulas as every other backtest path in the repo.
             from rdagent.components.backtesting.vbt_backtest import (
-                backtest_signal_ftmo,
                 DEFAULT_TXN_COST_BPS,
+                backtest_signal_ftmo,
             )
 
-            close = self.load_ohlcv_close()
+            # Reuse the already-loaded close from above; create a synthetic proxy if unavailable
             if close is None:
                 logger.warning("OHLCV data unavailable, using factor-mean proxy")
                 proxy = df_factors.mean(axis=1).astype(float)
@@ -890,7 +891,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
             logger.info(
                 f"[DEBUG] {strategy_name}: bt stats: "
                 f"sharpe={sharpe:.4f} dd={max_dd:.4f} wr={win_rate:.4f} "
-                f"trades={num_real_trades} total_ret={bt['total_return']:.4%}"
+                f"trades={num_real_trades} total_ret={bt['total_return']:.4%}",
             )
 
             metrics = {
@@ -920,7 +921,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
                     f"[DEBUG] {strategy_name}: rejection breakdown: "
                     f"sharpe={sharpe:.4f} (need>={self.min_sharpe}), "
                     f"dd={max_dd:.4f} (need>={self.max_drawdown}), "
-                    f"wr={win_rate:.4f} (need>={self.min_win_rate})"
+                    f"wr={win_rate:.4f} (need>={self.min_win_rate})",
                 )
                 metrics["reason"] = self._get_rejection_reason(sharpe, max_dd, win_rate)
 
@@ -932,7 +933,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
             return {
                 "strategy_name": strategy_name,
                 "status": "rejected",
-                "reason": f"Evaluation error: {str(e)}",
+                "reason": f"Evaluation error: {e!s}",
                 "factors_used": [],
             }
 
@@ -951,7 +952,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
             reasons.append(f"Win Rate {win_rate:.2%} < {self.min_win_rate:.2%}")
         return "; ".join(reasons) if reasons else "Unknown"
 
-    def _generate_strategy_name(self, factors: List[Dict[str, Any]], idx: int) -> str:
+    def _generate_strategy_name(self, factors: list[dict[str, Any]], idx: int) -> str:
         """Generate a strategy name from its factors."""
         # Extract key words from factor names
         words = []
@@ -962,7 +963,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
             for p in parts:
                 # Extract capitalized words
                 cap_words = [w for w in p.split() if w[0:1].isupper()]
-                words.extend(cap_words if cap_words else [p])
+                words.extend(cap_words or [p])
 
         # Take up to 3 unique words
         unique_words = list(dict.fromkeys(words))[:3]
@@ -975,7 +976,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
         count: int = 10,
         workers: int = 2,  # Reduced from 4 to 2 to avoid LLM server overload
         progress_callback=None,
-    ) -> List[Dict[str, Any]]:
+    ) -> list[dict[str, Any]]:
         """
         Generate and evaluate trading strategies.
 
@@ -1028,7 +1029,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
                         logger.info(
                             f"Strategy ACCEPTED: {result['strategy_name']} | "
                             f"Sharpe={result['sharpe_ratio']:.2f} | "
-                            f"DD={result['max_drawdown']:.2%}"
+                            f"DD={result['max_drawdown']:.2%}",
                         )
                     else:
                         # Also save rejected strategies for debugging
@@ -1036,7 +1037,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
                         logger.warning(
                             f"Strategy REJECTED: {result['strategy_name']} - {result.get('reason', 'unknown')} | "
                             f"Sharpe={result.get('sharpe_ratio', 'N/A')} | "
-                            f"DD={result.get('max_drawdown', 'N/A')}"
+                            f"DD={result.get('max_drawdown', 'N/A')}",
                         )
 
                     if progress_callback:
@@ -1052,12 +1053,12 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
 
         logger.info(
             f"Strategy generation complete: {strategies_accepted}/{strategies_generated} accepted "
-            f"({strategies_accepted/max(strategies_generated,1)*100:.1f}%)"
+            f"({strategies_accepted/max(strategies_generated,1)*100:.1f}%)",
         )
 
         return results
 
-    def _generate_strategy_configs(self, factors: List[Dict], count: int) -> List[List[Dict]]:
+    def _generate_strategy_configs(self, factors: list[dict], count: int) -> list[list[dict]]:
         """
         Generate strategy configurations from factor combinations.
 
@@ -1085,7 +1086,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
         np.random.shuffle(configs)
         return configs[: count * 2]  # Generate extras
 
-    def _generate_and_evaluate_single(self, idx: int, factors: List[Dict]) -> Dict[str, Any]:
+    def _generate_and_evaluate_single(self, idx: int, factors: list[dict]) -> dict[str, Any]:
         """Generate and evaluate a single strategy."""
         strategy_name = self._generate_strategy_name(factors, idx + 1)
 
@@ -1108,7 +1109,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
         # by finding optimal entry/exit thresholds, signal smoothing, etc.
         if self.use_optuna:
             initial_status = result.get("status", "rejected")
-            initial_sharpe = result.get("sharpe_ratio", float('-inf'))
+            initial_sharpe = result.get("sharpe_ratio", float("-inf"))
             logger.info(f"Running Optuna optimization for {strategy_name} (initial: {initial_status}, Sharpe={initial_sharpe:.4f})...")
             optimizer = OptunaOptimizer(n_trials=self.optuna_trials)
 
@@ -1117,7 +1118,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
 
             if factor_values is not None:
                 optimized = optimizer.optimize_strategy(result, factor_values)
-                optimized_sharpe = optimized.get("sharpe_ratio", float('-inf'))
+                optimized_sharpe = optimized.get("sharpe_ratio", float("-inf"))
                 optimized_status = optimized.get("status", "rejected")
                 best_params = optimized.get("best_params", {})
 
@@ -1126,14 +1127,14 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
                     improvement = optimized_sharpe - initial_sharpe
                     logger.info(
                         f"Optuna {'RESCUED' if optimized_status == 'accepted' and initial_status == 'rejected' else 'improved'} "
-                        f"{strategy_name}: Sharpe {initial_sharpe:.4f} → {optimized_sharpe:.4f} (+{improvement:.4f})"
+                        f"{strategy_name}: Sharpe {initial_sharpe:.4f} → {optimized_sharpe:.4f} (+{improvement:.4f})",
                     )
 
                     # Re-evaluate with best parameters to get comparable metrics
                     if best_params:
                         patched_code = self._patch_strategy_code(code, best_params)
                         re_eval = self._evaluate_with_patched_code(patched_code, strategy_name, factors)
-                        if re_eval.get("sharpe_ratio", float('-inf')) > initial_sharpe:
+                        if re_eval.get("sharpe_ratio", float("-inf")) > initial_sharpe:
                             result.update(re_eval)
                             result["code"] = patched_code
                             result["best_params"] = best_params
@@ -1142,7 +1143,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
                                 result.pop("reason", None)
                             logger.info(
                                 f"Re-evaluated {strategy_name} with best params: "
-                                f"Sharpe {initial_sharpe:.4f} → {re_eval.get('sharpe_ratio', 0):.4f}"
+                                f"Sharpe {initial_sharpe:.4f} → {re_eval.get('sharpe_ratio', 0):.4f}",
                             )
                         else:
                             result.update(optimized)
@@ -1162,7 +1163,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
 
         return result
 
-    def _prepare_factor_values(self, factors: List[Dict]) -> Optional[pd.DataFrame]:
+    def _prepare_factor_values(self, factors: list[dict]) -> pd.DataFrame | None:
         """Prepare factor values DataFrame for Optuna optimization."""
         factor_values = {}
         for f in factors:
@@ -1181,7 +1182,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
             return df.dropna()
         return None
 
-    def _patch_strategy_code(self, code: str, params: Dict[str, Any]) -> str:
+    def _patch_strategy_code(self, code: str, params: dict[str, Any]) -> str:
         """Patch strategy code with Optuna's best parameters."""
         import re
         patched = code
@@ -1192,26 +1193,26 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
         signal_window = params.get("signal_window", 3)
 
         param_patterns = [
-            (r'entry_thresh\s*=\s*[\d.]+', f'entry_thresh = {entry_thresh}'),
-            (r'exit_thresh\s*=\s*[\d.]+', f'exit_thresh = {exit_thresh}'),
-            (r'window\s*=\s*\d+', f'window = {zscore_window}'),
-            (r'signal_window\s*=\s*\d+', f'signal_window = {signal_window}'),
+            (r"entry_thresh\s*=\s*[\d.]+", f"entry_thresh = {entry_thresh}"),
+            (r"exit_thresh\s*=\s*[\d.]+", f"exit_thresh = {exit_thresh}"),
+            (r"window\s*=\s*\d+", f"window = {zscore_window}"),
+            (r"signal_window\s*=\s*\d+", f"signal_window = {signal_window}"),
         ]
         for pattern, replacement in param_patterns:
             patched = re.sub(pattern, replacement, patched)
 
         # Patch .rolling(N) calls for common window sizes
-        rolling_pattern = r'\.rolling\((\d+)\)'
+        rolling_pattern = r"\.rolling\((\d+)\)"
         def replace_rolling(match):
             val = int(match.group(1))
             if val in (20, 30, 50, 100, 200):
-                return f'.rolling({zscore_window})'
+                return f".rolling({zscore_window})"
             return match.group(0)
         patched = re.sub(rolling_pattern, replace_rolling, patched)
 
         return patched
 
-    def _evaluate_with_patched_code(self, patched_code: str, strategy_name: str, factors: List[Dict]) -> Dict[str, Any]:
+    def _evaluate_with_patched_code(self, patched_code: str, strategy_name: str, factors: list[dict]) -> dict[str, Any]:
         """Re-evaluate strategy with patched parameters using full OHLCV backtest."""
         try:
             factor_names = [f["factor_name"] for f in factors if f["factor_name"] != "timestamp"]
@@ -1222,7 +1223,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
                     factor_values[fname] = series
 
             if not factor_values:
-                return {"sharpe_ratio": float('-inf'), "status": "rejected"}
+                return {"sharpe_ratio": float("-inf"), "status": "rejected"}
 
             common_idx = None
             for name, s in factor_values.items():
@@ -1232,14 +1233,14 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
                     common_idx = common_idx.intersection(s.index)
 
             if common_idx is None or len(common_idx) < 100:
-                return {"sharpe_ratio": float('-inf'), "status": "rejected"}
+                return {"sharpe_ratio": float("-inf"), "status": "rejected"}
 
             df_factors = pd.DataFrame({
                 name: s.reindex(common_idx) for name, s in factor_values.items()
             }).dropna()
 
             for col in df_factors.columns:
-                df_factors[col] = pd.to_numeric(df_factors[col], errors='coerce')
+                df_factors[col] = pd.to_numeric(df_factors[col], errors="coerce")
 
             close = self.load_ohlcv_close()
             if close is not None:
@@ -1247,7 +1248,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
 
             df_factors = df_factors.dropna()
             if len(df_factors) < 1000:
-                return {"sharpe_ratio": float('-inf'), "status": "rejected"}
+                return {"sharpe_ratio": float("-inf"), "status": "rejected"}
 
             if close is not None:
                 close = close.reindex(df_factors.index)
@@ -1256,21 +1257,21 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
             try:
                 exec(patched_code, {"np": np, "pd": pd, "numpy": np}, local_vars)
             except Exception:
-                return {"sharpe_ratio": float('-inf'), "status": "rejected"}
+                return {"sharpe_ratio": float("-inf"), "status": "rejected"}
 
             if "signal" not in local_vars:
-                return {"sharpe_ratio": float('-inf'), "status": "rejected"}
+                return {"sharpe_ratio": float("-inf"), "status": "rejected"}
 
             signal = local_vars["signal"]
 
             from rdagent.components.backtesting.vbt_backtest import (
-                backtest_signal_ftmo,
                 DEFAULT_TXN_COST_BPS,
+                backtest_signal_ftmo,
             )
 
             close_for_bt = close.reindex(signal.index).ffill() if close is not None else None
             if close_for_bt is None:
-                return {"sharpe_ratio": float('-inf'), "status": "rejected"}
+                return {"sharpe_ratio": float("-inf"), "status": "rejected"}
 
             bt = backtest_signal_ftmo(
                 close=close_for_bt,
@@ -1278,7 +1279,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
                 txn_cost_bps=float(os.getenv("TXN_COST_BPS", DEFAULT_TXN_COST_BPS)),
             )
             if bt.get("status") != "success":
-                return {"sharpe_ratio": float('-inf'), "status": "rejected"}
+                return {"sharpe_ratio": float("-inf"), "status": "rejected"}
 
             sharpe = bt["sharpe"]
             max_dd = bt["max_drawdown"]
@@ -1307,9 +1308,9 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
 
         except Exception as e:
             logger.debug(f"Re-evaluation failed for {strategy_name}: {e}")
-            return {"sharpe_ratio": float('-inf'), "status": "rejected"}
+            return {"sharpe_ratio": float("-inf"), "status": "rejected"}
 
-    def _save_strategy(self, result: Dict[str, Any]) -> None:
+    def _save_strategy(self, result: dict[str, Any]) -> None:
         """Save accepted strategy to JSON file."""
         timestamp = int(time.time())
         safe_name = result["strategy_name"].replace("/", "_").replace(" ", "_")[:60]
@@ -1325,7 +1326,7 @@ signal = signal.rolling(window=3, min_periods=1).mean().round().astype(int)
 
         logger.info(f"Saved strategy to {filepath}")
 
-    def get_strategy_summary(self, results: List[Dict[str, Any]]) -> Dict[str, Any]:
+    def get_strategy_summary(self, results: list[dict[str, Any]]) -> dict[str, Any]:
         """
         Generate summary statistics from strategy generation results.
 

rdagent/components/document_reader/document_reader.py

@@ -85,13 +85,16 @@ def load_and_process_one_pdf_by_azure_document_intelligence(
 
 
 def load_and_process_pdfs_by_azure_document_intelligence(path: Path) -> dict[str, str]:
-    assert RD_AGENT_SETTINGS.azure_document_intelligence_key is not None
-    assert RD_AGENT_SETTINGS.azure_document_intelligence_endpoint is not None
+    if RD_AGENT_SETTINGS.azure_document_intelligence_key is None:
+        raise AssertionError("azure_document_intelligence_key must be set")
+    if RD_AGENT_SETTINGS.azure_document_intelligence_endpoint is None:
+        raise AssertionError("azure_document_intelligence_endpoint must be set")
 
     content_dict = {}
     ab_path = path.resolve()
     if ab_path.is_file():
-        assert ".pdf" in ab_path.suffixes, "The file must be a PDF file."
+        if ".pdf" not in ab_path.suffixes:
+            raise ValueError("The file must be a PDF file.")
         proc = load_and_process_one_pdf_by_azure_document_intelligence
         content_dict[str(ab_path)] = proc(
             ab_path,

rdagent/components/loader/task_loader.py

@@ -87,7 +87,8 @@ class ModelWsLoader(WsLoader[ModelTask, ModelFBWorkspace]):
         self.path = Path(path)
 
     def load(self, task: ModelTask) -> ModelFBWorkspace:
-        assert task.name is not None
+        if task.name is None:
+            raise AssertionError("task.name should not be None")
         mti = ModelFBWorkspace(task)
         mti.prepare()
         with open(self.path / f"{task.name}.py", "r") as f:

rdagent/core/utils.py

@@ -83,10 +83,24 @@ def import_class(class_path: str) -> Any:
     Returns
     -------
         class of `class_path`
+
+    Raises
+    ------
+    ImportError
+        If module or class cannot be found.
     """
-    module_path, class_name = class_path.rsplit(".", 1)
-    module = importlib.import_module(module_path)
-    return getattr(module, class_name)
+    try:
+        module_path, class_name = class_path.rsplit(".", 1)
+    except ValueError:
+        raise ImportError(f"Invalid class path: {class_path!r}")
+    try:
+        module = importlib.import_module(module_path)
+    except ModuleNotFoundError as e:
+        raise ImportError(f"Module not found: {module_path!r}") from e
+    try:
+        return getattr(module, class_name)
+    except AttributeError as e:
+        raise ImportError(f"Class not found: {class_name!r} in {module_path!r}") from e
 
 
 class CacheSeedGen:

rdagent/oai/backend/base.py

@@ -541,7 +541,8 @@ class APIBackend(ABC):
         **kwargs,
     ) -> str | list[list[float]]:
         """This function to share operation between embedding and chat completion"""
-        assert not (chat_completion and embedding), "chat_completion and embedding cannot be True at the same time"
+        if chat_completion and embedding:
+            raise ValueError("chat_completion and embedding cannot be True at the same time")
         max_retry = LLM_SETTINGS.max_retry if LLM_SETTINGS.max_retry is not None else max_retry
         timeout_count = 0
         violation_count = 0

rdagent/scenarios/data_science/debug/data.py

@@ -268,7 +268,8 @@ class JsonReducer(DataReducer):
                 parent[key] = sampled  # type: ignore  # parent 是 list，key 是 index, list.__setitem__(key, sampled)
             self.sampled_files.extend([self.extract_filename(i) for i in sampled])
             break
-        assert len(self.sampled_files) > 0
+        if len(self.sampled_files) <= 0:
+            raise AssertionError("sampled_files must contain at least one file")
         return data
 
     def _find_all_lists(

rdagent/scenarios/data_science/example/source_data/arf-12-hours-prediction-task/prepare.py

@@ -56,14 +56,17 @@ sparse.save_npz(public / "test" / "X.npz", X_test)
 sparse.save_npz(public / "train" / "X.npz", X_train)
 df_train.to_csv(public / "train" / "ARF_12h.csv", index=False)
 
-assert (
-    X_train.shape[0] == df_train.shape[0]
-), f"Mismatch: X_train rows ({X_train.shape[0]}) != df_train rows ({df_train.shape[0]})"
-assert (
-    X_test.shape[0] == df_test.shape[0]
-), f"Mismatch: X_test rows ({X_test.shape[0]}) != df_test rows ({df_test.shape[0]})"
-assert df_test.shape[1] == 2, "Public test set should have 2 columns"
-assert df_train.shape[1] == 3, "Public train set should have 3 columns"
-assert len(df_train) + len(df_test) == len(
-    df_label
-), "Length of new_train and new_test should equal length of old_train"
+if X_train.shape[0] != df_train.shape[0]:
+    raise ValueError(
+        f"Mismatch: X_train rows ({X_train.shape[0]}) != df_train rows ({df_train.shape[0]})"
+    )
+if X_test.shape[0] != df_test.shape[0]:
+    raise ValueError(
+        f"Mismatch: X_test rows ({X_test.shape[0]}) != df_test rows ({df_test.shape[0]})"
+    )
+if df_test.shape[1] != 2:
+    raise ValueError("Public test set should have 2 columns")
+if df_train.shape[1] != 3:
+    raise ValueError("Public train set should have 3 columns")
+if len(df_train) + len(df_test) != len(df_label):
+    raise ValueError("Length of new_train and new_test should equal length of old_train")

rdagent/scenarios/data_science/loop.py

@@ -320,7 +320,8 @@ class DataScienceRDLoop(RDLoop):
             # only clean current workspace without affecting other loops.
             for k in "direct_exp_gen", "coding", "running":
                 if k in prev_out and prev_out[k] is not None:
-                    assert isinstance(prev_out[k], DSExperiment)
+                    if not isinstance(prev_out[k], DSExperiment):
+                        raise TypeError(f"prev_out[{k!r}] must be an instance of DSExperiment")
                     clean_workspace(prev_out[k].experiment_workspace.workspace_path)
 
             # Backup the workspace (only necessary files are included)

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

@@ -213,7 +213,8 @@ class DSTrace(Trace[DataScienceScen, KnowledgeBase]):
         self, component: COMPONENT, search_list: list[tuple[DSExperiment, ExperimentFeedback]] = []
     ) -> bool:
         for exp, fb in search_list:
-            assert isinstance(exp.hypothesis, DSHypothesis), "Hypothesis should be DSHypothesis (and not None)"
+            if not isinstance(exp.hypothesis, DSHypothesis):
+                raise TypeError("Hypothesis should be DSHypothesis (and not None)")
             if exp.hypothesis.component == component and fb:
                 return True
         return False

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

@@ -377,7 +377,8 @@ class ExpGen2TraceAndMergeV2(ExpGen):
             if DS_RD_SETTING.enable_multi_version_exp_gen:
                 exp_gen_version_list = DS_RD_SETTING.exp_gen_version_list.split(",")
                 for version in exp_gen_version_list:
-                    assert version in ["v3", "v2", "v1"]
+                    if version not in ["v3", "v2", "v1"]:
+                        raise ValueError(f"version must be 'v1', 'v2', or 'v3', got {version!r}")
 
                 if len(trace.hist) == 0:
                     # set the proposal version for the first sub-trace

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

@@ -339,7 +339,8 @@ class DSProposalV1ExpGen(ExpGen):
             eda_output = sota_exp.experiment_workspace.file_dict.get("EDA.md", None)
         scenario_desc = trace.scen.get_scenario_all_desc(eda_output=eda_output)
 
-        assert sota_exp is not None, "SOTA experiment is not provided."
+        if sota_exp is None:
+            raise ValueError("SOTA experiment is not provided.")
         last_exp = trace.last_exp()
         # exp_and_feedback = trace.hist[-1]
         # last_exp = exp_and_feedback[0]
@@ -445,8 +446,10 @@ class DSProposalV1ExpGen(ExpGen):
                         json_target_type=dict[str, dict[str, str | dict] | str],
                     )
                 )
-                assert "hypothesis_proposal" in resp_dict, "Hypothesis proposal not provided."
-                assert "task_design" in resp_dict, "Task design not provided."
+                if "hypothesis_proposal" not in resp_dict:
+                    raise ValueError("Hypothesis proposal not provided.")
+                if "task_design" not in resp_dict:
+                    raise ValueError("Task design not provided.")
                 task_class = component_info["task_class"]
                 hypothesis_proposal = resp_dict.get("hypothesis_proposal", {})
                 hypothesis = DSHypothesis(
@@ -1149,8 +1152,10 @@ You help users retrieve relevant knowledge from community discussions and public
         )
 
         response_dict = json.loads(response)
-        assert response_dict.get("component") in HypothesisComponent.__members__, f"Invalid component"
-        assert response_dict.get("hypothesis") is not None, f"Invalid hypothesis"
+        if response_dict.get("component") not in HypothesisComponent.__members__:
+            raise ValueError(f"Invalid component: {response_dict.get('component')}")
+        if response_dict.get("hypothesis") is None:
+            raise ValueError("Invalid hypothesis")
         return response_dict
 
     # END: for support llm-based hypothesis selection  -----
@@ -1253,7 +1258,8 @@ You help users retrieve relevant knowledge from community discussions and public
             description=task_desc,
         )
 
-        assert isinstance(task, PipelineTask), f"Task {task_name} is not a PipelineTask, got {type(task)}"
+        if not isinstance(task, PipelineTask):
+            raise TypeError(f"Task {task_name} is not a PipelineTask, got {type(task)}")
         # only for llm with response schema.(TODO: support for non-schema llm?)
         # If the LLM provides a "packages" field (list[str]), compute runtime environment now and cache it for subsequent prompts in later loops.
         if isinstance(task_dict, dict) and "packages" in task_dict and isinstance(task_dict["packages"], list):

rdagent/scenarios/kaggle/experiment/scenario.py

@@ -165,7 +165,8 @@ class KGScenario(Scenario):
         return data_info
 
     def output_format(self, tag=None) -> str:
-        assert tag in [None, "feature", "model"]
+        if tag not in [None, "feature", "model"]:
+            raise ValueError(f"tag must be None, 'feature', or 'model', got {tag!r}")
         feature_output_format = f"""The feature code should output following the format:
 {T(".prompts:kg_feature_output_format").r()}"""
         model_output_format = f"""The model code should output following the format:\n""" + T(
@@ -180,7 +181,8 @@ class KGScenario(Scenario):
             return model_output_format
 
     def interface(self, tag=None) -> str:
-        assert tag in [None, "feature", "XGBoost", "RandomForest", "LightGBM", "NN"]
+        if tag not in [None, "feature", "XGBoost", "RandomForest", "LightGBM", "NN"]:
+            raise ValueError(f"tag must be None, 'feature', 'XGBoost', 'RandomForest', 'LightGBM', or 'NN', got {tag!r}")
         feature_interface = f"""The feature code should follow the interface:
 {T(".prompts:kg_feature_interface").r()}"""
         if tag == "feature":
@@ -195,7 +197,8 @@ class KGScenario(Scenario):
             return model_interface
 
     def simulator(self, tag=None) -> str:
-        assert tag in [None, "feature", "model"]
+        if tag not in [None, "feature", "model"]:
+            raise ValueError(f"tag must be None, 'feature', or 'model', got {tag!r}")
 
         kg_feature_simulator = (
             "The feature code will be sent to the simulator:\n" + T(".prompts:kg_feature_simulator").r()

rdagent/scenarios/kaggle/kaggle_crawler.py

@@ -87,7 +87,12 @@ def crawl_descriptions(
         content = e.get_attribute("innerHTML")
         contents.append(content)
 
-    assert len(subtitles) == len(contents) + 1 and subtitles[-1] == "Citation"
+    if not (len(subtitles) == len(contents) + 1 and subtitles[-1] == "Citation"):
+        raise AssertionError(
+            f"Expected len(contents)+1 == len(subtitles) and last subtitle == 'Citation', "
+            f"got len(subtitles)={len(subtitles)}, len(contents)={len(contents)}, "
+            f"last subtitle={subtitles[-1]!r}"
+        )
     for i in range(len(subtitles) - 1):
         descriptions[subtitles[i]] = contents[i]
 

rdagent/scenarios/kaggle/proposal/proposal.py

@@ -307,7 +307,8 @@ class KGHypothesisGen(FactorAndModelHypothesisGen):
 class KGHypothesis2Experiment(FactorAndModelHypothesis2Experiment):
     def prepare_context(self, hypothesis: Hypothesis, trace: Trace) -> Tuple[dict, bool]:
         scenario = trace.scen.get_scenario_all_desc(filtered_tag="hypothesis_and_experiment")
-        assert isinstance(hypothesis, KGHypothesis)
+        if not isinstance(hypothesis, KGHypothesis):
+            raise TypeError("hypothesis must be an instance of KGHypothesis")
         experiment_output_format = (
             T("scenarios.kaggle.prompts:feature_experiment_output_format").r()
             if hypothesis.action in [KG_ACTION_FEATURE_ENGINEERING, KG_ACTION_FEATURE_PROCESSING]

rdagent/scenarios/qlib/developer/factor_runner.py

@@ -1,7 +1,7 @@
-import sys
-import os
 import logging
+import os
 from pathlib import Path
+
 """
 Qlib Factor Runner - Executes factor backtests in Docker.
 
@@ -11,15 +11,8 @@ NOTE: The @cache_with_pickle decorator was REMOVED from develop() because:
 - Docker-level caching (QlibDockerConf.enable_cache=False) is sufficient
 - The pickle cache caused 240+ factor generations but ZERO Docker backtests
 """
-from pathlib import Path
-from typing import Optional
 
 import pandas as pd
-from pandarallel import pandarallel
-
-
-pandarallel.initialize(verbose=1)
-
 from rdagent.app.qlib_rd_loop.conf import FactorBasePropSetting
 from rdagent.components.runner import CachedRunner
 from rdagent.core.exception import FactorEmptyError
@@ -74,7 +67,7 @@ def _shift_daily_constant_factor_if_needed(factor_col: "pd.Series", factor_name:
 
         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."
+            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
@@ -122,13 +115,13 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
     """
 
     def calculate_information_coefficient(
-        self, concat_feature: pd.DataFrame, SOTA_feature_column_size: int, new_feature_columns_size: int
+        self, concat_feature: pd.DataFrame, SOTA_feature_column_size: int, new_feature_columns_size: int,
     ) -> pd.DataFrame:
         res = pd.Series(index=range(SOTA_feature_column_size * new_feature_columns_size))
         for col1 in range(SOTA_feature_column_size):
             for col2 in range(SOTA_feature_column_size, SOTA_feature_column_size + new_feature_columns_size):
                 res.loc[col1 * new_feature_columns_size + col2 - SOTA_feature_column_size] = concat_feature.iloc[
-                    :, col1
+                    :, col1,
                 ].corr(concat_feature.iloc[:, col2])
         return res
 
@@ -137,16 +130,21 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
         # if the IC is larger than a threshold, remove the new_feature column
         # return the new_feature
 
+        from pandarallel import pandarallel
+        pandarallel.initialize(verbose=1)
+
         concat_feature = pd.concat([SOTA_feature, new_feature], axis=1)
         IC_max = (
             concat_feature.groupby("datetime")
             .parallel_apply(
-                lambda x: self.calculate_information_coefficient(x, SOTA_feature.shape[1], new_feature.shape[1])
+                lambda x: self.calculate_information_coefficient(x, SOTA_feature.shape[1], new_feature.shape[1]),
             )
             .mean()
         )
         IC_max.index = pd.MultiIndex.from_product([range(SOTA_feature.shape[1]), range(new_feature.shape[1])])
         IC_max = IC_max.unstack().max(axis=0)
+        if not hasattr(IC_max, "index"):
+            return new_feature
         return new_feature.iloc[:, IC_max[IC_max < 0.99].index]
 
     def develop(self, exp: QlibFactorExperiment) -> QlibFactorExperiment:
@@ -161,7 +159,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
         self._ensure_results_dirs()
 
         if exp.based_experiments and exp.based_experiments[-1].result is None:
-            logger.info(f"Baseline experiment execution ...")
+            logger.info("Baseline experiment execution ...")
             exp.based_experiments[-1] = self.develop(exp.based_experiments[-1])
 
         fbps = FactorBasePropSetting()
@@ -185,11 +183,11 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
                 base_exp for base_exp in exp.based_experiments if isinstance(base_exp, QlibFactorExperiment)
             ]
             if len(sota_factor_experiments_list) > 1:
-                logger.info(f"SOTA factor processing ...")
+                logger.info("SOTA factor processing ...")
                 SOTA_factor = process_factor_data(sota_factor_experiments_list)
 
             # Process the new factors data
-            logger.info(f"New factor processing ...")
+            logger.info("New factor processing ...")
             new_factors = process_factor_data(exp)
 
             if new_factors.empty:
@@ -200,7 +198,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
                 new_factors = self.deduplicate_new_factors(SOTA_factor, new_factors)
                 if new_factors.empty:
                     raise FactorEmptyError(
-                        "The factors generated in this round are highly similar to the previous factors. Please change the direction for creating new factors."
+                        "The factors generated in this round are highly similar to the previous factors. Please change the direction for creating new factors.",
                     )
                 combined_factors = pd.concat([SOTA_factor, new_factors], axis=1).dropna()
             else:
@@ -211,7 +209,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
             combined_factors = combined_factors.loc[:, ~combined_factors.columns.duplicated(keep="last")]
             new_columns = pd.MultiIndex.from_product([["feature"], combined_factors.columns])
             combined_factors.columns = new_columns
-            logger.info(f"Factor data processing completed.")
+            logger.info("Factor data processing completed.")
 
             num_features = len(exp.base_features) + len(combined_factors.columns)
 
@@ -230,10 +228,10 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
                     sota_model_exp = base_exp
                     exist_sota_model_exp = True
                     break
-            logger.info(f"Experiment execution ...")
+            logger.info("Experiment execution ...")
             if exist_sota_model_exp:
                 exp.experiment_workspace.inject_files(
-                    **{"model.py": sota_model_exp.sub_workspace_list[0].file_dict["model.py"]}
+                    **{"model.py": sota_model_exp.sub_workspace_list[0].file_dict["model.py"]},
                 )
                 sota_training_hyperparameters = sota_model_exp.sub_tasks[0].training_hyperparameters
                 if sota_training_hyperparameters:
@@ -244,19 +242,19 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
                             "early_stop": str(sota_training_hyperparameters.get("early_stop", 10)),
                             "batch_size": str(sota_training_hyperparameters.get("batch_size", 256)),
                             "weight_decay": str(sota_training_hyperparameters.get("weight_decay", 0.0001)),
-                        }
+                        },
                     )
                 sota_model_type = sota_model_exp.sub_tasks[0].model_type
                 if sota_model_type == "TimeSeries":
                     env_to_use.update(
-                        {"dataset_cls": "TSDatasetH", "num_features": num_features, "step_len": 20, "num_timesteps": 20}
+                        {"dataset_cls": "TSDatasetH", "num_features": num_features, "step_len": 20, "num_timesteps": 20},
                     )
                 elif sota_model_type == "Tabular":
                     env_to_use.update({"dataset_cls": "DatasetH", "num_features": num_features})
 
                 # model + combined factors
                 result, stdout = exp.experiment_workspace.execute(
-                    qlib_config_name="conf_combined_factors_sota_model.yaml", run_env=env_to_use
+                    qlib_config_name="conf_combined_factors_sota_model.yaml", run_env=env_to_use,
                 )
             else:
                 # LGBM + combined factors
@@ -265,7 +263,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
                     run_env=env_to_use,
                 )
         else:
-            logger.info(f"Experiment execution ...")
+            logger.info("Experiment execution ...")
             if exp.base_feature_codes:
                 factors = process_factor_data(exp)
                 factors = factors.sort_index()
@@ -275,7 +273,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
                 target_path = exp.experiment_workspace.workspace_path / "combined_factors_df.parquet"
                 # Save the combined factors to the workspace
                 factors.to_parquet(target_path, engine="pyarrow")
-                logger.info(f"Factor data processing completed.")
+                logger.info("Factor data processing completed.")
                 result, stdout = exp.experiment_workspace.execute(
                     qlib_config_name="conf_combined_factors.yaml",
                     run_env=env_to_use,
@@ -288,10 +286,10 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
 
         # Handle Qlib Docker backtest failure gracefully
         if result is None:
-            factor_name = getattr(exp.hypothesis, 'hypothesis', 'unknown')
+            factor_name = getattr(exp.hypothesis, "hypothesis", "unknown")
             logger.warning(
                 f"Qlib Docker backtest returned None for '{factor_name}'. "
-                f"Attempting direct factor evaluation..."
+                f"Attempting direct factor evaluation...",
             )
 
             # Try to compute metrics directly from the factor's result.h5
@@ -303,7 +301,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
             else:
                 logger.error(
                     f"Both Qlib Docker backtest and direct evaluation failed for '{factor_name}'. "
-                    f"Skipping this factor and continuing."
+                    f"Skipping this factor and continuing.",
                 )
                 # Save failed run info for debugging
                 self._save_failed_run(exp, stdout, error_type="result_none")
@@ -321,7 +319,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
         if validation_result.get("has_issues"):
             logger.warning(
                 f"Result validation warnings for factor '{getattr(exp.hypothesis, 'hypothesis', 'unknown')}': "
-                f"{validation_result['warnings']}"
+                f"{validation_result['warnings']}",
             )
             # Save warning info for debugging
             self._save_failed_run(exp, stdout, error_type="validation_warnings", validation=validation_result)
@@ -372,43 +370,43 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
         details = {}
 
         factor_name = "unknown"
-        if hasattr(exp, 'hypothesis') and exp.hypothesis is not None:
-            factor_name = getattr(exp.hypothesis, 'hypothesis', 'unknown')
+        if hasattr(exp, "hypothesis") and exp.hypothesis is not None:
+            factor_name = getattr(exp.hypothesis, "hypothesis", "unknown")
 
         if isinstance(result, pd.Series):
             # Check IC
-            ic_value = result.get('IC', None)
-            details['ic_raw'] = ic_value
+            ic_value = result.get("IC", None)
+            details["ic_raw"] = ic_value
             if ic_value is None or (isinstance(ic_value, float) and (ic_value != ic_value)):  # NaN check
                 warnings.append("IC is None/NaN — factor has no predictive power")
             else:
                 try:
                     ic_float = float(ic_value)
-                    details['ic'] = ic_float
+                    details["ic"] = ic_float
                     if abs(ic_float) < 0.001:
                         warnings.append(
-                            f"IC is near zero ({ic_float:.6f}) — factor may not predict returns"
+                            f"IC is near zero ({ic_float:.6f}) — factor may not predict returns",
                         )
                 except (ValueError, TypeError):
                     warnings.append(f"IC value is not numeric: {ic_value}")
 
             # Check positions (1day.pos)
-            pos_value = result.get('1day.pos', None)
-            details['positions_raw'] = pos_value
+            pos_value = result.get("1day.pos", None)
+            details["positions_raw"] = pos_value
             if pos_value is not None:
                 try:
                     pos_float = float(pos_value)
-                    details['positions'] = pos_float
+                    details["positions"] = pos_float
                     if pos_float == 0:
                         warnings.append(
                             "1day.pos == 0 — model opened ZERO positions (stayed neutral). "
                             "Possible causes: (1) topk too high for single-asset, "
-                            "(2) signal threshold too restrictive, (3) no valid predictions"
+                            "(2) signal threshold too restrictive, (3) no valid predictions",
                         )
                     elif pos_float < 10:
                         warnings.append(
                             f"1day.pos = {pos_float:.0f} — very few positions opened. "
-                            f"Check signal threshold and topk settings"
+                            f"Check signal threshold and topk settings",
                         )
                 except (ValueError, TypeError):
                     pass  # pos might be a string
@@ -416,24 +414,24 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
             # Check if result is essentially empty (all values None or NaN)
             non_null_count = result.notna().sum()
             total_count = len(result)
-            details['non_null_metrics'] = int(non_null_count)
-            details['total_metrics'] = int(total_count)
+            details["non_null_metrics"] = int(non_null_count)
+            details["total_metrics"] = int(total_count)
             if non_null_count < 3:
                 warnings.append(
                     f"Result has only {non_null_count}/{total_count} non-null metrics — "
-                    f"backtest likely produced empty results"
+                    f"backtest likely produced empty results",
                 )
 
             # Check for key metrics
-            required_metrics = ['IC', '1day.excess_return_with_cost.shar', '1day.pos']
+            required_metrics = ["IC", "1day.excess_return_with_cost.shar", "1day.pos"]
             for metric_name in required_metrics:
                 val = result.get(metric_name, None)
-                details[f'has_{metric_name}'] = val is not None
+                details[f"has_{metric_name}"] = val is not None
 
         elif isinstance(result, dict):
             # Dict-based result validation
-            ic_value = result.get('IC', result.get('ic', None))
-            details['ic_raw'] = ic_value
+            ic_value = result.get("IC", result.get("ic", None))
+            details["ic_raw"] = ic_value
             if ic_value is None:
                 warnings.append("IC is None — factor has no predictive power")
 
@@ -443,7 +441,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
             "details": details,
         }
 
-    def _evaluate_factor_directly(self, exp, stdout: str) -> Optional[pd.Series]:
+    def _evaluate_factor_directly(self, exp, stdout: str) -> pd.Series | None:
         """
         Evaluate factor directly from its result.h5 file when Qlib Docker fails.
 
@@ -475,7 +473,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
             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'):
+                    if ws is not None and hasattr(ws, "workspace_path"):
                         candidate = ws.workspace_path / "result.h5"
                         if candidate.exists():
                             workspace_path = ws.workspace_path
@@ -579,7 +577,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
 
             logger.info(
                 f"Direct evaluation: IC={ic:.6f}, Sharpe={sharpe:.4f}, "
-                f"AnnRet={annualized_return:.4f}%, WR={win_rate:.2%}"
+                f"AnnRet={annualized_return:.4f}%, WR={win_rate:.2%}",
             )
             return result
 
@@ -588,7 +586,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
             return None
 
     def _save_failed_run(self, exp, stdout: str, error_type: str = "unknown",
-                         validation: Optional[dict] = None) -> None:
+                         validation: dict | None = None) -> None:
         """
         Save failed run information to results/failed_runs.json for debugging.
 
@@ -615,20 +613,20 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
 
             # Get factor name
             factor_name = "unknown"
-            if hasattr(exp, 'hypothesis') and exp.hypothesis is not None:
-                factor_name = getattr(exp.hypothesis, 'hypothesis', 'unknown')
+            if hasattr(exp, "hypothesis") and exp.hypothesis is not None:
+                factor_name = getattr(exp.hypothesis, "hypothesis", "unknown")
 
             # Build failed run record
             failed_record = {
                 "timestamp": datetime.now().isoformat(),
                 "factor_name": factor_name,
                 "error_type": error_type,
-                "stdout": stdout if stdout else "(empty)",
+                "stdout": stdout or "(empty)",
                 "validation": validation,
                 "experiment_details": {
-                    "base_features": list(getattr(exp, 'base_features', {}).keys()) if hasattr(exp, 'base_features') else [],
-                    "hypothesis": getattr(exp.hypothesis, 'hypothesis', str(getattr(exp, 'hypothesis', 'N/A')))
-                                  if hasattr(exp, 'hypothesis') else "N/A",
+                    "base_features": list(getattr(exp, "base_features", {}).keys()) if hasattr(exp, "base_features") else [],
+                    "hypothesis": getattr(exp.hypothesis, "hypothesis", str(getattr(exp, "hypothesis", "N/A")))
+                                  if hasattr(exp, "hypothesis") else "N/A",
                 },
             }
 
@@ -651,11 +649,11 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
 
             failed_file.write_text(
                 json.dumps(existing_records, indent=2, default=str, ensure_ascii=False),
-                encoding="utf-8"
+                encoding="utf-8",
             )
             logger.info(
                 f"Failed run saved: {factor_name} (type={error_type}) "
-                f"→ {failed_file}"
+                f"→ {failed_file}",
             )
 
         except Exception as e:
@@ -678,23 +676,23 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
             containing metric names like 'IC', '1day.excess_return_with_cost.shar', etc.
         """
         try:
-            import json
-            import pandas as pd
             from pathlib import Path
+
+            import pandas as pd
             from rdagent.components.backtesting import ResultsDatabase
 
             # 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'])
+            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):
+            if getattr(exp, "rejected_by_protection", False):
                 logger.info(
                     f"Factor rejected by protection, skipping DB save: "
-                    f"{getattr(exp, 'protection_reason', 'unknown')}"
+                    f"{getattr(exp, 'protection_reason', 'unknown')}",
                 )
                 return
 
@@ -710,47 +708,47 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
             # Extract metrics from result (pd.Series from qlib_res.csv)
             metrics = {}
             if isinstance(result, pd.Series):
-                metrics['ic'] = self._safe_float(result.get('IC', None))
-                metrics['sharpe_ratio'] = self._safe_float(
-                    result.get('1day.excess_return_with_cost.shar',
-                    result.get('1day.excess_return_with_cost.sharpe', None))
+                metrics["ic"] = self._safe_float(result.get("IC", None))
+                metrics["sharpe_ratio"] = self._safe_float(
+                    result.get("1day.excess_return_with_cost.shar",
+                    result.get("1day.excess_return_with_cost.sharpe", None)),
                 )
-                metrics['annualized_return'] = self._safe_float(
-                    result.get('1day.excess_return_with_cost.annualized_return', None)
+                metrics["annualized_return"] = self._safe_float(
+                    result.get("1day.excess_return_with_cost.annualized_return", None),
                 )
-                metrics['max_drawdown'] = self._safe_float(
-                    result.get('1day.excess_return_with_cost.max_drawdown', None)
+                metrics["max_drawdown"] = self._safe_float(
+                    result.get("1day.excess_return_with_cost.max_drawdown", None),
                 )
-                metrics['win_rate'] = self._safe_float(result.get('win_rate', None))
-                metrics['information_ratio'] = self._safe_float(
-                    result.get('1day.excess_return_with_cost.information_ratio', None)
+                metrics["win_rate"] = self._safe_float(result.get("win_rate", None))
+                metrics["information_ratio"] = self._safe_float(
+                    result.get("1day.excess_return_with_cost.information_ratio", None),
                 )
-                metrics['volatility'] = self._safe_float(
-                    result.get('1day.excess_return_with_cost.std',
-                    result.get('1day.excess_return_with_cost.volatility', None))
+                metrics["volatility"] = self._safe_float(
+                    result.get("1day.excess_return_with_cost.std",
+                    result.get("1day.excess_return_with_cost.volatility", None)),
                 )
                 # Store raw metrics for JSON export
-                metrics['raw_metrics'] = result.to_dict()
+                metrics["raw_metrics"] = result.to_dict()
             elif isinstance(result, dict):
-                metrics['ic'] = self._safe_float(result.get('IC', result.get('ic', None)))
-                metrics['sharpe_ratio'] = self._safe_float(
-                    result.get('sharpe', result.get('sharpe_ratio', None))
+                metrics["ic"] = self._safe_float(result.get("IC", result.get("ic", None)))
+                metrics["sharpe_ratio"] = self._safe_float(
+                    result.get("sharpe", result.get("sharpe_ratio", None)),
                 )
-                metrics['annualized_return'] = self._safe_float(result.get('annualized_return', None))
-                metrics['max_drawdown'] = self._safe_float(result.get('max_drawdown', None))
-                metrics['win_rate'] = self._safe_float(result.get('win_rate', None))
-                metrics['information_ratio'] = None
-                metrics['volatility'] = None
-                metrics['raw_metrics'] = result
+                metrics["annualized_return"] = self._safe_float(result.get("annualized_return", None))
+                metrics["max_drawdown"] = self._safe_float(result.get("max_drawdown", None))
+                metrics["win_rate"] = self._safe_float(result.get("win_rate", None))
+                metrics["information_ratio"] = None
+                metrics["volatility"] = None
+                metrics["raw_metrics"] = result
 
             # Result validation before saving (warnings, not blocking)
             self._log_result_warnings(factor_name, result, metrics)
 
             # Only save if we have at least IC or Sharpe
-            if metrics.get('ic') is None and metrics.get('sharpe_ratio') is None:
+            if metrics.get("ic") is None and metrics.get("sharpe_ratio") is None:
                 logger.warning(
                     f"No valid IC/Sharpe for factor '{factor_name}', skipping DB save. "
-                    f"IC={metrics.get('ic')}, Sharpe={metrics.get('sharpe_ratio')}"
+                    f"IC={metrics.get('ic')}, Sharpe={metrics.get('sharpe_ratio')}",
                 )
                 return
 
@@ -761,19 +759,19 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
             db_file = db_path / "backtest_results.db"
 
             # Parallel run isolation: use run-specific subdirectory if PARALLEL_RUN_ID is set
-            run_id = os.getenv("PARALLEL_RUN_ID", "0")
-            if run_id != "0":
+            parallel_run_id = os.getenv("PARALLEL_RUN_ID", "0")
+            if parallel_run_id != "0":
                 # For parallel runs, save to isolated results directory
-                isolated_db_path = project_root / "results" / "runs" / f"run{run_id}" / "db"
+                isolated_db_path = project_root / "results" / "runs" / f"run{parallel_run_id}" / "db"
                 isolated_db_path.mkdir(parents=True, exist_ok=True)
                 db_file = isolated_db_path / "backtest_results.db"
 
             # Save to database
             db = ResultsDatabase(db_path=str(db_file))
-            run_id = db.add_backtest(factor_name=factor_name[:100], metrics=metrics)
+            db_run_id = db.add_backtest(factor_name=factor_name[:100], metrics=metrics)
             logger.info(
                 f"Factor result saved to DB: {factor_name[:60]} "
-                f"(IC={metrics.get('ic')}, Sharpe={metrics.get('sharpe_ratio')}, run_id={run_id})"
+                f"(IC={metrics.get('ic')}, Sharpe={metrics.get('sharpe_ratio')}, run_id={db_run_id})"
             )
 
             # Extract factor code and description from experiment
@@ -781,10 +779,10 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
 
             # Also write a JSON summary to results/factors/ for file-based access
             self._save_factor_json(
-                factor_name, metrics, run_id,
+                factor_name, metrics, db_run_id,
                 factor_code=factor_code,
                 factor_description=factor_description,
-                exp=exp
+                exp=exp,
             )
 
             # Save factor values as parquet for strategy building
@@ -796,7 +794,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
             import traceback
             logger.error(
                 f"Database save failed for factor '{getattr(exp.hypothesis, 'hypothesis', 'unknown')}': {e}\n"
-                f"Traceback: {traceback.format_exc()}"
+                f"Traceback: {traceback.format_exc()}",
             )
 
     def _save_factor_json(self, factor_name: str, metrics: dict, run_id: int,
@@ -907,14 +905,14 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
                 factor_description = match.group(1).strip()[:500]
             else:
                 # Try comments
-                lines = factor_code.split('\n')
+                lines = factor_code.split("\n")
                 desc_lines = []
                 for line in lines[:20]:
                     stripped = line.strip()
-                    if stripped.startswith('#') and not stripped.startswith('#!'):
+                    if stripped.startswith("#") and not stripped.startswith("#!"):
                         desc_lines.append(stripped[1:].strip())
                 if desc_lines:
-                    factor_description = ' '.join(desc_lines)[:500]
+                    factor_description = " ".join(desc_lines)[:500]
 
         return factor_code, factor_description
 
@@ -926,8 +924,9 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
         the complete backtest range (not just the debug 2024 subset).
         """
         import os as _os
-        import subprocess
         import shutil
+        import subprocess
+        import sys
         import tempfile
 
         try:
@@ -935,7 +934,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
             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'):
+                    if ws is not None and hasattr(ws, "workspace_path"):
                         fp = ws.workspace_path / "factor.py"
                         if fp.exists():
                             workspace_path = ws.workspace_path
@@ -967,12 +966,17 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
                 shutil.copy(str(full_data), str(tmp / "intraday_pv.h5"))
 
                 ret = subprocess.run(
-                    ["sys.executable", "factor.py"],
+                    [sys.executable, "factor.py"],
                     cwd=str(tmp),
                     capture_output=True,
                     timeout=300,
+                    check=False,
                 )
                 if ret.returncode != 0:
+                    logger.warning(
+                        f"Full-data factor run failed (exit {ret.returncode}): "
+                        f"{ret.stderr[:500] if ret.stderr else '(no stderr)'}"
+                    )
                     # Fall back to debug-data result if full-data run fails
                     result_h5 = workspace_path / "result.h5"
                     if not result_h5.exists():
@@ -1023,7 +1027,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
         warnings_list = []
 
         # Check IC
-        ic = metrics.get('ic')
+        ic = metrics.get("ic")
         if ic is None:
             warnings_list.append("IC is None — factor has no predictive power")
         elif abs(ic) < 0.001:
@@ -1031,7 +1035,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
 
         # Check positions (1day.pos) — CRITICAL for EURUSD
         if isinstance(result, pd.Series):
-            pos_value = result.get('1day.pos', None)
+            pos_value = result.get("1day.pos", None)
             if pos_value is not None:
                 try:
                     pos_float = float(pos_value)
@@ -1039,23 +1043,23 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
                         warnings_list.append(
                             "WARNING: 1day.pos == 0 — ZERO positions opened! "
                             "Model stayed completely neutral. Check Qlib config: "
-                            "ensure topk=1 and market=eurusd for single-asset trading."
+                            "ensure topk=1 and market=eurusd for single-asset trading.",
                         )
                     elif pos_float < 10:
                         warnings_list.append(
                             f"Low position count: 1day.pos = {pos_float:.0f} — "
-                            f"model traded very rarely"
+                            f"model traded very rarely",
                         )
                 except (ValueError, TypeError):
                     pass
 
         # Check Sharpe
-        sharpe = metrics.get('sharpe_ratio')
+        sharpe = metrics.get("sharpe_ratio")
         if sharpe is not None and abs(sharpe) < 0.1:
             warnings_list.append(f"Sharpe near zero ({sharpe:.4f}) — no risk-adjusted edge")
 
         # Check max drawdown
-        mdd = metrics.get('max_drawdown')
+        mdd = metrics.get("max_drawdown")
         if mdd is not None and mdd < -0.5:
             warnings_list.append(f"Extreme drawdown: {mdd:.2%} — high risk factor")
 
@@ -1070,7 +1074,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
             return None
         try:
             f = float(value)
-            if pd.isna(f) or f == float('inf') or f == float('-inf'):
+            if pd.isna(f) or f == float("inf") or f == float("-inf"):
                 return None
             return f
         except (ValueError, TypeError):
@@ -1113,7 +1117,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
 
         if protection_result.should_block:
             logger.warning(
-                f"Factor {factor_name} rejected by protection manager: {protection_result.reason}"
+                f"Factor {factor_name} rejected by protection manager: {protection_result.reason}",
             )
             # Mark factor as rejected by protection
             exp.rejected_by_protection = True
@@ -1138,8 +1142,8 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
         from pathlib import Path
 
         factor_name = "unknown"
-        if hasattr(exp, 'hypothesis') and exp.hypothesis is not None:
-            factor_name = getattr(exp.hypothesis, 'hypothesis', 'unknown')
+        if hasattr(exp, "hypothesis") and exp.hypothesis is not None:
+            factor_name = getattr(exp.hypothesis, "hypothesis", "unknown")
 
         # Build log entry
         log_entry = {
@@ -1151,42 +1155,42 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
             "annualized_return": None,
             "max_drawdown": None,
             "win_rate": None,
-            "rejected_by_protection": getattr(exp, 'rejected_by_protection', False),
-            "protection_reason": getattr(exp, 'protection_reason', None),
+            "rejected_by_protection": getattr(exp, "rejected_by_protection", False),
+            "protection_reason": getattr(exp, "protection_reason", None),
         }
 
         # Extract metrics if available
         if result is not None:
-            if hasattr(result, 'get'):  # pd.Series or dict
-                ic_val = result.get('IC', result.get('ic', None))
-                log_entry['ic'] = self._safe_float(ic_val) if ic_val is not None else None
+            if hasattr(result, "get"):  # pd.Series or dict
+                ic_val = result.get("IC", result.get("ic", None))
+                log_entry["ic"] = self._safe_float(ic_val) if ic_val is not None else None
 
-                sharpe_val = result.get('1day.excess_return_with_cost.shar',
-                                        result.get('1day.excess_return_with_cost.sharpe',
-                                        result.get('sharpe', None)))
-                log_entry['sharpe'] = self._safe_float(sharpe_val) if sharpe_val is not None else None
+                sharpe_val = result.get("1day.excess_return_with_cost.shar",
+                                        result.get("1day.excess_return_with_cost.sharpe",
+                                        result.get("sharpe", None)))
+                log_entry["sharpe"] = self._safe_float(sharpe_val) if sharpe_val is not None else None
 
-                ann_ret = result.get('1day.excess_return_with_cost.annualized_return',
-                                     result.get('annualized_return', None))
-                log_entry['annualized_return'] = self._safe_float(ann_ret) if ann_ret is not None else None
+                ann_ret = result.get("1day.excess_return_with_cost.annualized_return",
+                                     result.get("annualized_return", None))
+                log_entry["annualized_return"] = self._safe_float(ann_ret) if ann_ret is not None else None
 
-                mdd = result.get('1day.excess_return_with_cost.max_drawdown',
-                                 result.get('max_drawdown', None))
-                log_entry['max_drawdown'] = self._safe_float(mdd) if mdd is not None else None
+                mdd = result.get("1day.excess_return_with_cost.max_drawdown",
+                                 result.get("max_drawdown", None))
+                log_entry["max_drawdown"] = self._safe_float(mdd) if mdd is not None else None
 
-                wr = result.get('win_rate', None)
-                log_entry['win_rate'] = self._safe_float(wr) if wr is not None else None
+                wr = result.get("win_rate", None)
+                log_entry["win_rate"] = self._safe_float(wr) if wr is not None else None
 
             # Determine status
-            if log_entry['ic'] is not None or log_entry['sharpe'] is not None:
-                log_entry['status'] = "success"
-            elif getattr(exp, 'rejected_by_protection', False):
-                log_entry['status'] = "rejected_protection"
+            if log_entry["ic"] is not None or log_entry["sharpe"] is not None:
+                log_entry["status"] = "success"
+            elif getattr(exp, "rejected_by_protection", False):
+                log_entry["status"] = "rejected_protection"
             else:
-                log_entry['status'] = "no_valid_metrics"
+                log_entry["status"] = "no_valid_metrics"
         else:
-            log_entry['status'] = "execution_failed"
-            log_entry['reason'] = "Result was None"
+            log_entry["status"] = "execution_failed"
+            log_entry["reason"] = "Result was None"
 
         # Write to results/logs/
         try:
@@ -1209,7 +1213,7 @@ class QlibFactorRunner(CachedRunner[QlibFactorExperiment]):
 
             logger.info(
                 f"Run log written for '{factor_name[:50]}': "
-                f"status={log_entry['status']}, IC={log_entry['ic']}, Sharpe={log_entry['sharpe']}"
+                f"status={log_entry['status']}, IC={log_entry['ic']}, Sharpe={log_entry['sharpe']}",
             )
         except Exception as e:
             logger.error(f"Failed to write run log: {e}")

rdagent/scenarios/qlib/developer/model_runner.py

@@ -203,12 +203,14 @@ class QlibModelRunner(CachedRunner[QlibModelExperiment]):
 
             # Save to database
             db = ResultsDatabase()
-            run_id = db.add_backtest(factor_name=factor_name[:100], metrics=metrics)
-            logger.info(
-                f"Model result saved to DB: {factor_name[:50]} "
-                f"(IC={metrics.get('ic')}, Sharpe={metrics.get('sharpe_ratio')}, run_id={run_id})"
-            )
-            db.close()
+            try:
+                run_id = db.add_backtest(factor_name=factor_name[:100], metrics=metrics)
+                logger.info(
+                    f"Model result saved to DB: {factor_name[:50]} "
+                    f"(IC={metrics.get('ic')}, Sharpe={metrics.get('sharpe_ratio')}, run_id={run_id})"
+                )
+            finally:
+                db.close()
 
         except Exception as e:
             logger.warning(f"Database save failed for model {getattr(exp.hypothesis, 'hypothesis', 'unknown')}: {e}")

rdagent/scenarios/qlib/experiment/quant_experiment.py

@@ -56,7 +56,8 @@ class QlibQuantScenario(Scenario):
         )
 
     def background(self, tag=None) -> str:
-        assert tag in [None, "factor", "model"]
+        if tag not in [None, "factor", "model"]:
+            raise ValueError(f"tag must be None, 'factor', or 'model', got {tag!r}")
         quant_background = "The background of the scenario is as follows:\n" + T(".prompts:qlib_quant_background").r(
             runtime_environment=self.get_runtime_environment(),
         )
@@ -83,7 +84,8 @@ class QlibQuantScenario(Scenario):
         return self._source_data
 
     def output_format(self, tag=None) -> str:
-        assert tag in [None, "factor", "model"]
+        if tag not in [None, "factor", "model"]:
+            raise ValueError(f"tag must be None, 'factor', or 'model', got {tag!r}")
         factor_output_format = (
             "The factor code should output the following format:\n" + T(".prompts:qlib_factor_output_format").r()
         )
@@ -99,7 +101,8 @@ class QlibQuantScenario(Scenario):
             return model_output_format
 
     def interface(self, tag=None) -> str:
-        assert tag in [None, "factor", "model"]
+        if tag not in [None, "factor", "model"]:
+            raise ValueError(f"tag must be None, 'factor', or 'model', got {tag!r}")
         factor_interface = (
             "The factor code should be written in the following interface:\n" + T(".prompts:qlib_factor_interface").r()
         )
@@ -115,7 +118,8 @@ class QlibQuantScenario(Scenario):
             return model_interface
 
     def simulator(self, tag=None) -> str:
-        assert tag in [None, "factor", "model"]
+        if tag not in [None, "factor", "model"]:
+            raise ValueError(f"tag must be None, 'factor', or 'model', got {tag!r}")
         factor_simulator = "The factor code will be sent to the simulator:\n" + T(".prompts:qlib_factor_simulator").r()
         model_simulator = "The model code will be sent to the simulator:\n" + T(".prompts:qlib_model_simulator").r()
 
@@ -185,7 +189,8 @@ class QlibQuantScenario(Scenario):
             return common_description(action) + interface(action) + output(action) + simulator(action)
 
     def get_runtime_environment(self, tag: str = None) -> str:
-        assert tag in [None, "factor", "model"]
+        if tag not in [None, "factor", "model"]:
+            raise ValueError(f"tag must be None, 'factor', or 'model', got {tag!r}")
 
         if tag is None or tag == "factor":
             # Use factor env to get the runtime environment

rdagent/scenarios/qlib/proposal/quant_proposal.py

@@ -1,8 +1,7 @@
-import logging
 import json
+import logging
 import os
 import random
-from typing import Tuple
 
 from rdagent.app.qlib_rd_loop.conf import QUANT_PROP_SETTING
 from rdagent.components.proposal import FactorAndModelHypothesisGen
@@ -42,7 +41,7 @@ class QlibQuantHypothesis(Hypothesis):
         action: str,
     ) -> None:
         super().__init__(
-            hypothesis, reason, concise_reason, concise_observation, concise_justification, concise_knowledge
+            hypothesis, reason, concise_reason, concise_observation, concise_justification, concise_knowledge,
         )
         self.action = action
 
@@ -54,10 +53,10 @@ Reason: {self.reason}
 
 
 class QlibQuantHypothesisGen(FactorAndModelHypothesisGen):
-    def __init__(self, scen: Scenario) -> Tuple[dict, bool]:
+    def __init__(self, scen: Scenario) -> None:
         super().__init__(scen)
 
-    def prepare_context(self, trace: Trace) -> Tuple[dict, bool]:
+    def prepare_context(self, trace: Trace) -> tuple[dict, bool]:
 
         # ========= Bandit ==========
         if QUANT_PROP_SETTING.action_selection == "bandit":
@@ -85,7 +84,7 @@ class QlibQuantHypothesisGen(FactorAndModelHypothesisGen):
 
             last_hypothesis_and_feedback = (
                 T("scenarios.qlib.prompts:last_hypothesis_and_feedback").r(
-                    experiment=trace.hist[-1][0], feedback=trace.hist[-1][1]
+                    experiment=trace.hist[-1][0], feedback=trace.hist[-1][1],
                 )
                 if len(trace.hist) > 0
                 else "No previous hypothesis and feedback available since it's the first round."
@@ -195,7 +194,7 @@ class QlibQuantHypothesisGen(FactorAndModelHypothesisGen):
         for i in range(len(trace.hist) - 1, -1, -1):
             if trace.hist[i][0].hypothesis.action == action:
                 last_hypothesis_and_feedback = T("scenarios.qlib.prompts:last_hypothesis_and_feedback").r(
-                    experiment=trace.hist[i][0], feedback=trace.hist[i][1]
+                    experiment=trace.hist[i][0], feedback=trace.hist[i][1],
                 )
                 break
 
@@ -204,7 +203,7 @@ class QlibQuantHypothesisGen(FactorAndModelHypothesisGen):
             for i in range(len(trace.hist) - 1, -1, -1):
                 if trace.hist[i][0].hypothesis.action == "model" and trace.hist[i][1].decision is True:
                     sota_hypothesis_and_feedback = T("scenarios.qlib.prompts:sota_hypothesis_and_feedback").r(
-                        experiment=trace.hist[i][0], feedback=trace.hist[i][1]
+                        experiment=trace.hist[i][0], feedback=trace.hist[i][1],
                     )
                     break
 

rdagent/utils/env.py

@@ -436,13 +436,10 @@ class Env(Generic[ASpecificEnvConf]):
         else:
             timeout_cmd = f"timeout --kill-after=10 {self.conf.running_timeout_period} {entry}"
         entry_add_timeout = (
-            f"/bin/sh -c '"  # start of the sh command
-            + f"{timeout_cmd}; entry_exit_code=$?; "
+            "/bin/sh -c '"  # start of the sh command
+            + timeout_cmd.replace("'", "'\\''") + "; entry_exit_code=$?; "
             + (
                 f"{_get_chmod_cmd(self.conf.mount_path)}; "
-                # We don't have to change the permission of the cache and input folder to remove it
-                # + f"if [ -d {self.conf.mount_path}/cache ]; then chmod 777 {self.conf.mount_path}/cache; fi; " +
-                #     f"if [ -d {self.conf.mount_path}/input ]; then chmod 777 {self.conf.mount_path}/input; fi; "
                 if isinstance(self.conf, DockerConf)
                 else ""
             )
@@ -620,7 +617,8 @@ class LocalEnv(Env[ASpecificLocalConf]):
         for lp, rp in running_extra_volume.items():
             volumes[lp] = rp
 
-        assert local_path is not None, "local_path should not be None"
+        if local_path is None:
+            raise ValueError("local_path should not be None")
         volumes = normalize_volumes(volumes, local_path)
 
         @contextlib.contextmanager
@@ -925,7 +923,11 @@ def _prepare_conda_env(env_name: str, requirements_file: Path, python_version: s
     """
     # 1. Create conda environment if not exists
     env_list = subprocess.run(["conda", "env", "list"], capture_output=True, text=True, check=False)
-    env_exists = any(line.split()[0] == env_name for line in env_list.stdout.splitlines() if line and not line.startswith("#"))
+    env_exists = any(
+        line.split()[0] == env_name
+        for line in env_list.stdout.splitlines()
+        if line and not line.startswith("#") and len(line.split()) > 0
+    )
     if not env_exists:
         print(f"[yellow]Creating conda env '{env_name}' (Python {python_version})...[/yellow]")
         subprocess.check_call(["conda", "create", "-y", "-n", env_name, f"python={python_version}"])
@@ -1191,7 +1193,7 @@ class DockerEnv(Env[DockerConf]):
             with Progress(SpinnerColumn(), TextColumn("{task.description}")) as p:
                 task = p.add_task("[cyan]Building image...")
                 for part in resp_stream:
-                    lines = part.decode("utf-8").split("\r\n")
+                    lines = part.decode("utf-8", errors="replace").split("\r\n")
                     for line in lines:
                         if line.strip():
                             status_dict = json.loads(line)
@@ -1472,7 +1474,8 @@ class DockerEnv(Env[DockerConf]):
                 cpu_count=self.conf.cpu_count,  # Set CPU limit
                 **self._gpu_kwargs(client),
             )
-            assert container is not None  # Ensure container was created successfully
+            if container is None:
+                raise AssertionError("Docker container was not created successfully")
             logs = container.logs(stream=True)
             print(Rule("[bold green]Docker Logs Begin[/bold green]", style="dark_orange"))
             table = Table(title="Run Info", show_header=False)
@@ -1522,8 +1525,8 @@ class DockerEnv(Env[DockerConf]):
 class QTDockerEnv(DockerEnv):
     """Qlib Torch Docker"""
 
-    def __init__(self, conf: DockerConf = QlibDockerConf()):
-        super().__init__(conf)
+    def __init__(self, conf: DockerConf | None = None):
+        super().__init__(conf if conf is not None else QlibDockerConf())
 
     def prepare(self, *args, **kwargs) -> None:  # type: ignore[no-untyped-def]
         """
@@ -1542,15 +1545,15 @@ class QTDockerEnv(DockerEnv):
 class KGDockerEnv(DockerEnv):
     """Kaggle Competition Docker"""
 
-    def __init__(self, competition: str | None = None, conf: DockerConf = KGDockerConf()):
-        super().__init__(conf)
+    def __init__(self, competition: str | None = None, conf: DockerConf | None = None):
+        super().__init__(conf if conf is not None else KGDockerConf())
 
 
 class MLEBDockerEnv(DockerEnv):
     """MLEBench Docker"""
 
-    def __init__(self, conf: DockerConf = MLEBDockerConf()):
-        super().__init__(conf)
+    def __init__(self, conf: DockerConf | None = None):
+        super().__init__(conf if conf is not None else MLEBDockerConf())
 
 
 class FTDockerEnv(DockerEnv):
@@ -1566,8 +1569,8 @@ class FTDockerEnv(DockerEnv):
         export FT_DOCKER_save_logs_to_file=false # disable log file
     """
 
-    def __init__(self, conf: DockerConf = FTDockerConf()):
-        super().__init__(conf)
+    def __init__(self, conf: DockerConf | None = None):
+        super().__init__(conf if conf is not None else FTDockerConf())
 
 
 class BenchmarkDockerEnv(DockerEnv):
@@ -1584,5 +1587,5 @@ class BenchmarkDockerEnv(DockerEnv):
         export BENCHMARK_DOCKER_terminal_tail_lines=100  # show last 100 lines
     """
 
-    def __init__(self, conf: DockerConf = BenchmarkDockerConf()):
-        super().__init__(conf)
+    def __init__(self, conf: DockerConf | None = None):
+        super().__init__(conf if conf is not None else BenchmarkDockerConf())

rdagent/utils/workflow/loop.py

@@ -15,19 +15,19 @@ import multiprocessing.queues
 import os
 import pickle
 from collections import defaultdict
+from collections.abc import Callable
 from dataclasses import dataclass
 from datetime import datetime, timezone
 from pathlib import Path
-from typing import Any, Callable, Optional, Union, cast
+from typing import Any, cast
 
 import psutil
-from tqdm.auto import tqdm
-
 from rdagent.core.conf import RD_AGENT_SETTINGS
 from rdagent.log import rdagent_logger as logger
 from rdagent.log.conf import LOG_SETTINGS
 from rdagent.log.timer import RD_Agent_TIMER_wrapper, RDAgentTimer
 from rdagent.utils.workflow.tracking import WorkflowTracker
+from tqdm.auto import tqdm
 
 
 class LoopMeta(type):
@@ -98,7 +98,7 @@ class LoopBase:
     skip_loop_error: tuple[type[BaseException], ...] = ()  # you can define a list of error that will skip current loop
     skip_loop_error_stepname: str | None = None  # if skip_loop_error exception happens, what's the next step to work on
     withdraw_loop_error: tuple[
-        type[BaseException], ...
+        type[BaseException], ...,
     ] = ()  # you can define a list of error that will withdraw current loop
 
     EXCEPTION_KEY = "_EXCEPTION"
@@ -129,8 +129,8 @@ class LoopBase:
         self.tracker = WorkflowTracker(self)  # Initialize tracker with this LoopBase instance
 
         # progress control
-        self.loop_n: Optional[int] = None  # remain loop count
-        self.step_n: Optional[int] = None  # remain step count
+        self.loop_n: int | None = None  # remain loop count
+        self.step_n: int | None = None  # remain step count
 
         self.semaphores: dict[str, asyncio.Semaphore] = {}
 
@@ -169,7 +169,7 @@ class LoopBase:
             self._pbar.close()
             del self._pbar
 
-    def _check_exit_conditions_on_step(self, loop_id: Optional[int] = None, step_id: Optional[int] = None) -> None:
+    def _check_exit_conditions_on_step(self, loop_id: int | None = None, step_id: int | None = None) -> None:
         """Check if the loop should continue or terminate.
 
         Raises
@@ -188,8 +188,7 @@ class LoopBase:
             if self.timer.is_timeout():
                 logger.warning("Timeout, exiting the loop.")
                 raise self.LoopTerminationError("Timer timeout")
-            else:
-                logger.info(f"Timer remaining time: {self.timer.remain_time()}")
+            logger.info(f"Timer remaining time: {self.timer.remain_time()}")
 
     async def _run_step(self, li: int, force_subproc: bool = False) -> None:
         """Execute a single step (next unrun step) in the workflow (async version with force_subproc option).
@@ -217,7 +216,7 @@ class LoopBase:
 
             with logger.tag(f"Loop_{li}.{name}"):
                 start = datetime.now(timezone.utc)
-                func: Callable[..., Any] = cast(Callable[..., Any], getattr(self, name))
+                func: Callable[..., Any] = cast("Callable[..., Any]", getattr(self, name))
 
                 next_step_idx = si + 1
                 step_forward = True
@@ -233,15 +232,14 @@ class LoopBase:
                             # Using deepcopy is to avoid triggering errors like "RuntimeError: dictionary changed size during iteration"
                             # GUESS: Some content in self.loop_prev_out[li] may be in the middle of being changed.
                             result = await curr_loop.run_in_executor(
-                                pool, copy.deepcopy(func), copy.deepcopy(self.loop_prev_out[li])
+                                pool, copy.deepcopy(func), copy.deepcopy(self.loop_prev_out[li]),
                             )
+                    # auto determine whether to run async or sync
+                    elif asyncio.iscoroutinefunction(func):
+                        result = await func(self.loop_prev_out[li])
                     else:
-                        # auto determine whether to run async or sync
-                        if asyncio.iscoroutinefunction(func):
-                            result = await func(self.loop_prev_out[li])
-                        else:
-                            # Default: run sync function directly
-                            result = func(self.loop_prev_out[li])
+                        # Default: run sync function directly
+                        result = func(self.loop_prev_out[li])
                     # Store result in the nested dictionary
                     self.loop_prev_out[li][name] = result
                 except Exception as e:
@@ -251,14 +249,13 @@ class LoopBase:
                             next_step_idx = self.steps.index(self.skip_loop_error_stepname)
                             if next_step_idx <= si:
                                 raise RuntimeError(
-                                    f"Cannot skip backwards or to same step. Current: {si} ({name}), Target: {next_step_idx} ({self.skip_loop_error_stepname})"
+                                    f"Cannot skip backwards or to same step. Current: {si} ({name}), Target: {next_step_idx} ({self.skip_loop_error_stepname})",
                                 ) from e
+                        # Default: jump to feedback step if exists, otherwise jump to the last step (record)
+                        elif "feedback" in self.steps:
+                            next_step_idx = self.steps.index("feedback")
                         else:
-                            # Default: jump to feedback step if exists, otherwise jump to the last step (record)
-                            if "feedback" in self.steps:
-                                next_step_idx = self.steps.index("feedback")
-                            else:
-                                next_step_idx = len(self.steps) - 1
+                            next_step_idx = len(self.steps) - 1
                         self.loop_prev_out[li][name] = None
                         self.loop_prev_out[li][self.EXCEPTION_KEY] = e
                     elif isinstance(e, self.withdraw_loop_error):
@@ -409,6 +406,8 @@ class LoopBase:
                 self.close_pbar()
 
     def withdraw_loop(self, loop_idx: int) -> None:
+        if loop_idx <= 0:
+            raise RuntimeError(f"Cannot withdraw loop {loop_idx}: no previous loop exists.")
         prev_session_dir = self.session_folder / str(loop_idx - 1)
         prev_path = min(
             (p for p in prev_session_dir.glob("*_*") if p.is_file()),
@@ -501,7 +500,7 @@ class LoopBase:
             session_folder = path.parent.parent
 
         with path.open("rb") as f:
-            session = cast(LoopBase, pickle.load(f))
+            session = cast("LoopBase", pickle.load(f))
 
         # set session folder
         if checkout:

rdagent/utils/workflow/tracking.py

@@ -9,7 +9,6 @@ import datetime
 from typing import TYPE_CHECKING
 
 import pytz
-
 from rdagent.core.conf import RD_AGENT_SETTINGS
 from rdagent.log.timer import RD_Agent_TIMER_wrapper
 
@@ -84,12 +83,14 @@ class WorkflowTracker:
             # Log timer status if timer is started
             if self.loop_base.timer.started:
                 remain_time = self.loop_base.timer.remain_time()
-                assert remain_time is not None
-                mlflow.log_metric("remain_time", remain_time.total_seconds())
-                mlflow.log_metric(
-                    "remain_percent",
-                    remain_time / self.loop_base.timer.all_duration * 100,
-                )
+                if remain_time is None:
+                    logger.warning("remain_time is None despite timer.started, skipping timer metrics")
+                else:
+                    mlflow.log_metric("remain_time", remain_time.total_seconds())
+                    mlflow.log_metric(
+                        "remain_percent",
+                        remain_time / self.loop_base.timer.all_duration * 100,
+                    )
 
         # Keep only the log_workflow_state method as it's the primary entry point now
         except Exception as e:

scripts/predix_parallel.py

@@ -19,19 +19,16 @@ import sys
 import time
 from datetime import datetime
 from pathlib import Path
-from typing import Dict, List, Optional
 
 from dotenv import load_dotenv
 from rich.console import Console
 from rich.live import Live
+from rich.markdown import Markdown
 from rich.panel import Panel
 from rich.table import Table
-from rich.markdown import Markdown
-from rich.layout import Layout
-from rich.progress import Progress, SpinnerColumn, TextColumn, BarColumn, TaskProgressColumn
 
 # Load environment variables from .env file
-load_dotenv(Path(__file__).parent / ".env")
+load_dotenv(Path(__file__).parent.parent / ".env")
 
 console = Console()
 
@@ -43,12 +40,12 @@ class RunState:
         self.run_id = run_id
         self.api_key_idx = api_key_idx
         self.model = model
-        self.process: Optional[subprocess.Popen] = None
+        self.process: subprocess.Popen | None = None
         self.status: str = "pending"  # pending, running, success, failed, stopped
-        self.start_time: Optional[datetime] = None
-        self.end_time: Optional[datetime] = None
-        self.exit_code: Optional[int] = None
-        self.error_message: Optional[str] = None
+        self.start_time: datetime | None = None
+        self.end_time: datetime | None = None
+        self.exit_code: int | None = None
+        self.error_message: str | None = None
         self.log_file: str = f"fin_quant_run{run_id}.log"
 
     @property
@@ -105,8 +102,8 @@ class ParallelRunner:
         self.num_runs = num_runs
         self.num_api_keys = num_api_keys
         self.model = model
-        self.runs: List[RunState] = []
-        self.project_root = Path(__file__).parent
+        self.runs: list[RunState] = []
+        self.project_root = Path(__file__).parent.parent
         self._shutdown_requested = False
 
         # Read API keys from environment
@@ -115,10 +112,10 @@ class ParallelRunner:
         # Validate we have enough API keys
         if self.model == "openrouter" and len(self.api_keys) < num_api_keys:
             console.print(
-                f"[yellow]⚠️  Requested {num_api_keys} API keys, but only {len(self.api_keys)} found in .env[/yellow]"
+                f"[yellow]⚠️  Requested {num_api_keys} API keys, but only {len(self.api_keys)} found in .env[/yellow]",
             )
             console.print(
-                f"[dim]Distributing across {len(self.api_keys)} available key(s)[/dim]"
+                f"[dim]Distributing across {len(self.api_keys)} available key(s)[/dim]",
             )
             self.num_api_keys = len(self.api_keys)
 
@@ -129,7 +126,7 @@ class ParallelRunner:
             run_state = RunState(run_id=i, api_key_idx=api_key_idx, model=model)
             self.runs.append(run_state)
 
-    def _load_api_keys(self) -> List[str]:
+    def _load_api_keys(self) -> list[str]:
         """Load API keys from environment variables."""
         keys = []
 
@@ -149,7 +146,7 @@ class ParallelRunner:
 
         return keys
 
-    def _build_env(self, run_state: RunState) -> Dict[str, str]:
+    def _build_env(self, run_state: RunState) -> dict[str, str]:
         """
         Build isolated environment for a subprocess.
 
@@ -174,16 +171,14 @@ class ParallelRunner:
         env["RD_AGENT_WORKSPACE"] = str(workspace_dir)
 
         # Configure API key for this run
-        if self.model == "openrouter" and run_state.api_key_idx < len(self.api_keys):
-            api_key = self.api_keys[run_state.api_key_idx]
-            env["OPENAI_API_KEY"] = api_key
-            env["OPENAI_API_BASE"] = "https://openrouter.ai/api/v1"
-            env["CHAT_MODEL"] = os.getenv("OPENROUTER_MODEL", "openrouter/google/gemma-4-26b-a4b-it:free")
-
-            # If we configured multiple API keys AND have enough keys, use load balancing
+        if self.model == "openrouter":
             if self.num_api_keys >= 2 and len(self.api_keys) >= 2:
                 env["OPENAI_API_KEY"] = f"{self.api_keys[0]},{self.api_keys[1]}"
                 env["LITELLM_PARALLEL_CALLS"] = "2"
+            elif run_state.api_key_idx < len(self.api_keys):
+                env["OPENAI_API_KEY"] = self.api_keys[run_state.api_key_idx]
+            env["OPENAI_API_BASE"] = "https://openrouter.ai/api/v1"
+            env["CHAT_MODEL"] = os.getenv("OPENROUTER_MODEL", "openrouter/google/gemma-4-26b-a4b-it:free")
         elif self.model == "local":
             env["OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY", "local")
             env["OPENAI_API_BASE"] = os.getenv("OPENAI_API_BASE", "http://localhost:8081/v1")
@@ -191,7 +186,7 @@ class ParallelRunner:
 
         return env
 
-    def _build_command(self, run_state: RunState) -> List[str]:
+    def _build_command(self, run_state: RunState) -> list[str]:
         """
         Build the subprocess command to run predix quant.
 
@@ -235,20 +230,24 @@ class ParallelRunner:
         log_path = self.project_root / run_state.log_file
         log_f = open(log_path, "a", encoding="utf-8")
 
-        # Start subprocess
-        run_state.process = subprocess.Popen(
-            cmd,
-            env=env,
-            cwd=str(self.project_root),
-            stdout=log_f,
-            stderr=subprocess.STDOUT,
-        )
-        run_state.status = "running"
-        run_state.start_time = datetime.now()
+        try:
+            # Start subprocess
+            run_state.process = subprocess.Popen(
+                cmd,
+                env=env,
+                cwd=str(self.project_root),
+                stdout=log_f,
+                stderr=subprocess.STDOUT,
+            )
+            run_state.status = "running"
+            run_state.start_time = datetime.now()
+        except Exception:
+            log_f.close()
+            raise
 
         console.print(
             f"[dim]  ▶️  Run {run_state.run_id} started (PID: {run_state.process.pid}, "
-            f"API Key: {run_state.api_key_idx + 1}, Model: {run_state.model})[/dim]"
+            f"API Key: {run_state.api_key_idx + 1}, Model: {run_state.model})[/dim]",
         )
 
     def _check_run(self, run_state: RunState) -> None:
@@ -273,14 +272,14 @@ class ParallelRunner:
                 run_state.status = "success"
                 console.print(
                     f"[bold green]  ✅ Run {run_state.run_id} completed "
-                    f"({run_state.elapsed})[/bold green]"
+                    f"({run_state.elapsed})[/bold green]",
                 )
             else:
                 run_state.status = "failed"
                 run_state.error_message = f"Exit code: {poll_result}"
                 console.print(
                     f"[bold red]  ❌ Run {run_state.run_id} failed "
-                    f"({run_state.elapsed}, exit code: {poll_result})[/bold red]"
+                    f"({run_state.elapsed}, exit code: {poll_result})[/bold red]",
                 )
 
     def _stop_run(self, run_state: RunState) -> None:
@@ -386,7 +385,7 @@ class ParallelRunner:
             if run.status == "running":
                 self._stop_run(run)
 
-    def run(self) -> Dict[str, int]:
+    def run(self) -> dict[str, int]:
         """
         Execute all parallel runs and show live dashboard.
 
@@ -400,7 +399,7 @@ class ParallelRunner:
         signal.signal(signal.SIGTERM, self._signal_handler)
 
         console.print(f"\n[bold cyan]{'=' * 60}[/bold cyan]")
-        console.print(f"[bold cyan]🔀 Predix Parallel Runner[/bold cyan]")
+        console.print("[bold cyan]🔀 Predix Parallel Runner[/bold cyan]")
         console.print(f"[bold cyan]{'=' * 60}[/bold cyan]")
         console.print(f"  Runs: {self.num_runs}")
         console.print(f"  API Keys: {self.num_api_keys} ({len(self.api_keys)} available)")
@@ -451,7 +450,7 @@ class ParallelRunner:
         stopped_count = sum(1 for r in self.runs if r.status == "stopped")
 
         console.print(f"\n[bold cyan]{'=' * 60}[/bold cyan]")
-        console.print(f"[bold cyan]📊 Parallel Run Summary[/bold cyan]")
+        console.print("[bold cyan]📊 Parallel Run Summary[/bold cyan]")
         console.print(f"[bold cyan]{'=' * 60}[/bold cyan]")
         console.print(f"  ✅ Success: {success_count}/{self.num_runs}")
         console.print(f"  ❌ Failed: {failed_count}/{self.num_runs}")
@@ -463,7 +462,7 @@ class ParallelRunner:
             if run.start_time and run.end_time:
                 delta = run.end_time - run.start_time
                 console.print(
-                    f"    Run #{run.run_id}: {run.status} ({delta.total_seconds():.0f}s)"
+                    f"    Run #{run.run_id}: {run.status} ({delta.total_seconds():.0f}s)",
                 )
 
         return {
@@ -478,7 +477,7 @@ def main(
     runs: int = 5,
     api_keys: int = 2,
     model: str = "openrouter",
-) -> Dict[str, int]:
+) -> dict[str, int]:
     """
     Run multiple factor experiments in parallel.
 
@@ -504,7 +503,7 @@ if __name__ == "__main__":
     import argparse
 
     parser = argparse.ArgumentParser(
-        description="Predix Parallel Runner - Run multiple factor experiments concurrently"
+        description="Predix Parallel Runner - Run multiple factor experiments concurrently",
     )
     parser.add_argument(
         "--runs", "-n",
@@ -542,7 +541,7 @@ if __name__ == "__main__":
     elif args.runs > 25:
         console.print(f"\n[yellow]⚠️  {args.runs} runs - high resource usage expected[/yellow]")
         console.print(f"   Estimated RAM: ~{args.runs * 0.65:.0f} GB")
-        console.print(f"   Use --force to confirm.\n")
+        console.print("   Use --force to confirm.\n")
         import time
         time.sleep(2)
 

scripts/predix_rebacktest_strategies.py

@@ -16,7 +16,8 @@ def load_factors(names, vdir):
                     if df is not None and len(df.columns) > 0:
                         dfs[n] = df.iloc[:, 0]
                         break
-                except: pass
+                except Exception:
+                    pass
     return dfs
 
 def fix_code(code, available):
@@ -79,7 +80,7 @@ except Exception as e:
             if r.returncode != 0:
                 return None
             sig = pd.read_pickle(str(tdp / "s.pkl"))
-        except:
+        except Exception:
             return None
     
     fwd = df.mean(axis=1).shift(-96).dropna()
@@ -124,7 +125,7 @@ def main(count=None):
             d = json.load(open(f))
             if isinstance(d, dict) and 'strategy_name' in d:
                 files.append(f)
-        except: pass
+        except Exception: pass
     if count: files = files[:count]
     
     print(f"Re-evaluating {len(files)} strategies...\n")
@@ -147,7 +148,7 @@ def main(count=None):
                         with open(f, 'w') as out: json.dump(data, out, indent=2, ensure_ascii=False)
                         updated += 1
                         results.append({'name':data['strategy_name'], **bt})
-            except:
+            except Exception:
                 pass
             p.update(task, advance=1)
     

test/backtesting/test_ftmo_oos.py

@@ -16,7 +16,9 @@ import pytest
 
 from rdagent.components.backtesting.vbt_backtest import (
     OOS_START_DEFAULT,
+    _apply_ftmo_mask,
     backtest_signal_ftmo,
+    FTMO_INITIAL_CAPITAL,
     FTMO_MAX_DAILY_LOSS,
     FTMO_MAX_TOTAL_LOSS,
     monte_carlo_trade_pvalue,
@@ -238,3 +240,141 @@ def test_wf_consistency_range(close_6yr):
     c = r.get("wf_oos_consistency")
     if c is not None:
         assert 0.0 <= c <= 1.0
+
+
+# ---------------------------------------------------------------------------
+# Direct _apply_ftmo_mask unit tests
+# ---------------------------------------------------------------------------
+
+class TestApplyFtmoMask:
+    """Direct unit tests for _apply_ftmo_mask — the core FTMO daily/total loss engine."""
+
+    @pytest.fixture
+    def flat_close(self) -> pd.Series:
+        n = 3000
+        idx = pd.date_range("2024-01-01", periods=n, freq="1min")
+        return pd.Series(1.10, index=idx)
+
+    def test_returns_compliance_dict(self, flat_close):
+        signal = _random_signal(flat_close.index)
+        masked, info = _apply_ftmo_mask(signal, flat_close, leverage=1.0, txn_cost_bps=2.14)
+        assert "ftmo_daily_breaches" in info
+        assert "ftmo_total_breached" in info
+        assert "ftmo_total_breach_ts" in info
+        assert "ftmo_compliant" in info
+
+    def test_flat_market_zero_signal_fully_compliant(self, flat_close):
+        """No trades → always compliant."""
+        signal = pd.Series(0.0, index=flat_close.index)
+        masked, info = _apply_ftmo_mask(signal, flat_close, leverage=1.0, txn_cost_bps=2.14)
+        assert info["ftmo_daily_breaches"] == 0
+        assert info["ftmo_total_breached"] is False
+        assert info["ftmo_compliant"] is True
+        # All signals should remain zero
+        assert (masked == 0).all()
+
+    def test_daily_loss_breach_zeroes_rest_of_day(self):
+        """When daily loss exceeds 5%, rest of that day's signals are zeroed."""
+        n = 3000
+        idx = pd.date_range("2024-01-01", periods=n, freq="1min")
+        # Price drops sharply in first few bars to trigger daily loss
+        price = pd.Series(1.10, index=idx, dtype=float)
+        price.iloc[3:20] = 0.00  # crash from 1.10 to 0.00 → massive loss
+        signal = pd.Series(1.0, index=idx)  # always long at 30x leverage
+
+        masked, info = _apply_ftmo_mask(signal, price, leverage=30.0, txn_cost_bps=0)
+        assert info["ftmo_daily_breaches"] > 0
+        # After breach, signals on same day must be zeroed
+        breach_day = idx[0].date()
+        same_day_late = (idx[-1] if idx[-1].date() == breach_day else idx[20])
+        if same_day_late.date() == breach_day:
+            assert masked.loc[same_day_late] == 0
+
+    def test_total_loss_breach_zeroes_all_remaining(self):
+        """When total loss exceeds 10%, ALL subsequent signals are zeroed."""
+        n = 5000
+        idx = pd.date_range("2024-01-01", periods=n, freq="1min")
+        # Price crashes → max position → total loss limit breached
+        price = pd.Series(1.10, index=idx, dtype=float)
+        price.iloc[5:50] = 0.50  # >10% drop with 30x leverage
+        signal = pd.Series(1.0, index=idx)
+
+        masked, info = _apply_ftmo_mask(signal, price, leverage=30.0, txn_cost_bps=0)
+        assert info["ftmo_total_breached"] is True
+        assert info["ftmo_total_breach_ts"] is not None
+        # After breach, ALL later signals must be zero
+        assert (masked.iloc[100:] == 0).all()
+
+    def test_total_breach_respected_across_days(self):
+        """Total breach persists across day boundaries — no new trades after breach."""
+        n = 5000
+        idx = pd.date_range("2024-01-01", periods=n, freq="1min")
+        price = pd.Series(1.10, index=idx, dtype=float)
+        price.iloc[5:50] = 0.50
+        signal = pd.Series(1.0, index=idx)
+
+        masked, info = _apply_ftmo_mask(signal, price, leverage=30.0, txn_cost_bps=0)
+        # All signals after breach index must be zero
+        breach_ts = pd.Timestamp(info["ftmo_total_breach_ts"])
+        assert (masked.loc[masked.index > breach_ts] == 0).all()
+
+    def test_daily_loss_resets_on_new_day(self):
+        """Daily loss limit resets at day boundary — new day starts fresh (unless total breached)."""
+        n = 5000
+        idx = pd.date_range("2024-01-01", periods=n, freq="1min")
+        price = pd.Series(1.10, index=idx, dtype=float)
+        # Trigger daily breach on day 1 by dropping 1%
+        price.iloc[5:20] = 1.09  # ~1% drop with 30x → ~30% loss
+        signal = pd.Series(1.0, index=idx)
+
+        masked, info = _apply_ftmo_mask(signal, price, leverage=30.0, txn_cost_bps=0)
+        assert info["ftmo_daily_breaches"] >= 1
+        # Day 2 signals should be active again if not total-breached
+        day2_mask = idx.date > idx[0].date()
+        if day2_mask.any() and not info["ftmo_total_breached"]:
+            day2 = idx[day2_mask][0]
+            assert masked.loc[day2] != 0
+
+    def test_compliant_flag_false_after_daily_breach(self):
+        """Even one daily breach makes ftmo_compliant=False."""
+        n = 3000
+        idx = pd.date_range("2024-01-01", periods=n, freq="1min")
+        price = pd.Series(1.10, index=idx, dtype=float)
+        price.iloc[3:20] = 0.00
+        signal = pd.Series(1.0, index=idx)
+
+        masked, info = _apply_ftmo_mask(signal, price, leverage=30.0, txn_cost_bps=0)
+        assert info["ftmo_compliant"] is False
+
+    def test_compliant_flag_false_after_total_breach(self):
+        """Total breach makes ftmo_compliant=False."""
+        n = 5000
+        idx = pd.date_range("2024-01-01", periods=n, freq="1min")
+        price = pd.Series(1.10, index=idx, dtype=float)
+        price.iloc[5:50] = 0.50
+        signal = pd.Series(1.0, index=idx)
+
+        masked, info = _apply_ftmo_mask(signal, price, leverage=30.0, txn_cost_bps=0)
+        assert info["ftmo_compliant"] is False
+
+    def test_transaction_costs_reduce_equity(self):
+        """Transaction costs should reduce equity — compliant scenario with fees."""
+        n = 1000
+        idx = pd.date_range("2024-01-01", periods=n, freq="1min")
+        price = pd.Series(1.10, index=idx, dtype=float)
+        # Alternating signal → lots of position changes → high costs
+        signal = pd.Series([1.0 if i % 2 == 0 else -1.0 for i in range(n)], index=idx)
+
+        masked, info = _apply_ftmo_mask(signal, price, leverage=1.0, txn_cost_bps=10.0)
+        # With high costs and flat market, equity should drop
+        assert "ftmo_daily_breaches" in info
+
+    def test_output_mask_has_same_index(self):
+        n = 2000
+        idx = pd.date_range("2024-01-01", periods=n, freq="1min")
+        price = pd.Series(1.10, index=idx)
+        signal = _random_signal(idx, seed=1)
+
+        masked, info = _apply_ftmo_mask(signal, price, leverage=1.0, txn_cost_bps=2.14)
+        assert len(masked) == len(signal)
+        assert masked.index.equals(signal.index)

test/backtesting/test_results_db.py

@@ -399,3 +399,91 @@ class TestDatabaseIntegrity:
 
 # Import am Anfang der Datei für die Tests
 from rdagent.components.backtesting.results_db import ResultsDatabase
+
+
+class TestAddColumnIfNotExists:
+    """Direct tests for _add_column_if_not_exists migration helper."""
+
+    def test_add_new_column_succeeds(self):
+        """Adding a new column to an existing table should work."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            import os
+            db_path = os.path.join(tmpdir, "test.db")
+            db = ResultsDatabase(db_path=db_path)
+            try:
+                db._add_column_if_not_exists("backtest_runs", "test_new_col", "REAL")
+                c = db.conn.cursor()
+                c.execute("PRAGMA table_info(backtest_runs)")
+                cols = [row[1] for row in c.fetchall()]
+                assert "test_new_col" in cols
+            finally:
+                db.close()
+
+    def test_existing_column_noop(self):
+        """Adding an already existing column should succeed (no-op)."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            import os
+            db_path = os.path.join(tmpdir, "test.db")
+            db = ResultsDatabase(db_path=db_path)
+            try:
+                # First call adds, second call should be no-op
+                db._add_column_if_not_exists("backtest_runs", "ic", "REAL")
+                db._add_column_if_not_exists("backtest_runs", "ic", "REAL")
+                c = db.conn.cursor()
+                c.execute("PRAGMA table_info(backtest_runs)")
+                cols = [row[1] for row in c.fetchall()]
+                assert cols.count("ic") == 1  # should exist exactly once
+            finally:
+                db.close()
+
+    def test_invalid_table_raises(self):
+        with tempfile.TemporaryDirectory() as tmpdir:
+            import os
+            db_path = os.path.join(tmpdir, "test.db")
+            db = ResultsDatabase(db_path=db_path)
+            try:
+                with pytest.raises(ValueError, match="Unknown table"):
+                    db._add_column_if_not_exists("nonexistent_table", "col", "REAL")
+            finally:
+                db.close()
+
+    def test_invalid_column_name_raises(self):
+        with tempfile.TemporaryDirectory() as tmpdir:
+            import os
+            db_path = os.path.join(tmpdir, "test.db")
+            db = ResultsDatabase(db_path=db_path)
+            try:
+                with pytest.raises(ValueError, match="Invalid column name"):
+                    db._add_column_if_not_exists("backtest_runs", "bad;column", "REAL")
+            finally:
+                db.close()
+
+    def test_invalid_column_type_raises(self):
+        with tempfile.TemporaryDirectory() as tmpdir:
+            import os
+            db_path = os.path.join(tmpdir, "test.db")
+            db = ResultsDatabase(db_path=db_path)
+            try:
+                with pytest.raises(ValueError, match="Invalid column type"):
+                    db._add_column_if_not_exists("backtest_runs", "col", "INVALID_TYPE")
+            finally:
+                db.close()
+
+    def test_all_allowed_types_work(self):
+        """REAL, TEXT, INTEGER, BLOB should all be valid types."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            import os
+            db_path = os.path.join(tmpdir, "test.db")
+            db = ResultsDatabase(db_path=db_path)
+            try:
+                for col_type in ("REAL", "TEXT", "INTEGER", "BLOB"):
+                    db._add_column_if_not_exists(
+                        "backtest_runs", f"test_{col_type.lower()}", col_type,
+                    )
+                c = db.conn.cursor()
+                c.execute("PRAGMA table_info(backtest_runs)")
+                cols = {row[1] for row in c.fetchall()}
+                for col_type in ("REAL", "TEXT", "INTEGER", "BLOB"):
+                    assert f"test_{col_type.lower()}" in cols
+            finally:
+                db.close()

test/local/test_background_tasks.py

@@ -0,0 +1,277 @@
+"""
+Tests for background task infrastructure (parallel runner, CLI paths, env loading).
+
+Verifies bugs that were previously present:
+- predix_parallel.py: project_root pointing to scripts/ instead of repo root
+- predix_parallel.py: .env loaded from scripts/ instead of repo root
+- predix_parallel.py: API key round-robin overwritten by comma-separated list
+- cli.py: project_root depth wrong (4 .parent hops instead of 3)
+- cli.py start_loop: hardcoded "python" instead of sys.executable
+- cli.py parallel: hardcoded model=local
+"""
+
+import os
+import sys
+from pathlib import Path
+from unittest.mock import Mock, patch
+
+import pytest
+
+
+# ── predix_parallel.py ──────────────────────────────────────────────────
+
+
+class TestParallelRunnerProjectRoot:
+    """Verify ParallelRunner.project_root points to the repo root, not scripts/."""
+
+    def test_project_root_is_repo_root(self):
+        """Bug: project_root was Path(__file__).parent (= scripts/)."""
+        from scripts.predix_parallel import ParallelRunner
+
+        runner = ParallelRunner(num_runs=1, num_api_keys=1, model="local")
+        root = runner.project_root
+
+        # Must contain predix.py (repo root), NOT be the scripts/ dir
+        assert (root / "predix.py").exists(), (
+            f"project_root={root} does not contain predix.py — "
+            f"likely still pointing to scripts/ instead of repo root"
+        )
+        assert root.name != "scripts", (
+            f"project_root={root} ends with 'scripts/' — should be repo root"
+        )
+
+    def test_build_command_points_to_predix_py(self):
+        """Bug: command pointed to scripts/predix.py which doesn't exist."""
+        from scripts.predix_parallel import ParallelRunner, RunState
+
+        runner = ParallelRunner(num_runs=1, num_api_keys=1, model="local")
+        run = RunState(run_id=1, api_key_idx=0, model="local")
+        cmd = runner._build_command(run)
+
+        predix_path = Path(cmd[1])
+        assert predix_path.exists(), (
+            f"Command references {predix_path} which does not exist — "
+            f"project_root likely still wrong"
+        )
+        assert predix_path.name == "predix.py"
+        assert predix_path.parent.name != "scripts", (
+            "predix.py should be in repo root, not scripts/"
+        )
+
+    def test_env_loading_from_repo_root(self):
+        """Bug: load_dotenv loaded scripts/.env which doesn't exist."""
+        # load_dotenv is called at module import time, so we just verify
+        # that after import, the env reflects any .env at repo root.
+        # The key test: the call should not raise FileNotFoundError.
+        repo_root = Path(__file__).parent.parent.parent
+        env_path = repo_root / ".env"
+        assert env_path.exists(), (
+            f".env not found at {env_path} — repo root detection may be wrong"
+        )
+
+
+class TestParallelRunnerAPIKeys:
+    """Verify API key distribution logic."""
+
+    def test_single_api_key_no_overwrite(self):
+        """Bug: with num_api_keys=1, individual key was set then overwritten."""
+        from scripts.predix_parallel import ParallelRunner, RunState
+
+        with patch.dict(os.environ, {}, clear=True):
+            os.environ["OPENROUTER_API_KEY"] = "sk-test-key-1"
+            runner = ParallelRunner(num_runs=2, num_api_keys=1, model="openrouter")
+            # Reset api_keys since _load_api_keys already ran in __init__
+            runner.api_keys = ["sk-test-key-1"]
+            runner.num_api_keys = 1
+
+            env = runner._build_env(RunState(run_id=1, api_key_idx=0, model="openrouter"))
+
+            assert env["OPENAI_API_KEY"] == "sk-test-key-1", (
+                "Single key should be assigned directly, not overwritten"
+            )
+            assert "LITELLM_PARALLEL_CALLS" not in env, (
+                "LITELLM_PARALLEL_CALLS should not be set for single key"
+            )
+
+    def test_multi_api_key_comma_separated(self):
+        """With 2+ keys, all runs get comma-separated list for load balancing."""
+        from scripts.predix_parallel import ParallelRunner, RunState
+
+        with patch.dict(os.environ, {}, clear=True):
+            os.environ["OPENROUTER_API_KEY"] = "sk-key-a"
+            os.environ["OPENROUTER_API_KEY_2"] = "sk-key-b"
+            runner = ParallelRunner(num_runs=3, num_api_keys=2, model="openrouter")
+
+            env = runner._build_env(RunState(run_id=1, api_key_idx=0, model="openrouter"))
+
+            assert env["OPENAI_API_KEY"] == "sk-key-a,sk-key-b", (
+                "Multiple keys should be comma-separated for LiteLLM load balancing"
+            )
+            assert env.get("LITELLM_PARALLEL_CALLS") == "2"
+
+    def test_round_robin_api_key_index(self):
+        """Verify round-robin API key index assignment is computed correctly."""
+        from scripts.predix_parallel import ParallelRunner
+
+        with patch.dict(os.environ, {}, clear=True):
+            os.environ["OPENROUTER_API_KEY"] = "a"
+            os.environ["OPENROUTER_API_KEY_2"] = "b"
+            runner = ParallelRunner(num_runs=5, num_api_keys=2, model="openrouter")
+
+            # 5 runs, 2 keys → indices: 0, 1, 0, 1, 0
+            expected = [0, 1, 0, 1, 0]
+            actual = [r.api_key_idx for r in runner.runs]
+            assert actual == expected, f"Round-robin mismatch: {actual} != {expected}"
+
+
+class TestParallelRunnerLogFileHandling:
+    """Verify log files and results go to the right place."""
+
+    def test_log_file_paths_in_repo_root(self):
+        """Bug: logs went to scripts/fin_quant_runN.log."""
+        from scripts.predix_parallel import ParallelRunner
+
+        runner = ParallelRunner(num_runs=2, num_api_keys=1, model="local")
+
+        for run in runner.runs:
+            log_file = run.log_file
+            # log_file is relative — should be "fin_quant_runN.log"
+            assert "scripts" not in log_file, (
+                f"Log file {log_file} should not be in scripts/"
+            )
+            assert log_file.startswith("fin_quant_run"), (
+                f"Unexpected log file name: {log_file}"
+            )
+
+
+# ── cli.py ──────────────────────────────────────────────────────────────
+
+
+class TestCLIProjectRoot:
+    """Verify CLI commands resolve project_root to the actual repo root."""
+
+    REPO_ROOT = Path(__file__).parent.parent.parent
+
+    def test_cli_project_root_depth(self):
+        """Bug: 4x .parent put project_root one level above the repo."""
+        # The fixed code uses .parent.parent.parent (3 hops) from rdagent/app/cli.py
+        cli_file = self.REPO_ROOT / "rdagent" / "app" / "cli.py"
+        assert cli_file.exists(), f"cli.py not found at {cli_file}"
+
+        # Simulate what the fixed code does
+        resolved = cli_file.parent.parent.parent
+        assert resolved == self.REPO_ROOT, (
+            f"3 .parent hops from cli.py should yield repo root, got {resolved}"
+        )
+
+        # The bug used 4 hops which would overshoot
+        buggy = cli_file.parent.parent.parent.parent
+        assert buggy != self.REPO_ROOT, (
+            "4 .parent hops should NOT yield repo root "
+            f"(got {buggy}, expected {self.REPO_ROOT.parent})"
+        )
+        assert (buggy / "Predix").exists() or buggy == self.REPO_ROOT.parent, (
+            f"4 .parent hops overshoots repo root: {buggy}"
+        )
+
+    def test_cli_start_loop_uses_sys_executable(self):
+        """Bug: start_loop used hardcoded 'python' instead of sys.executable."""
+        from rdagent.app.cli import start_loop_cli
+        import inspect
+
+        source = inspect.getsource(start_loop_cli)
+
+        # The fixed code uses sys.executable in the generator list
+        assert "sys.executable" in source, (
+            "start_loop_cli should use sys.executable, not hardcoded 'python'"
+        )
+        # Should NOT contain the old hardcoded pattern
+        assert 'f"python ' not in source, (
+            "start_loop_cli should not contain hardcoded 'python' string"
+        )
+
+    def test_cli_parallel_not_hardcoded_model(self):
+        """Bug: parallel_cli hardcoded -m local in subprocess command."""
+        from rdagent.app.cli import parallel_cli
+        import inspect
+
+        source = inspect.getsource(parallel_cli)
+
+        # The fixed code no longer passes -m local as a cmd argument
+        assert '-m", "local"' not in source and '-m", \n        "local"' not in source and '"-m", "local"' not in source and '"local"]' not in source, (
+            "parallel_cli should not hardcode model=local in subprocess command list"
+        )
+
+    def test_cli_scripts_exist_at_resolved_paths(self):
+        """Verify scripts referenced by CLI commands exist at the resolved paths."""
+        from rdagent.app.cli import eval_all_cli, batch_backtest_cli, simple_eval_cli
+        from rdagent.app.cli import rebacktest_cli, report_cli, parallel_cli
+        import inspect
+
+        # All these commands use Path(__file__).parent.parent.parent as project_root
+        commands = {
+            "eval_all": "scripts/predix_full_eval.py",
+            "batch_backtest": "scripts/predix_batch_backtest.py",
+            "simple_eval": "scripts/predix_simple_eval.py",
+            "rebacktest": "scripts/predix_rebacktest_strategies.py",
+            "report": "scripts/predix_strategy_report.py",
+            "parallel": "scripts/predix_parallel.py",
+        }
+
+        for cmd_name, script_path in commands.items():
+            full_path = self.REPO_ROOT / script_path
+            assert full_path.exists(), (
+                f"CLI command '{cmd_name}' references {full_path} which does not exist. "
+                f"project_root depth may be wrong."
+            )
+
+    def test_start_loop_generator_script_exists(self):
+        """Bug: wrong project_root meant generator script not found."""
+        from rdagent.app.cli import start_loop_cli
+        import inspect
+
+        source = inspect.getsource(start_loop_cli)
+        # The generator should reference scripts/predix_smart_strategy_gen.py
+        assert "predix_smart_strategy_gen.py" in source, (
+            "start_loop_cli should reference predix_smart_strategy_gen.py"
+        )
+
+        script = self.REPO_ROOT / "scripts" / "predix_smart_strategy_gen.py"
+        assert script.exists(), (
+            f"Generator script not found at {script}"
+        )
+
+    def test_start_loop_uses_child_proc_not_pkill(self):
+        """Bug: cleanup used pkill -f which killed all instances system-wide."""
+        from rdagent.app.cli import start_loop_cli
+        import inspect
+
+        source = inspect.getsource(start_loop_cli)
+
+        # Fixed code uses child_proc.terminate() / child_proc.kill()
+        assert "child_proc" in source, (
+            "start_loop_cli should use child_proc variable for targeted cleanup"
+        )
+        # Should NOT contain the old broad pkill
+        assert "pkill" not in source, (
+            "start_loop_cli should not use broad pkill for process management"
+        )
+
+
+# ── Integration: full import checks ─────────────────────────────────────
+
+
+class TestImportsDontCrash:
+    """Verify that importing the fixed modules doesn't crash."""
+
+    def test_import_parallel_runner(self):
+        """ParallelRunner should import without errors."""
+        from scripts.predix_parallel import ParallelRunner, RunState
+        runner = ParallelRunner(num_runs=1, num_api_keys=1, model="local")
+        assert runner.num_runs == 1
+        assert len(runner.runs) == 1
+
+    def test_import_cli_app(self):
+        """CLI app should import without errors."""
+        from rdagent.app.cli import app
+        assert app is not None

test/local/test_bug_fixes.py

@@ -0,0 +1,281 @@
+"""
+Tests for bug fixes in strategy_orchestrator, factor_runner, backtest_engine,
+results_db, model_runner, optuna_optimizer, env, and related modules.
+
+Verifies:
+- strategy_orchestrator.py compiles (IndentationError was fixed)
+- factor_runner.py uses sys.executable (variable), not literal string
+- backtest_engine.py path depth is 4 (not 3) .parent hops
+- results_db.py path depth for factors/failed dirs is 4 (not 3)
+- model_runner.py DB connection closed via try/finally
+- factor_runner.py variable shadowing eliminated (run_id vs db_run_id)
+- optuna_optimizer.py no longer shadows imported logger
+- env.py Docker build output handles non-UTF-8 bytes
+- env.py conda env list parsing guards against empty lines
+- strategy_orchestrator.py exec() exception logged at ERROR level
+- strategy_orchestrator.py template validation warns on unreplaced {{...}}
+- factor_runner.py IC_max guard against scalar (AttributeError)
+- predix_parallel.py handle leak on Popen failure
+- predix_rebacktest_strategies.py bare except replaced with except Exception
+"""
+
+import ast
+import inspect
+import os
+import sys
+from pathlib import Path
+from unittest.mock import patch
+
+import pytest
+
+REPO_ROOT = Path(__file__).parent.parent.parent
+
+
+# ── Fix 1: strategy_orchestrator.py IndentationError ──────────────────────
+
+
+class TestStrategyOrchestratorSyntax:
+    def test_file_compiles(self):
+        """Bug: IndentationError at line 764 prevented the entire file from importing."""
+        import py_compile
+        py_compile.compile(
+            str(REPO_ROOT / "rdagent/components/coder/strategy_orchestrator.py"),
+            doraise=True,
+        )
+
+    def test_module_imports(self):
+        """Verify StrategyOrchestrator can be imported after syntax fix."""
+        from rdagent.components.coder.strategy_orchestrator import StrategyOrchestrator
+        assert StrategyOrchestrator is not None
+
+
+# ── Fix 2: factor_runner.py literal "sys.executable" ──────────────────────
+
+
+class TestFactorRunnerSysExecutable:
+    def test_not_literal_string(self):
+        """Bug: ["sys.executable", ...] was a literal string, not the variable."""
+        source = (REPO_ROOT / "rdagent/scenarios/qlib/developer/factor_runner.py").read_text()
+        # The fix should NOT contain the quoted literal 'sys.executable'
+        assert '"sys.executable"' not in source, (
+            "factor_runner.py still contains literal string 'sys.executable' — "
+            "should be sys.executable (variable)"
+        )
+        # Should use sys.executable (variable, part of a list)
+        assert "sys.executable" in source
+
+    def test_subprocess_run_with_check_false(self):
+        """Bug: subprocess.run without explicit check=False."""
+        source = (REPO_ROOT / "rdagent/scenarios/qlib/developer/factor_runner.py").read_text()
+        # The fix added check=False to the full-data factor run
+        assert 'check=False' in source, (
+            "subprocess.run should have explicit check=False for full-data factor run"
+        )
+
+
+# ── Fix 3: backtest_engine.py path depth ─────────────────────────────────
+
+
+class TestBacktestEnginePathDepth:
+    def test_path_depth_is_4(self):
+        """Bug: 3 .parent hops ended at rdagent/ instead of repo root."""
+        from rdagent.components.backtesting.backtest_engine import FactorBacktester
+
+        fb = FactorBacktester()
+        results = fb.results_path
+
+        # results_path should be under the repo root, not under rdagent/
+        assert REPO_ROOT in results.parents or results.parent == REPO_ROOT / "results", (
+            f"results_path={results} is not under repo root {REPO_ROOT}. "
+            "Path depth may still be wrong."
+        )
+        # The path should NOT be inside rdagent/
+        assert "rdagent/results" not in str(results).replace(str(REPO_ROOT), ""), (
+            f"results_path={results} appears to be inside rdagent/ directory"
+        )
+
+
+# ── Fix 4: results_db.py path depth ──────────────────────────────────────
+
+
+class TestResultsDBPathDepth:
+    def test_factors_dir_depth(self):
+        """Bug: 3 .parent hops from results_db.py ended at rdagent/."""
+        from rdagent.components.backtesting.results_db import ResultsDatabase
+        source = inspect.getsource(ResultsDatabase.generate_results_summary)
+
+        # After fix, should use .parent.parent.parent.parent (4 hops)
+        assert ".parent.parent.parent.parent" in source, (
+            "ResultsDatabase.generate_results_summary should use 4 .parent hops, not 3"
+        )
+
+
+# ── Fix 5: model_runner.py DB close ──────────────────────────────────────
+
+
+class TestModelRunnerDBClose:
+    def test_try_finally_for_db_close(self):
+        """Bug: db.close() was after add_backtest, not in finally block."""
+        source = (REPO_ROOT / "rdagent/scenarios/qlib/developer/model_runner.py").read_text()
+
+        # After fix, db.close() should be in a finally block or try/finally context
+        assert "finally:" in source, (
+            "model_runner.py should use try/finally to close DB connection"
+        )
+        assert "db.close()" in source
+
+
+# ── Fix 6: factor_runner.py variable shadowing ───────────────────────────
+
+
+class TestFactorRunnerShadowing:
+    def test_no_run_id_shadowing(self):
+        """Bug: run_id was reassigned from parallel run ID to DB row ID."""
+        source = (REPO_ROOT / "rdagent/scenarios/qlib/developer/factor_runner.py").read_text()
+
+        # After fix, parallel_run_id and db_run_id are separate variables
+        assert "parallel_run_id" in source, (
+            "factor_runner.py should use parallel_run_id for parallel run isolation"
+        )
+        assert "db_run_id" in source, (
+            "factor_runner.py should use db_run_id for DB row ID"
+        )
+
+
+# ── Fix 7: optuna_optimizer.py logger shadowing ──────────────────────────
+
+
+class TestOptunaLoggerShadowing:
+    def test_logger_not_reassigned(self):
+        """Bug: logger was reassigned from rdagent_logger to raw logging.getLogger."""
+        source = (REPO_ROOT / "rdagent/components/coder/optuna_optimizer.py").read_text()
+
+        # After fix, the module-level logger should be rdagent_logger
+        assert "from rdagent.log import rdagent_logger as logger" in source
+        # The second assignment should use a different name
+        assert "_optuna_logger" in source, (
+            "optuna_optimizer.py should not shadow the rdagent logger"
+        )
+
+
+# ── Fix 8: env.py UnicodeDecodeError ─────────────────────────────────────
+
+
+class TestEnvUnicodeDecode:
+    def test_decode_with_errors_replace(self):
+        """Bug: part.decode('utf-8') could raise UnicodeDecodeError."""
+        source = (REPO_ROOT / "rdagent/utils/env.py").read_text()
+
+        # After fix, decode uses errors="replace"
+        assert 'decode("utf-8", errors="replace")' in source, (
+            "env.py should handle non-UTF-8 Docker build output with errors='replace'"
+        )
+
+
+# ── Fix 9: env.py conda env list parsing ─────────────────────────────────
+
+
+class TestEnvCondaParsing:
+    def test_guard_against_empty_lines(self):
+        """Bug: line.split()[0] crashed on empty tokens."""
+        source = (REPO_ROOT / "rdagent/utils/env.py").read_text()
+
+        # After fix, guards against empty split results
+        assert "len(line.split()) > 0" in source, (
+            "env.py should guard against empty split results in conda env list parsing"
+        )
+
+
+# ── Fix 10: strategy_orchestrator.py exec() logging ──────────────────────
+
+
+class TestStrategyOrchExecLogging:
+    def test_error_logging_in_exec_handler(self):
+        """Bug: exec() exception was silently swallowed."""
+        source = (REPO_ROOT / "rdagent/components/coder/strategy_orchestrator.py").read_text()
+
+        # After fix, logger.error is called inside the except block
+        assert "logger.error" in source, (
+            "strategy_orchestrator.py should log exec() errors at ERROR level"
+        )
+
+
+# ── Fix 11: strategy_orchestrator.py template validation ─────────────────
+
+
+class TestStrategyOrchTemplateValidation:
+    def test_unreplaced_template_warning(self):
+        """Bug: no validation that {{...}} placeholders were replaced."""
+        source = (REPO_ROOT / "rdagent/components/coder/strategy_orchestrator.py").read_text()
+
+        # After fix, warns on unreplaced template variables
+        assert "Unreplaced template variables" in source, (
+            "strategy_orchestrator.py should warn on unreplaced {{...}} placeholders"
+        )
+
+
+# ── Fix 12: factor_runner.py IC_max guard ────────────────────────────────
+
+
+class TestFactorRunnerICMaxGuard:
+    def test_hasattr_guard(self):
+        """Bug: IC_max[...].index failed with AttributeError on scalar result."""
+        source = (REPO_ROOT / "rdagent/scenarios/qlib/developer/factor_runner.py").read_text()
+
+        # After fix, guards against scalar IC_max result
+        assert 'hasattr(IC_max, "index")' in source, (
+            "factor_runner.py should guard IC_max.index access with hasattr"
+        )
+
+
+# ── Fix 13: predix_parallel.py handle leak ───────────────────────────────
+
+
+class TestParallelRunnerHandleLeak:
+    def test_log_f_close_on_popen_failure(self):
+        """Bug: open() file handle leaked if Popen failed."""
+        source = (REPO_ROOT / "scripts/predix_parallel.py").read_text()
+
+        # After fix, log_f.close() is called before re-raise
+        assert "log_f.close()" in source, (
+            "predix_parallel.py should close log file handle on Popen failure"
+        )
+
+
+# ── Fix 14: predix_rebacktest_strategies.py bare except ──────────────────
+
+
+class TestRebacktestBareExcept:
+    def test_not_bare_except(self):
+        """Bug: bare except: pass swallowed all errors including SystemExit."""
+        source = (REPO_ROOT / "scripts/predix_rebacktest_strategies.py").read_text()
+
+        # After fix, should use except Exception, not bare except
+        assert "except Exception:" in source
+        assert "except:" not in source, (
+            "predix_rebacktest_strategies.py should not use bare except:"
+        )
+
+
+# ── Integration: import checks ───────────────────────────────────────────
+
+
+class TestAllImportsDontCrash:
+    def test_strategy_orchestrator_imports(self):
+        """Verify all fixed modules import without errors."""
+        import rdagent.components.coder.strategy_orchestrator  # noqa: F401
+
+    def test_factor_runner_imports(self):
+        import rdagent.scenarios.qlib.developer.factor_runner  # noqa: F401
+
+    def test_backtest_engine_imports(self):
+        import rdagent.components.backtesting.backtest_engine  # noqa: F401
+
+    def test_results_db_imports(self):
+        import rdagent.components.backtesting.results_db  # noqa: F401
+
+    def test_model_runner_imports(self):
+        import rdagent.scenarios.qlib.developer.model_runner  # noqa: F401
+
+    def test_optuna_optimizer_imports(self):
+        import rdagent.components.coder.optuna_optimizer  # noqa: F401

test/qlib/test_core_framework.py

@@ -0,0 +1,243 @@
+"""Tests for rdagent.core — the core framework abstractions."""
+
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+from unittest.mock import MagicMock
+
+import pytest
+
+PROJECT_ROOT = Path(__file__).parent.parent.parent
+sys.path.insert(0, str(PROJECT_ROOT))
+
+
+# =============================================================================
+# Feedback base class
+# =============================================================================
+
+
+class TestFeedback:
+    def test_default_is_acceptable_returns_true(self):
+        from rdagent.core.evaluation import Feedback
+        fb = Feedback()
+        assert fb.is_acceptable() is True
+
+    def test_default_finished_returns_true(self):
+        from rdagent.core.evaluation import Feedback
+        fb = Feedback()
+        assert fb.finished() is True
+
+    def test_default_bool_is_true(self):
+        from rdagent.core.evaluation import Feedback
+        fb = Feedback()
+        assert bool(fb) is True
+
+
+# =============================================================================
+# EvoStep dataclass
+# =============================================================================
+
+
+class TestEvoStep:
+    def test_default_construction(self):
+        from rdagent.core.evolving_framework import EvoStep
+        es = EvoStep(evolvable_subjects="mock_evo")
+        assert es.evolvable_subjects == "mock_evo"
+        assert es.queried_knowledge is None
+        assert es.feedback is None
+
+    def test_full_construction(self):
+        from rdagent.core.evolving_framework import EvoStep, QueriedKnowledge
+        qk = QueriedKnowledge()
+        es = EvoStep(evolvable_subjects="evo", queried_knowledge=qk, feedback="fb")
+        assert es.queried_knowledge is qk
+        assert es.feedback == "fb"
+
+    def test_equality_by_reference(self):
+        from rdagent.core.evolving_framework import EvoStep
+        es1 = EvoStep(evolvable_subjects="a")
+        es2 = EvoStep(evolvable_subjects="a")
+        assert es1 == es2
+
+
+# =============================================================================
+# Scenario base class
+# =============================================================================
+
+
+class TestScenario:
+    def test_source_data_default_returns_empty_string(self):
+        from rdagent.core.scenario import Scenario
+        
+        class MinimalScenario(Scenario):
+            @property
+            def background(self) -> str: return "bg"
+            @property
+            def rich_style_description(self) -> str: return "rich"
+            def get_scenario_all_desc(self, **kwargs) -> str: return "all"
+            def get_runtime_environment(self) -> str: return "env"
+
+        scen = MinimalScenario()
+        assert scen.source_data == ""
+
+    def test_source_data_property_calls_get_source_data_desc(self):
+        from rdagent.core.scenario import Scenario
+
+        class FakeScenario(Scenario):
+            @property
+            def background(self) -> str: return "bg"
+            @property
+            def rich_style_description(self) -> str: return "rich"
+            def get_scenario_all_desc(self, **kwargs) -> str: return "all"
+            def get_runtime_environment(self) -> str: return "env"
+            def get_source_data_desc(self, task=None) -> str: return "custom_data"
+
+        scen = FakeScenario()
+        assert scen.source_data == "custom_data"
+
+
+# =============================================================================
+# EvolvingStrategy base class
+# =============================================================================
+
+
+class TestEvolvingStrategy:
+    def test_init_stores_scenario(self):
+        from rdagent.core.evolving_framework import EvolvingStrategy
+        
+        class MinimalStrategy(EvolvingStrategy):
+            def evolve_iter(self, evo, queried_knowledge=None, evolving_trace=None):
+                yield evo
+        
+        mock_scen = MagicMock()
+        es = MinimalStrategy(mock_scen)
+        assert es.scen is mock_scen
+
+
+# =============================================================================
+# IterEvaluator base class
+# =============================================================================
+
+
+class TestIterEvaluator:
+    def test_evaluate_returns_feedback(self):
+        from rdagent.core.evaluation import Feedback
+        from rdagent.core.evolving_framework import IterEvaluator, EvolvableSubjects
+
+        class MyFeedback(Feedback):
+            pass
+
+        class MyEvaluator(IterEvaluator):
+            def evaluate_iter(self):
+                evo = yield MyFeedback()
+                yield MyFeedback()
+                return MyFeedback()
+
+        eva = MyEvaluator()
+        result = eva.evaluate(EvolvableSubjects())
+        assert isinstance(result, MyFeedback)
+
+    def test_evaluate_iter_send_none_stops(self):
+        """Sending None mid-iteration triggers StopIteration with final feedback."""
+        from rdagent.core.evaluation import Feedback
+        from rdagent.core.evolving_framework import IterEvaluator
+
+        class MyEvaluator(IterEvaluator):
+            def evaluate_iter(self):
+                yield Feedback()            # kick-off (none)
+                evo_next = yield Feedback()  # partial eval
+                if evo_next is None:
+                    return Feedback()        # early return
+                return Feedback()            # normal path
+
+        eva = MyEvaluator()
+        gen = eva.evaluate_iter()
+        next(gen)           # kick-off → first Feedback
+        gen.send("any")     # evo gets "any" → second Feedback (evo_next NOT assigned yet)
+        with pytest.raises(StopIteration):
+            gen.send(None)  # evo_next = None → return → StopIteration
+
+
+# =============================================================================
+# Developer base class
+# =============================================================================
+
+
+class TestDeveloper:
+    def test_develop_raises_not_implemented(self):
+        from rdagent.core.developer import Developer
+
+        class MinimalDeveloper(Developer):
+            def develop(self, exp):
+                return super().develop(exp)
+
+        dev = MinimalDeveloper(MagicMock())
+        with pytest.raises(NotImplementedError):
+            dev.develop(MagicMock())
+
+
+# =============================================================================
+# Knowledge / QueriedKnowledge
+# =============================================================================
+
+
+class TestKnowledgeHierarchy:
+    def test_knowledge_pass_through(self):
+        from rdagent.core.evolving_framework import Knowledge, QueriedKnowledge
+        k = Knowledge()
+        qk = QueriedKnowledge()
+        assert isinstance(k, Knowledge)
+        assert isinstance(qk, QueriedKnowledge)
+
+
+# =============================================================================
+# EvolvingAgent (abstract interface)
+# =============================================================================
+
+
+class TestEvolvingAgent:
+    def test_ragevo_agent_init(self):
+        from rdagent.core.evolving_agent import RAGEvoAgent
+        mock_strategy = MagicMock()
+        mock_rag = MagicMock()
+        agent = RAGEvoAgent.__new__(RAGEvoAgent)
+        RAGEvoAgent.__init__(agent, max_loop=5, evolving_strategy=mock_strategy, rag=mock_rag)
+        assert agent.max_loop == 5
+        assert agent.evolving_strategy is mock_strategy
+        assert agent.rag is mock_rag
+
+    def test_ragevo_agent_default_knowledge_flags(self):
+        from rdagent.core.evolving_agent import RAGEvoAgent
+        agent = RAGEvoAgent.__new__(RAGEvoAgent)
+        RAGEvoAgent.__init__(agent, max_loop=3, evolving_strategy=MagicMock(), rag=MagicMock())
+        assert agent.with_knowledge is False
+        assert agent.knowledge_self_gen is False
+        assert agent.enable_filelock is False
+
+    def test_ragevo_agent_with_knowledge_enabled(self):
+        from rdagent.core.evolving_agent import RAGEvoAgent
+        agent = RAGEvoAgent.__new__(RAGEvoAgent)
+        RAGEvoAgent.__init__(
+            agent, max_loop=3, evolving_strategy=MagicMock(), rag=MagicMock(),
+            with_knowledge=True, knowledge_self_gen=True,
+            enable_filelock=True, filelock_path="/tmp/test.lock",
+        )
+        assert agent.with_knowledge is True
+        assert agent.knowledge_self_gen is True
+        assert agent.enable_filelock is True
+        assert agent.filelock_path == "/tmp/test.lock"
+
+
+# =============================================================================
+# EvolvableSubjects clone
+# =============================================================================
+
+
+class TestEvolvableSubjects:
+    def test_clone_produces_deep_copy(self):
+        from rdagent.core.evolving_framework import EvolvableSubjects
+        es = EvolvableSubjects()
+        clone = es.clone()
+        assert clone is not es
+        assert type(clone) is type(es)

test/qlib/test_costeer_feedback.py

@@ -0,0 +1,338 @@
+"""Tests for CoSTEER feedback types and EvolvingItem."""
+
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+from unittest.mock import MagicMock
+
+import pytest
+
+PROJECT_ROOT = Path(__file__).parent.parent.parent
+sys.path.insert(0, str(PROJECT_ROOT))
+
+
+# =============================================================================
+# CoSTEERSingleFeedback
+# =============================================================================
+
+
+class TestCoSTEERSingleFeedback:
+    def test_construction_with_valid_fields(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedback
+        fb = CoSTEERSingleFeedback(
+            execution="exec ok",
+            return_checking="return ok",
+            code="code ok",
+            final_decision=True,
+        )
+        assert fb.execution == "exec ok"
+        assert fb.return_checking == "return ok"
+        assert fb.code == "code ok"
+        assert fb.final_decision is True
+
+    def test_bool_returns_final_decision(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedback
+        fb_true = CoSTEERSingleFeedback(execution="x", return_checking="x", code="x", final_decision=True)
+        fb_false = CoSTEERSingleFeedback(execution="x", return_checking="x", code="x", final_decision=False)
+        assert bool(fb_true) is True
+        assert bool(fb_false) is False
+
+    def test_default_values(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedback
+        fb = CoSTEERSingleFeedback(execution="x", return_checking=None, code="x")
+        assert fb.final_decision is None
+        assert fb.raw_execution == ""
+        assert fb.source_feedback == {}
+
+    def test_val_and_update_init_dict_converts_boolean(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedback
+        assert CoSTEERSingleFeedback.val_and_update_init_dict(
+            {"execution": "x", "return_checking": "y", "code": "z", "final_decision": "true"}
+        )["final_decision"] is True
+        assert CoSTEERSingleFeedback.val_and_update_init_dict(
+            {"execution": "x", "return_checking": "y", "code": "z", "final_decision": "false"}
+        )["final_decision"] is False
+        assert CoSTEERSingleFeedback.val_and_update_init_dict(
+            {"execution": "x", "return_checking": "y", "code": "z", "final_decision": "True"}
+        )["final_decision"] is True
+        assert CoSTEERSingleFeedback.val_and_update_init_dict(
+            {"execution": "x", "return_checking": "y", "code": "z", "final_decision": "False"}
+        )["final_decision"] is False
+
+    def test_val_and_update_init_dict_rejects_non_boolean(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedback
+        with pytest.raises(ValueError):
+            CoSTEERSingleFeedback.val_and_update_init_dict(
+                {"execution": "x", "return_checking": "y", "code": "z", "final_decision": 42}
+            )
+
+    def test_val_and_update_init_dict_missing_final_decision_raises(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedback
+        with pytest.raises(ValueError, match="final_decision"):
+            CoSTEERSingleFeedback.val_and_update_init_dict(
+                {"execution": "x", "return_checking": "y", "code": "z"}
+            )
+
+    def test_val_and_update_init_dict_json_dumps_non_string_attrs(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedback
+        import json
+        data = {
+            "execution": {"key": "val"},
+            "return_checking": ["list"],
+            "code": 123,
+            "final_decision": True,
+        }
+        result = CoSTEERSingleFeedback.val_and_update_init_dict(data)
+        for attr in ("execution", "return_checking", "code"):
+            # Should have been converted to JSON string
+            assert isinstance(result[attr], str)
+            _ = json.loads(result[attr])  # valid JSON
+
+    def test_merge_all_true_decisions(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedback
+        fb1 = CoSTEERSingleFeedback(execution="a", return_checking="ra", code="c1", final_decision=True)
+        fb2 = CoSTEERSingleFeedback(execution="b", return_checking="rb", code="c2", final_decision=True)
+        merged = CoSTEERSingleFeedback.merge([fb1, fb2])
+        assert merged.final_decision is True
+
+    def test_merge_one_false_decision(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedback
+        fb1 = CoSTEERSingleFeedback(execution="a", return_checking="ra", code="c1", final_decision=True)
+        fb2 = CoSTEERSingleFeedback(execution="b", return_checking="rb", code="c2", final_decision=False)
+        merged = CoSTEERSingleFeedback.merge([fb1, fb2])
+        assert merged.final_decision is False
+
+    def test_merge_concatenates_fields(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedback
+        fb1 = CoSTEERSingleFeedback(execution="A", return_checking="RA", code="C1", final_decision=True)
+        fb2 = CoSTEERSingleFeedback(execution="B", return_checking="RB", code="C2", final_decision=True)
+        merged = CoSTEERSingleFeedback.merge([fb1, fb2])
+        assert "A\n\nB" in merged.execution
+        assert "RA\n\nRB" in merged.return_checking
+        assert "C1\n\nC2" in merged.code
+
+    def test_merge_skips_none_fields(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedback
+        fb1 = CoSTEERSingleFeedback(execution="A", return_checking="RA", code="C1", final_decision=True)
+        fb2 = CoSTEERSingleFeedback(execution="B", return_checking=None, code="C2", final_decision=True)
+        merged = CoSTEERSingleFeedback.merge([fb1, fb2])
+        assert merged.execution == "A\n\nB"
+        assert merged.return_checking == "RA"
+        assert merged.code == "C1\n\nC2"
+
+    def test_merge_aggregates_source_feedback(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedback
+        fb1 = CoSTEERSingleFeedback(execution="a", return_checking="x", code="c", final_decision=True,
+                                     source_feedback={"eval1": True})
+        fb2 = CoSTEERSingleFeedback(execution="b", return_checking="y", code="d", final_decision=True,
+                                     source_feedback={"eval2": False})
+        merged = CoSTEERSingleFeedback.merge([fb1, fb2])
+        assert merged.source_feedback == {"eval1": True, "eval2": False}
+
+    def test_str_contains_all_sections(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedback
+        fb = CoSTEERSingleFeedback(execution="exec", return_checking="ret", code="code", final_decision=True)
+        s = str(fb)
+        assert "Execution" in s
+        assert "Return Checking" in s
+        assert "Code" in s
+        assert "Final Decision" in s
+        assert "SUCCESS" in s
+
+    def test_str_shows_fail_for_false(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedback
+        fb = CoSTEERSingleFeedback(execution="exec", return_checking="ret", code="code", final_decision=False)
+        assert "FAIL" in str(fb)
+
+
+# =============================================================================
+# CoSTEERSingleFeedbackDeprecated
+# =============================================================================
+
+
+class TestCoSTEERSingleFeedbackDeprecated:
+    def test_property_getters(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedbackDeprecated
+        fb = CoSTEERSingleFeedbackDeprecated(
+            execution_feedback="exec",
+            code_feedback="code",
+            value_feedback="val",
+            shape_feedback="shape",
+            final_decision=True,
+            final_feedback="final",
+            value_generated_flag=True,
+            final_decision_based_on_gt=True,
+        )
+        assert fb.execution == "exec"
+        assert fb.code == "code"
+        assert fb.final_decision is True
+        assert fb.value_generated_flag is True
+        assert fb.final_decision_based_on_gt is True
+
+    def test_return_checking_when_value_generated(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedbackDeprecated
+        fb = CoSTEERSingleFeedbackDeprecated(
+            value_generated_flag=True, value_feedback="vals", shape_feedback="shapes",
+        )
+        rc = fb.return_checking
+        assert "vals" in rc
+        assert "shapes" in rc
+
+    def test_return_checking_when_no_value_generated(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedbackDeprecated
+        fb = CoSTEERSingleFeedbackDeprecated(value_generated_flag=False)
+        assert fb.return_checking is None
+
+    def test_setters_work(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedbackDeprecated
+        fb = CoSTEERSingleFeedbackDeprecated()
+        fb.execution = "new_exec"
+        fb.code = "new_code"
+        fb.return_checking = "new_rc"
+        assert fb.execution_feedback == "new_exec"
+        assert fb.code_feedback == "new_code"
+        assert fb.value_feedback == "new_rc"
+        assert fb.shape_feedback == "new_rc"
+
+    def test_str_contains_all_sections(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedbackDeprecated
+        fb = CoSTEERSingleFeedbackDeprecated(
+            execution_feedback="exec", shape_feedback="shape",
+            code_feedback="code", value_feedback="val",
+            final_feedback="final", final_decision=True,
+        )
+        s = str(fb)
+        for keyword in ("Execution", "Shape", "Code", "Value", "Final Decision", "SUCCESS"):
+            assert keyword in s
+
+
+# =============================================================================
+# CoSTEERMultiFeedback
+# =============================================================================
+
+
+class TestCoSTEERMultiFeedback:
+    def _make_fb(self, decision=True):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERSingleFeedback
+        return CoSTEERSingleFeedback(execution="x", return_checking="x", code="x", final_decision=decision)
+
+    def test_getitem(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERMultiFeedback
+        fb1, fb2 = self._make_fb(True), self._make_fb(False)
+        mf = CoSTEERMultiFeedback([fb1, fb2])
+        assert mf[0].final_decision is True
+        assert mf[1].final_decision is False
+
+    def test_len(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERMultiFeedback
+        assert len(CoSTEERMultiFeedback([self._make_fb()])) == 1
+        assert len(CoSTEERMultiFeedback([])) == 0
+
+    def test_append(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERMultiFeedback
+        mf = CoSTEERMultiFeedback([])
+        mf.append(self._make_fb(True))
+        assert len(mf) == 1
+        assert mf[0].final_decision is True
+
+    def test_iter(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERMultiFeedback
+        fbs = [self._make_fb(True), self._make_fb(True)]
+        mf = CoSTEERMultiFeedback(fbs)
+        assert list(mf) == fbs
+
+    def test_finished_all_true(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERMultiFeedback
+        mf = CoSTEERMultiFeedback([self._make_fb(True), self._make_fb(True)])
+        assert mf.finished() is True
+
+    def test_finished_one_false(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERMultiFeedback
+        mf = CoSTEERMultiFeedback([self._make_fb(True), self._make_fb(False)])
+        assert mf.finished() is False
+
+    def test_finished_with_none_skips(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERMultiFeedback
+        mf = CoSTEERMultiFeedback([self._make_fb(True), None])
+        assert mf.finished() is True  # None = skipped = accepted
+
+    def test_bool_all_true(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERMultiFeedback
+        assert bool(CoSTEERMultiFeedback([self._make_fb(True), self._make_fb(True)])) is True
+
+    def test_bool_one_false(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERMultiFeedback
+        assert bool(CoSTEERMultiFeedback([self._make_fb(True), self._make_fb(False)])) is False
+
+    def test_is_acceptable_delegates(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERMultiFeedback
+        mf = CoSTEERMultiFeedback([self._make_fb(True), self._make_fb(True)])
+        assert mf.is_acceptable() is True
+
+
+# =============================================================================
+# EvolvingItem
+# =============================================================================
+
+
+class TestEvolvingItem:
+    def test_construction_without_gt(self):
+        from rdagent.components.coder.CoSTEER.evolvable_subjects import EvolvingItem
+        from rdagent.core.experiment import Task
+
+        t1 = Task(name="task1")
+        t2 = Task(name="task2")
+        ei = EvolvingItem(sub_tasks=[t1, t2])
+        assert len(ei.sub_tasks) == 2
+        assert ei.sub_gt_implementations is None
+
+    def test_construction_with_matching_gt(self):
+        from rdagent.components.coder.CoSTEER.evolvable_subjects import EvolvingItem
+        from rdagent.core.experiment import Task, FBWorkspace
+        t1, t2 = Task(name="t1"), Task(name="t2")
+        ws1, ws2 = FBWorkspace(), FBWorkspace()
+        ei = EvolvingItem(sub_tasks=[t1, t2], sub_gt_implementations=[ws1, ws2])
+        assert ei.sub_gt_implementations == [ws1, ws2]
+
+    def test_mismatched_gt_length_resets_to_none(self):
+        from rdagent.components.coder.CoSTEER.evolvable_subjects import EvolvingItem
+        from rdagent.core.experiment import Task, FBWorkspace
+        t1, t2 = Task(name="t1"), Task(name="t2")
+        ei = EvolvingItem(sub_tasks=[t1, t2], sub_gt_implementations=[FBWorkspace()])
+        assert ei.sub_gt_implementations is None
+
+    def test_from_experiment(self):
+        from rdagent.components.coder.CoSTEER.evolvable_subjects import EvolvingItem
+        from rdagent.core.experiment import Experiment, Task
+
+        exp = Experiment(sub_tasks=[Task(name="x")])
+        exp.based_experiments = ["base"]
+        exp.experiment_workspace = "ws"
+        ei = EvolvingItem.from_experiment(exp)
+        assert len(ei.sub_tasks) == 1
+        assert ei.based_experiments == ["base"]
+        assert ei.experiment_workspace == "ws"
+
+
+# =============================================================================
+# CoSTEERQueriedKnowledge
+# =============================================================================
+
+
+class TestCoSTEERQueriedKnowledge:
+    def test_default_construction(self):
+        from rdagent.components.coder.CoSTEER.knowledge_management import CoSTEERQueriedKnowledge
+        qk = CoSTEERQueriedKnowledge()
+        assert qk.success_task_to_knowledge_dict == {}
+        assert qk.failed_task_info_set == set()
+
+    def test_with_data(self):
+        from rdagent.components.coder.CoSTEER.knowledge_management import CoSTEERQueriedKnowledge
+        qk = CoSTEERQueriedKnowledge(
+            success_task_to_knowledge_dict={"a": "knowledge_a"},
+            failed_task_info_set={"fail1", "fail2"},
+        )
+        assert qk.success_task_to_knowledge_dict["a"] == "knowledge_a"
+        assert "fail1" in qk.failed_task_info_set
+        assert "fail2" in qk.failed_task_info_set

test/qlib/test_costeer_strategy.py

@@ -0,0 +1,225 @@
+"""Tests for CoSTEER config, task, and evolve strategy population logic."""
+
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+from unittest.mock import MagicMock, patch
+
+import pytest
+
+PROJECT_ROOT = Path(__file__).parent.parent.parent
+sys.path.insert(0, str(PROJECT_ROOT))
+
+
+# =============================================================================
+# CoSTEERSettings
+# =============================================================================
+
+
+class TestCoSTEERSettings:
+    def test_default_values(self):
+        from rdagent.components.coder.CoSTEER.config import CoSTEERSettings
+        s = CoSTEERSettings()
+        assert s.max_loop == 1
+        assert s.fail_task_trial_limit == 5
+        assert s.v2_query_component_limit == 1
+        assert s.v2_query_error_limit == 1
+        assert s.v2_query_former_trace_limit == 3
+        assert s.v2_add_fail_attempt_to_latest_successful_execution is False
+        assert s.v2_knowledge_sampler == 1.0
+        assert s.coder_use_cache is False
+        assert s.enable_filelock is False
+
+    def test_singleton_instance(self):
+        from rdagent.components.coder.CoSTEER.config import CoSTEER_SETTINGS
+        from rdagent.components.coder.CoSTEER.config import CoSTEERSettings
+        assert isinstance(CoSTEER_SETTINGS, CoSTEERSettings)
+        assert CoSTEER_SETTINGS.max_loop == 1
+
+
+# =============================================================================
+# CoSTEERTask
+# =============================================================================
+
+
+class TestCoSTEERTask:
+    def test_base_code_stored(self):
+        from rdagent.components.coder.CoSTEER.task import CoSTEERTask
+        t = CoSTEERTask(name="test", base_code="print(1)")
+        assert t.base_code == "print(1)"
+
+    def test_base_code_none_by_default(self):
+        from rdagent.components.coder.CoSTEER.task import CoSTEERTask
+        t = CoSTEERTask(name="test")
+        assert t.base_code is None
+
+
+# =============================================================================
+# MultiProcessEvolvingStrategy.assign_code_list_to_evo
+# =============================================================================
+
+
+class TestAssignCodeListToEvo:
+    def test_empty_code_list_noops(self):
+        from rdagent.components.coder.CoSTEER.evolving_strategy import MultiProcessEvolvingStrategy
+        from rdagent.components.coder.CoSTEER.evolvable_subjects import EvolvingItem
+        from rdagent.core.experiment import Task
+
+        strat = MultiProcessEvolvingStrategy.__new__(MultiProcessEvolvingStrategy)
+        MultiProcessEvolvingStrategy.__init__(strat, scen=MagicMock(), settings=MagicMock())
+
+        ei = EvolvingItem(sub_tasks=[Task(name="t1")])
+        ei.experiment_workspace = MagicMock()
+        result = strat.assign_code_list_to_evo([{}], ei)
+        assert result is ei
+
+    def test_none_entry_is_skipped(self):
+        from rdagent.components.coder.CoSTEER.evolving_strategy import MultiProcessEvolvingStrategy
+        from rdagent.components.coder.CoSTEER.evolvable_subjects import EvolvingItem
+        from rdagent.core.experiment import Task
+
+        strat = MultiProcessEvolvingStrategy.__new__(MultiProcessEvolvingStrategy)
+        MultiProcessEvolvingStrategy.__init__(strat, scen=MagicMock(), settings=MagicMock())
+
+        ei = EvolvingItem(sub_tasks=[Task(name="t1")])
+        ei.experiment_workspace = MagicMock()
+        result = strat.assign_code_list_to_evo([None], ei)
+        assert result.sub_workspace_list[0] is None  # unchanged
+
+    def test_code_injects_files(self):
+        from rdagent.components.coder.CoSTEER.evolving_strategy import MultiProcessEvolvingStrategy
+        from rdagent.components.coder.CoSTEER.evolvable_subjects import EvolvingItem
+        from rdagent.core.experiment import Task
+
+        strat = MultiProcessEvolvingStrategy.__new__(MultiProcessEvolvingStrategy)
+        MultiProcessEvolvingStrategy.__init__(strat, scen=MagicMock(), settings=MagicMock())
+
+        ei = EvolvingItem(sub_tasks=[Task(name="t1")])
+        mock_ws = MagicMock()
+        ei.experiment_workspace = mock_ws
+
+        strat.assign_code_list_to_evo([{"factor.py": "x=1"}], ei)
+        mock_ws.inject_files.assert_called_once_with(**{"factor.py": "x=1"})
+
+    def test_change_summary_extracted(self):
+        from rdagent.components.coder.CoSTEER.evolving_strategy import MultiProcessEvolvingStrategy
+        from rdagent.components.coder.CoSTEER.evolvable_subjects import EvolvingItem
+        from rdagent.core.experiment import Task
+
+        strat = MultiProcessEvolvingStrategy.__new__(MultiProcessEvolvingStrategy)
+        MultiProcessEvolvingStrategy.__init__(strat, scen=MagicMock(), settings=MagicMock())
+
+        mock_ws = MagicMock()
+        ei = EvolvingItem(sub_tasks=[Task(name="t1")])
+        ei.experiment_workspace = mock_ws
+
+        strat.assign_code_list_to_evo([{"__change_summary__": "summary", "factor.py": "x"}], ei)
+        assert mock_ws.change_summary == "summary"
+        # change_summary should have been popped from dict
+        mock_ws.inject_files.assert_called_once_with(**{"factor.py": "x"})
+
+
+# =============================================================================
+# MultiProcessEvolvingStrategy.evolve_iter
+# =============================================================================
+
+
+class TestEvolveIter:
+    def _make_strat(self):
+        from rdagent.components.coder.CoSTEER.evolving_strategy import MultiProcessEvolvingStrategy
+        from rdagent.components.coder.CoSTEER.config import CoSTEERSettings
+        strat = MultiProcessEvolvingStrategy(
+            scen=MagicMock(), settings=CoSTEERSettings(), improve_mode=False,
+        )
+        return strat
+
+    def _make_evo(self, n_tasks=2):
+        from rdagent.components.coder.CoSTEER.evolvable_subjects import EvolvingItem
+        from rdagent.core.experiment import Task
+        tasks = [Task(name=f"task_{i}") for i in range(n_tasks)]
+        for t in tasks:
+            t.get_task_information = MagicMock(return_value=f"info_{t.name}")
+        ei = EvolvingItem(sub_tasks=tasks)
+        ei.experiment_workspace = MagicMock()
+        return ei
+
+    def test_raises_without_queried_knowledge(self):
+        strat = self._make_strat()
+        evo = self._make_evo()
+        with pytest.raises(ValueError, match="queried_knowledge"):
+            next(strat.evolve_iter(evo=evo, queried_knowledge=None))
+
+    def test_successful_tasks_not_scheduled(self):
+        from rdagent.components.coder.CoSTEER.knowledge_management import CoSTEERQueriedKnowledge
+        strat = self._make_strat()
+        evo = self._make_evo(n_tasks=1)
+        qk = CoSTEERQueriedKnowledge(
+            success_task_to_knowledge_dict={
+                "info_task_0": MagicMock(implementation=MagicMock(file_dict={"f.py": "x"})),
+            },
+        )
+        # evolve_iter is a generator, next() starts it
+        gen = strat.evolve_iter(evo=evo, queried_knowledge=qk)
+        # Should yield the evo (populated from success knowledge)
+        result = next(gen)
+        # The task was already successful, so no new scheduling
+        assert result is evo
+
+    def test_failed_tasks_skipped(self):
+        from rdagent.components.coder.CoSTEER.knowledge_management import CoSTEERQueriedKnowledge
+        strat = self._make_strat()
+        evo = self._make_evo(n_tasks=1)
+        qk = CoSTEERQueriedKnowledge(
+            failed_task_info_set={"info_task_0"},
+        )
+        gen = strat.evolve_iter(evo=evo, queried_knowledge=qk)
+        result = next(gen)
+        # Task skipped because it's in failed_set
+        assert result is evo
+
+    def test_improve_mode_skips_with_no_last_feedback(self):
+        from rdagent.components.coder.CoSTEER.knowledge_management import CoSTEERQueriedKnowledge
+        strat = self._make_strat()
+        strat.improve_mode = True
+        evo = self._make_evo(n_tasks=1)
+        qk = CoSTEERQueriedKnowledge()
+        gen = strat.evolve_iter(evo=evo, queried_knowledge=qk, evolving_trace=[])
+        result = next(gen)
+        # In improve_mode with no last_feedback, task should be skipped
+        # (code_list[0] should be {} — empty implementation)
+        assert result is evo
+
+    def test_non_improve_mode_schedules_new_tasks(self):
+        """Tasks not in success/failed should be scheduled."""
+        from rdagent.components.coder.CoSTEER.knowledge_management import CoSTEERQueriedKnowledge
+        strat = self._make_strat()
+        evo = self._make_evo(n_tasks=1)
+        qk = CoSTEERQueriedKnowledge()
+
+        with patch(
+            "rdagent.components.coder.CoSTEER.evolving_strategy.multiprocessing_wrapper",
+            return_value=[{"factor.py": "x=1"}],
+        ):
+            gen = strat.evolve_iter(evo=evo, queried_knowledge=qk)
+            result = next(gen)
+            assert result is evo
+
+
+# =============================================================================
+# CoSTEERMultiEvaluator (partial)
+# =============================================================================
+
+
+class TestCoSTEERMultiEvaluator:
+    def test_init_with_single_evaluator(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERMultiEvaluator
+        mock_eval = MagicMock()
+        eva = CoSTEERMultiEvaluator(single_evaluator=mock_eval, scen=MagicMock())
+        assert eva.single_evaluator is mock_eval
+
+    def test_init_with_evaluator_list(self):
+        from rdagent.components.coder.CoSTEER.evaluators import CoSTEERMultiEvaluator
+        mock_list = [MagicMock(), MagicMock()]
+        eva = CoSTEERMultiEvaluator(single_evaluator=mock_list, scen=MagicMock())
+        assert eva.single_evaluator is mock_list

test/qlib/test_factor_coder.py

@@ -0,0 +1,221 @@
+"""Tests for factor_coder — evaluators, task, workspace."""
+
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+from unittest.mock import MagicMock, patch
+
+import pytest
+
+PROJECT_ROOT = Path(__file__).parent.parent.parent
+sys.path.insert(0, str(PROJECT_ROOT))
+
+
+# =============================================================================
+# FactorTask
+# =============================================================================
+
+
+class TestFactorTask:
+    def test_construction_fields(self):
+        from rdagent.components.coder.factor_coder.factor import FactorTask
+        t = FactorTask(
+            factor_name="f1",
+            factor_description="desc",
+            factor_formulation="formula",
+            variables={"x": 1},
+            resource="r1",
+        )
+        assert t.factor_name == "f1"
+        assert t.factor_description == "desc"
+        assert t.factor_formulation == "formula"
+        assert t.variables == {"x": 1}
+        assert t.factor_resources == "r1"
+        assert t.factor_implementation is False
+        assert t.base_code is None  # from CoSTEERTask
+
+    def test_get_task_information(self):
+        from rdagent.components.coder.factor_coder.factor import FactorTask
+        t = FactorTask("f1", "desc", "formula", variables={"x": 1})
+        info = t.get_task_information()
+        assert "factor_name: f1" in info
+        assert "factor_description: desc" in info
+        assert "factor_formulation: formula" in info
+        assert "variables: {'x': 1}" in info
+
+    def test_get_task_brief_information(self):
+        from rdagent.components.coder.factor_coder.factor import FactorTask
+        t = FactorTask("f1", "desc", "formula")
+        info = t.get_task_brief_information()
+        assert "factor_name: f1" in info
+
+    def test_get_task_information_and_implementation_result(self):
+        from rdagent.components.coder.factor_coder.factor import FactorTask
+        t = FactorTask("f1", "desc", "formula")
+        result = t.get_task_information_and_implementation_result()
+        assert result["factor_name"] == "f1"
+        assert result["factor_description"] == "desc"
+        assert "factor_implementation" in result
+
+    def test_from_dict(self):
+        from rdagent.components.coder.factor_coder.factor import FactorTask
+        d = {
+            "factor_name": "f2",
+            "factor_description": "d2",
+            "factor_formulation": "f2",
+            "variables": {},
+            "resource": None,
+            "factor_implementation": True,
+        }
+        t = FactorTask.from_dict(d)
+        assert t.factor_name == "f2"
+        assert t.factor_implementation is True
+
+    def test_repr(self):
+        from rdagent.components.coder.factor_coder.factor import FactorTask
+        t = FactorTask("myfactor", "desc", "formula")
+        assert "FactorTask" in repr(t)
+        assert "myfactor" in repr(t)
+
+
+# =============================================================================
+# FactorFBWorkspace
+# =============================================================================
+
+
+class TestFactorFBWorkspace:
+    def test_init_sets_workspace_path(self):
+        from rdagent.components.coder.factor_coder.factor import FactorFBWorkspace, FactorTask
+        t = FactorTask("f1", "desc", "formula")
+        ws = FactorFBWorkspace(target_task=t)
+        assert ws.workspace_path is not None
+        # Directory is created lazily by execute(), not in __init__
+        assert isinstance(ws.workspace_path, Path)
+
+    def test_execute_returns_message_and_dataframe(self):
+        from rdagent.components.coder.factor_coder.factor import FactorFBWorkspace, FactorTask
+        t = FactorTask("f1", "desc", "formula")
+        t.version = 1
+        ws = FactorFBWorkspace(target_task=t)
+        # Inject valid factor code
+        ws.inject_files(**{
+            "factor.py": (
+                "import pandas as pd\n"
+                "import numpy as np\n"
+                "data = pd.read_hdf('intraday_pv.h5', key='data')\n"
+                "factor_val = data['$close'].pct_change()\n"
+                "factor_val = factor_val.to_frame('f1')\n"
+                "factor_val.to_hdf('result.h5', key='data', mode='w')\n"
+            ),
+        })
+        msg, df = ws.execute()
+        assert isinstance(msg, str)
+        assert df is not None
+
+    def test_execute_succeeds_and_returns_data(self):
+        from rdagent.components.coder.factor_coder.factor import FactorFBWorkspace, FactorTask
+        t = FactorTask("fl1", "desc", "formula")
+        ws = FactorFBWorkspace(target_task=t)
+        ws.inject_files(**{
+            "factor.py": (
+                "import pandas as pd\n"
+                "data = pd.read_hdf('intraday_pv.h5', key='data')\n"
+                "factor_val = data['$close'].pct_change().to_frame('fl1')\n"
+                "factor_val.to_hdf('result.h5', key='data', mode='w')\n"
+            ),
+        })
+        msg, df = ws.execute()
+        assert FactorFBWorkspace.FB_EXEC_SUCCESS in msg
+        assert FactorFBWorkspace.FB_OUTPUT_FILE_FOUND in msg
+        assert df is not None
+
+
+# =============================================================================
+# FactorEvaluatorForCoder (partial integration)
+# =============================================================================
+
+
+class TestFactorEvaluatorForCoder:
+    def test_init_creates_sub_evaluators(self):
+        from rdagent.components.coder.factor_coder.evaluators import FactorEvaluatorForCoder
+        mock_scen = MagicMock()
+        eva = FactorEvaluatorForCoder(scen=mock_scen)
+        assert eva.value_evaluator is not None
+        assert eva.code_evaluator is not None
+        assert eva.final_decision_evaluator is not None
+
+    def test_evaluate_with_none_implementation(self):
+        from rdagent.components.coder.factor_coder.evaluators import FactorEvaluatorForCoder
+        eva = FactorEvaluatorForCoder(scen=MagicMock())
+        assert eva.evaluate(target_task=MagicMock(), implementation=None) is None
+
+    def test_evaluate_returns_queried_knowledge_if_present(self):
+        from rdagent.components.coder.factor_coder.evaluators import FactorEvaluatorForCoder
+        from rdagent.components.coder.factor_coder.factor import FactorTask
+
+        eva = FactorEvaluatorForCoder(scen=MagicMock())
+
+        t = FactorTask("f1", "desc", "formula")
+        qk = MagicMock()
+        qk.success_task_to_knowledge_dict = {"info_f1": MagicMock(feedback="cached_fb")}
+        t.get_task_information = MagicMock(return_value="info_f1")
+        qk.failed_task_info_set = set()
+
+        fb = eva.evaluate(target_task=t, implementation=MagicMock(), queried_knowledge=qk)
+        assert fb == "cached_fb"  # returned from cache
+
+    def test_evaluate_skips_failed_task(self):
+        from rdagent.components.coder.factor_coder.evaluators import FactorEvaluatorForCoder
+        from rdagent.components.coder.factor_coder.factor import FactorTask
+
+        eva = FactorEvaluatorForCoder(scen=MagicMock())
+
+        t = FactorTask("f1", "desc", "formula")
+        qk = MagicMock()
+        qk.success_task_to_knowledge_dict = {}
+        t.get_task_information = MagicMock(return_value="info_f1")
+        qk.failed_task_info_set = {"info_f1"}
+
+        fb = eva.evaluate(target_task=t, implementation=MagicMock(), queried_knowledge=qk)
+        assert fb.final_decision is False
+        assert "failed too many times" in fb.execution_feedback
+
+
+# =============================================================================
+# FactorEvaluator (eva_utils) — constructors and identity
+# =============================================================================
+
+
+class TestFactorEvaluatorsInit:
+    def test_factor_inf_evaluator_init(self):
+        from rdagent.components.coder.factor_coder.eva_utils import FactorInfEvaluator
+        eva = FactorInfEvaluator()
+        assert str(eva) == "FactorInfEvaluator"
+
+    def test_factor_single_column_evaluator_init(self):
+        from rdagent.components.coder.factor_coder.eva_utils import FactorSingleColumnEvaluator
+        eva = FactorSingleColumnEvaluator()
+        assert str(eva) == "FactorSingleColumnEvaluator"
+
+    def test_factor_output_format_evaluator_init(self):
+        from rdagent.components.coder.factor_coder.eva_utils import FactorOutputFormatEvaluator
+        eva = FactorOutputFormatEvaluator()
+        assert str(eva) == "FactorOutputFormatEvaluator"
+
+    def test_factor_missing_values_evaluator_init(self):
+        from rdagent.components.coder.factor_coder.eva_utils import FactorMissingValuesEvaluator
+        eva = FactorMissingValuesEvaluator()
+        assert str(eva) == "FactorMissingValuesEvaluator"
+
+    def test_factor_correlation_evaluator_init(self):
+        from rdagent.components.coder.factor_coder.eva_utils import FactorCorrelationEvaluator
+        eva = FactorCorrelationEvaluator(hard_check=True)
+        assert eva.hard_check is True
+        assert str(eva) == "FactorCorrelationEvaluator"
+
+    def test_factor_value_evaluator_init(self):
+        from rdagent.components.coder.factor_coder.eva_utils import FactorValueEvaluator
+        mock_scen = MagicMock()
+        eva = FactorValueEvaluator(mock_scen)
+        assert eva.scen is mock_scen

test/qlib/test_factor_eval_bugs.py

@@ -0,0 +1,241 @@
+"""Tests for bugs found in the factor evaluation pipeline."""
+
+import sys
+from pathlib import Path
+from unittest.mock import MagicMock, patch
+
+import numpy as np
+import pandas as pd
+import pytest
+
+# Project root
+PROJECT_ROOT = Path(__file__).parent.parent.parent
+sys.path.insert(0, str(PROJECT_ROOT))
+
+
+# =============================================================================
+# Bug 1: Missing `import sys` in _save_factor_values (factor_runner.py:968)
+# =============================================================================
+
+class TestSaveFactorValuesMissingSysImport:
+    """Verify that _save_factor_values has `import sys` — uses sys.executable at line 968."""
+
+    def test_save_factor_values_has_sys_import(self):
+        import inspect
+
+        from rdagent.scenarios.qlib.developer.factor_runner import QlibFactorRunner
+
+        source = inspect.getsource(QlibFactorRunner._save_factor_values)
+
+        assert "import sys" in source, (
+            "BUG: _save_factor_values calls sys.executable but does not import sys. "
+            "This causes a NameError at runtime, silently swallowed by the try/except."
+        )
+
+    def test_save_factor_values_nameerror_when_called(self):
+        """Simulate calling _save_factor_values without sys available."""
+        from rdagent.scenarios.qlib.developer.factor_runner import QlibFactorRunner
+
+        runner = QlibFactorRunner.__new__(QlibFactorRunner)
+
+        mock_exp = MagicMock()
+        mock_exp.sub_workspace_list = []
+        mock_exp.experiment_workspace.workspace_path = None
+
+        # This should NOT raise NameError for 'sys' — if it does, the bug is present
+        try:
+            runner._save_factor_values("TestFactor", mock_exp)
+        except NameError as e:
+            if "sys" in str(e):
+                pytest.fail(
+                    "BUG CONFIRMED: _save_factor_values raises NameError because "
+                    "'sys' is not imported. The factor values parquet is never saved."
+                )
+            raise
+
+
+# =============================================================================
+# Bug 2: `acc_rate` undefined in FactorEqualValueRatioEvaluator (eva_utils.py:335-346)
+# =============================================================================
+
+class TestEqualValueRatioAccRateUndefined:
+    """Verify FactorEqualValueRatioEvaluator handles shape-mismatch correctly."""
+
+    def test_acc_rate_undefined_after_except(self):
+        """If gen_df.sub(gt_df) raises, acc_rate should still be defined (default -1)."""
+        from rdagent.components.coder.factor_coder.eva_utils import FactorEqualValueRatioEvaluator
+
+        evaluator = FactorEqualValueRatioEvaluator()
+
+        # Simulate the case where _get_df returns None for gt_df, which causes
+        # gen_df.sub(None) to raise AttributeError. The except clause must not
+        # reference an undefined acc_rate variable.
+        gen_df = pd.DataFrame({"x": [1.0, 2.0, 3.0]}, index=[0, 1, 2])
+
+        gt_ws = MagicMock()
+        imp_ws = MagicMock()
+
+        gt_ws.execute.return_value = ("", None)  # _get_df will set gt_df = None
+        imp_ws.execute.return_value = ("", gen_df)
+
+        # Should NOT raise NameError
+        try:
+            result = evaluator.evaluate(imp_ws, gt_ws)
+            assert isinstance(result, tuple)
+            assert len(result) == 2
+            feedback, metric = result
+            assert metric == -1, f"Expected -1 (fallback), got {metric}"
+        except NameError as e:
+            if "acc_rate" in str(e):
+                pytest.fail(
+                    "BUG CONFIRMED: FactorEqualValueRatioEvaluator references 'acc_rate' "
+                    "which is undefined when gen_df.sub(gt_df) raises an exception."
+                )
+            raise
+
+    def test_acc_rate_defined_when_shapes_match(self):
+        """Normal case: same shapes — acc_rate should be defined and returned."""
+        from rdagent.components.coder.factor_coder.eva_utils import FactorEqualValueRatioEvaluator
+
+        evaluator = FactorEqualValueRatioEvaluator()
+
+        gt_ws = MagicMock()
+        imp_ws = MagicMock()
+
+        gen_df = pd.DataFrame({"x": [1.0, 2.0, 3.0]}, index=[0, 1, 2])
+        gt_df = pd.DataFrame({"y": [1.0, 2.0, 3.0]}, index=[0, 1, 2])
+
+        gt_ws.execute.return_value = ("", gt_df)
+        imp_ws.execute.return_value = ("", gen_df)
+
+        result = evaluator.evaluate(imp_ws, gt_ws)
+        assert isinstance(result, tuple)
+        assert len(result) == 2
+        feedback, metric = result
+        # When values match within tolerance, metric should be a float near 1.0
+        assert isinstance(metric, float) or isinstance(metric, (int, np.integer))
+        assert metric >= 0
+
+
+# =============================================================================
+# Bug 3: Annualization factor hardcoded in _evaluate_factor_directly (factor_runner.py:553)
+# =============================================================================
+
+class TestAnnualizationFactorInDirectEval:
+    """Verify direct evaluation uses correct annualization with forward_return_bars."""
+
+    def test_annualization_factor_uses_forward_bars(self):
+        """The direct eval method hardcodes 96 instead of using forward_return_bars param."""
+        import inspect
+
+        from rdagent.scenarios.qlib.developer.factor_runner import QlibFactorRunner
+
+        source = inspect.getsource(QlibFactorRunner._evaluate_factor_directly)
+
+        # Check that the method uses `np.sqrt(252 * 1440 / 96)` which hardcodes 96
+        # This should ideally be parameterized or at least consistent with the
+        # forward return calculation at line ~530 which also uses 96.
+        assert "np.sqrt(252 * 1440 / 96)" in source or "np.sqrt(252*1440/96)" in source, (
+            "The annualization factor in _evaluate_factor_directly should match "
+            "the forward_return_bars used for computing forward returns."
+        )
+
+    def test_ann_factor_is_consistent_with_forward_ret(self):
+        """Verify both the forward return shift and annualization use 96 bars."""
+        import inspect
+
+        from rdagent.scenarios.qlib.developer.factor_runner import QlibFactorRunner
+
+        source = inspect.getsource(QlibFactorRunner._evaluate_factor_directly)
+
+        # forward return uses `.shift(-96)` at line ~530
+        assert '.shift(-96)' in source, "Forward return shift should use 96 bars (1 day)"
+
+        # annualization should also use 96
+        assert '1440 / 96' in source, (
+            "Annualization factor should use the same number (96) as the forward return shift"
+        )
+
+
+# =============================================================================
+# Bug 4: _fix_inf_nan_handling inserts code before .dropna() or .to_hdf() in wrong context
+# =============================================================================
+
+class TestInfNanHandlingInsertion:
+    """Verify inf/nan auto-fixer doesn't insert code in the wrong context."""
+
+    def test_no_insertion_before_dropna_when_no_column_found(self):
+        from rdagent.components.coder.factor_coder.auto_fixer import FactorAutoFixer
+
+        fixer = FactorAutoFixer()
+
+        # Code where the LAST assignment before .dropna() is NOT a df['col'] = pattern
+        # but dropna() still exists (e.g., on a temporary variable)
+        code = (
+            "def calc():\n"
+            "    df = pd.read_hdf('data.h5', key='data')\n"
+            "    temp = df['$close'].diff()\n"
+            "    temp = temp.dropna()\n"
+            "    df['result'] = temp * 2\n"
+            "    result = df[['result']]\n"
+        )
+
+        result = fixer.fix(code)
+
+        # The code should still be valid (no syntax error from misplaced insertion)
+        import ast
+        try:
+            ast.parse(result)
+        except SyntaxError as e:
+            pytest.fail(f"Auto-fixer produced invalid Python code: {e}")
+
+    def test_inf_handling_inserted_before_result_assignment(self):
+        from rdagent.components.coder.factor_coder.auto_fixer import FactorAutoFixer
+
+        fixer = FactorAutoFixer()
+
+        code = (
+            "def calc():\n"
+            "    df = pd.read_hdf('data.h5', key='data')\n"
+            "    df['myfactor'] = df['$close'] / df['sigma_60bar']\n"
+            "    df['myfactor'] = df['myfactor'] / df['sigma_5bar']\n"
+            "    result = df[['myfactor']]\n"
+        )
+
+        result = fixer.fix(code)
+
+        # Should have added inf handling before the result = df[[...]] line
+        # but not broken syntax
+        import ast
+        try:
+            ast.parse(result)
+        except SyntaxError as e:
+            pytest.fail(f"Auto-fixer produced invalid Python code: {e}")
+
+        assert "replace([np.inf, -np.inf]" in result
+
+
+# =============================================================================
+# Bug 5: scan_factors reads factor_code twice (predix_full_eval.py:174 + 195)
+# =============================================================================
+
+class TestScanFactorsDoubleRead:
+    """Verify scan_factors doesn't wastefully read factor file twice."""
+
+    def test_factor_code_read_only_when_needed(self):
+        """Confirm the scan_factors double-read behavior (line 174+195)."""
+        import inspect
+        from scripts import predix_full_eval
+
+        source = inspect.getsource(predix_full_eval.scan_factors)
+
+        # Count occurrences of `.read_text()`
+        count = source.count(".read_text()")
+        # Expected: at least 2 (line 174 in fallback, line 195 in FactorInfo)
+        # Bug: if factor_name comes from result.h5 (line 168-170), then line 174
+        # is skipped, but line 195 always reads again — that's one wasted read.
+        assert count == 2, (
+            f"scan_factors has {count} read_text() calls. "
+            "Expected exactly 2 (one for name extraction, one for FactorInfo). "
+            "Consider caching to avoid double reads."
+        )

test/qlib/test_model_coder.py

@@ -0,0 +1,182 @@
+"""Tests for model_coder — ModelTask, shape/value evaluators, config."""
+
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+from unittest.mock import MagicMock
+
+import numpy as np
+import pytest
+
+PROJECT_ROOT = Path(__file__).parent.parent.parent
+sys.path.insert(0, str(PROJECT_ROOT))
+
+
+# =============================================================================
+# ModelTask
+# =============================================================================
+
+
+class TestModelTask:
+    def test_construction_fields(self):
+        from rdagent.components.coder.model_coder.model import ModelTask
+        t = ModelTask(
+            name="m1",
+            description="desc",
+            architecture="LSTM",
+            hyperparameters={"lr": 0.001},
+            training_hyperparameters={"epochs": 10},
+            formulation="y = f(x)",
+            variables={"x": "feature"},
+            model_type="TimeSeries",
+        )
+        assert t.name == "m1"
+        assert t.description == "desc"
+        assert t.architecture == "LSTM"
+        assert t.hyperparameters == {"lr": 0.001}
+        assert t.training_hyperparameters == {"epochs": 10}
+        assert t.formulation == "y = f(x)"
+        assert t.variables == {"x": "feature"}
+        assert t.model_type == "TimeSeries"
+        assert t.base_code is None
+
+    def test_get_task_information(self):
+        from rdagent.components.coder.model_coder.model import ModelTask
+        t = ModelTask(
+            name="m1", description="desc", architecture="LSTM",
+            hyperparameters={}, training_hyperparameters={},
+            model_type="Tabular",
+        )
+        info = t.get_task_information()
+        assert "name: m1" in info
+        assert "architecture: LSTM" in info
+        assert "model_type: Tabular" in info
+
+    def test_get_task_information_with_optional_fields(self):
+        from rdagent.components.coder.model_coder.model import ModelTask
+        t = ModelTask(
+            name="m2", description="d2", architecture="GRU",
+            hyperparameters={}, training_hyperparameters={},
+            formulation="f1", variables={"v": 1}, model_type="Graph",
+        )
+        info = t.get_task_information()
+        assert "formulation: f1" in info
+        assert "variables: {'v': 1}" in info
+
+    def test_get_task_brief_information(self):
+        from rdagent.components.coder.model_coder.model import ModelTask
+        t = ModelTask(
+            name="m1", description="desc", architecture="LSTM",
+            hyperparameters={"lr": 0.01}, training_hyperparameters={"epochs": 5},
+        )
+        info = t.get_task_brief_information()
+        assert "name: m1" in info
+        assert "architecture: LSTM" in info
+        assert "hyperparameters" in info
+
+    def test_from_dict(self):
+        from rdagent.components.coder.model_coder.model import ModelTask
+        d = {
+            "name": "m3", "description": "d3", "architecture": "TCN",
+            "hyperparameters": {}, "training_hyperparameters": {},
+        }
+        t = ModelTask.from_dict(d)
+        assert t.name == "m3"
+
+    def test_repr(self):
+        from rdagent.components.coder.model_coder.model import ModelTask
+        t = ModelTask(
+            name="mymodel", description="d", architecture="LSTM",
+            hyperparameters={}, training_hyperparameters={},
+        )
+        assert "ModelTask" in repr(t)
+        assert "mymodel" in repr(t)
+
+
+# =============================================================================
+# Shape/Value evaluators (eva_utils)
+# =============================================================================
+
+
+class TestShapeEvaluator:
+    def test_correct_shape(self):
+        from rdagent.components.coder.model_coder.eva_utils import shape_evaluator
+        msg, ok = shape_evaluator(np.ones((32, 10)), target_shape=(32, 10))
+        assert ok is True
+        assert "correct" in msg.lower()
+
+    def test_incorrect_shape(self):
+        from rdagent.components.coder.model_coder.eva_utils import shape_evaluator
+        msg, ok = shape_evaluator(np.ones((32, 5)), target_shape=(32, 10))
+        assert ok is False
+        assert "incorrect" in msg.lower()
+
+    def test_none_prediction(self):
+        from rdagent.components.coder.model_coder.eva_utils import shape_evaluator
+        msg, ok = shape_evaluator(None, target_shape=(32, 10))
+        assert ok is False
+
+    def test_none_target_shape(self):
+        from rdagent.components.coder.model_coder.eva_utils import shape_evaluator
+        msg, ok = shape_evaluator(np.ones((3,)), target_shape=None)
+        assert ok is False
+
+    def test_float_array(self):
+        from rdagent.components.coder.model_coder.eva_utils import shape_evaluator
+        msg, ok = shape_evaluator(np.array([1.0, 2.0]), target_shape=(2,))
+        assert ok is True
+
+
+class TestValueEvaluator:
+    def test_none_prediction(self):
+        from rdagent.components.coder.model_coder.eva_utils import value_evaluator
+        msg, ok = value_evaluator(None, np.ones((3,)))
+        assert ok is False
+
+    def test_none_target(self):
+        from rdagent.components.coder.model_coder.eva_utils import value_evaluator
+        msg, ok = value_evaluator(np.ones((3,)), None)
+        assert ok is False
+
+    def test_small_difference_passes(self):
+        from rdagent.components.coder.model_coder.eva_utils import value_evaluator
+        msg, ok = value_evaluator(
+            np.array([1.0, 2.0, 3.0]),
+            np.array([1.0, 2.0, 3.01]),
+        )
+        assert bool(ok) is True  # diff < 0.1
+
+    def test_large_difference_fails(self):
+        from rdagent.components.coder.model_coder.eva_utils import value_evaluator
+        msg, ok = value_evaluator(
+            np.array([1.0, 2.0]),
+            np.array([10.0, 20.0]),
+        )
+        assert bool(ok) is False  # diff > 0.1
+
+
+# =============================================================================
+# ModelCoSTEERSettings
+# =============================================================================
+
+
+class TestModelCoSTEERSettings:
+    def test_default_env_type(self):
+        from rdagent.components.coder.model_coder.conf import ModelCoSTEERSettings
+        s = ModelCoSTEERSettings()
+        assert s.env_type == "conda"
+
+    def test_singleton(self):
+        from rdagent.components.coder.model_coder.conf import MODEL_COSTEER_SETTINGS
+        from rdagent.components.coder.model_coder.conf import ModelCoSTEERSettings
+        assert isinstance(MODEL_COSTEER_SETTINGS, ModelCoSTEERSettings)
+
+    def test_get_model_env_runs(self):
+        from rdagent.components.coder.model_coder.conf import get_model_env
+        # May succeed (conda available) or fail — either way, test the code path
+        try:
+            env = get_model_env()
+            assert env is not None
+        except Exception:
+            pass  # expected if docker/conda not available

test/qlib/test_qlib_pipeline.py

@@ -0,0 +1,186 @@
+"""Tests for qlib pipeline — feedback, bandit, quant_loop_factory."""
+
+from __future__ import annotations
+
+import sys
+import tempfile
+from pathlib import Path
+from unittest.mock import MagicMock, patch
+
+import numpy as np
+import pandas as pd
+import pytest
+
+PROJECT_ROOT = Path(__file__).parent.parent.parent
+sys.path.insert(0, str(PROJECT_ROOT))
+
+
+# =============================================================================
+# process_results (feedback.py)
+# =============================================================================
+
+
+class TestProcessResults:
+    def test_process_results_handles_named_series(self):
+        """process_results renames column "0" to "Current Result" — this works
+        when the Series name is '0' (string), which matches the rename dict."""
+        from rdagent.scenarios.qlib.developer.feedback import process_results
+        import pandas as pd
+
+        # process_results expects the Series to produce a DataFrame column named "0" (string)
+        # This happens when the Series has name '0'
+        current = pd.Series(
+            {"IC": 0.05, "1day.excess_return_with_cost.annualized_return": 0.12,
+             "1day.excess_return_with_cost.max_drawdown": -0.08},
+            name="0",
+        )
+        sota = pd.Series(
+            {"IC": 0.03, "1day.excess_return_with_cost.annualized_return": 0.10,
+             "1day.excess_return_with_cost.max_drawdown": -0.05},
+            name="0",
+        )
+
+        result = process_results(current, sota)
+        assert "IC of Current Result is" in result
+        assert "of SOTA Result is" in result
+
+    def test_raises_on_missing_metrics(self):
+        from rdagent.scenarios.qlib.developer.feedback import process_results
+
+        current = pd.Series({"IC": 0.05})
+        sota = pd.Series({"IC": 0.03})
+        with pytest.raises(KeyError):
+            process_results(current, sota)
+
+
+# =============================================================================
+# bandit.py — Metrics and extract_metrics_from_experiment
+# =============================================================================
+
+
+class TestBanditMetrics:
+    def test_default_values_are_zero(self):
+        from rdagent.scenarios.qlib.proposal.bandit import Metrics
+        m = Metrics()
+        assert m.ic == 0.0
+        assert m.sharpe == 0.0
+        assert m.mdd == 0.0
+
+    def test_as_vector_length(self):
+        from rdagent.scenarios.qlib.proposal.bandit import Metrics
+        m = Metrics(ic=0.1, sharpe=1.5)
+        v = m.as_vector()
+        assert len(v) == 8
+        assert v[0] == 0.1
+        assert v[7] == 1.5
+
+    def test_mdd_negated_in_vector(self):
+        from rdagent.scenarios.qlib.proposal.bandit import Metrics
+        m = Metrics(mdd=0.15)
+        v = m.as_vector()
+        assert v[6] == -0.15  # -self.mdd
+
+    def test_extract_metrics_from_experiment(self):
+        from rdagent.scenarios.qlib.proposal.bandit import extract_metrics_from_experiment
+
+        mock_exp = MagicMock()
+        mock_exp.result = {
+            "IC": 0.04,
+            "ICIR": 0.5,
+            "Rank IC": 0.03,
+            "Rank ICIR": 0.4,
+            "1day.excess_return_with_cost.annualized_return ": 0.10,
+            "1day.excess_return_with_cost.information_ratio": 0.6,
+            "1day.excess_return_with_cost.max_drawdown": -0.12,
+        }
+        m = extract_metrics_from_experiment(mock_exp)
+        assert m.ic == 0.04
+        assert m.rank_ic == 0.03
+        assert m.mdd == -0.12
+
+    def test_extract_metrics_returns_default_on_error(self):
+        from rdagent.scenarios.qlib.proposal.bandit import extract_metrics_from_experiment
+
+        mock_exp = MagicMock()
+        mock_exp.result = None  # Will cause AttributeError
+        m = extract_metrics_from_experiment(mock_exp)
+        assert m.ic == 0.0
+        assert m.sharpe == 0.0
+
+    def test_sharpe_computation(self):
+        from rdagent.scenarios.qlib.proposal.bandit import extract_metrics_from_experiment
+
+        mock_exp = MagicMock()
+        mock_exp.result = {
+            "IC": 0.0, "ICIR": 0.0, "Rank IC": 0.0, "Rank ICIR": 0.0,
+            "1day.excess_return_with_cost.annualized_return ": 0.15,
+            "1day.excess_return_with_cost.information_ratio": 0.0,
+            "1day.excess_return_with_cost.max_drawdown": -0.10,
+        }
+        m = extract_metrics_from_experiment(mock_exp)
+        assert m.sharpe == pytest.approx(1.5)  # 0.15 / 0.10
+
+
+# =============================================================================
+# LinearThompsonTwoArm
+# =============================================================================
+
+
+class TestLinearThompsonTwoArm:
+    def test_initialization(self):
+        from rdagent.scenarios.qlib.proposal.bandit import LinearThompsonTwoArm
+        bandit = LinearThompsonTwoArm(dim=5)
+        assert bandit.dim == 5
+        assert bandit.noise_var == 1.0
+        assert bandit.mean["factor"].shape == (5,)
+        assert bandit.mean["model"].shape == (5,)
+        assert bandit.precision["factor"].shape == (5, 5)
+
+    def test_sample_reward_returns_float(self):
+        from rdagent.scenarios.qlib.proposal.bandit import LinearThompsonTwoArm
+        bandit = LinearThompsonTwoArm(dim=3)
+        x = np.ones(3)
+        reward = bandit.sample_reward("factor", x)
+        assert isinstance(reward, float)
+
+    def test_arms_are_initialized_identically(self):
+        from rdagent.scenarios.qlib.proposal.bandit import LinearThompsonTwoArm
+        bandit = LinearThompsonTwoArm(dim=4)
+        assert np.array_equal(bandit.mean["factor"], bandit.mean["model"])
+        assert np.array_equal(bandit.precision["factor"], bandit.precision["model"])
+
+
+# =============================================================================
+# quant_loop_factory.py
+# =============================================================================
+
+
+class TestHasLocalComponents:
+    def test_returns_bool(self):
+        from rdagent.scenarios.qlib.quant_loop_factory import has_local_components
+        result = has_local_components()
+        assert isinstance(result, bool)
+
+    def test_returns_false_with_no_local_dir(self, monkeypatch):
+        from rdagent.scenarios.qlib import quant_loop_factory
+        monkeypatch.setattr(quant_loop_factory.Path, "exists", lambda self: False)
+        assert quant_loop_factory.has_local_components() is False
+
+
+class TestCountValidFactors:
+    def test_returns_zero_when_no_dir(self):
+        from rdagent.scenarios.qlib.quant_loop_factory import count_valid_factors
+        with patch("rdagent.scenarios.qlib.quant_loop_factory.Path.exists", return_value=False):
+            assert count_valid_factors() == 0
+
+    def test_returns_int(self):
+        from rdagent.scenarios.qlib.quant_loop_factory import count_valid_factors
+        result = count_valid_factors()
+        assert isinstance(result, int)
+        assert result >= 0
+
+
+class TestAdvancedLoopThreshold:
+    def test_constant_is_defined(self):
+        from rdagent.scenarios.qlib.quant_loop_factory import ADVANCED_LOOP_FACTOR_THRESHOLD
+        assert ADVANCED_LOOP_FACTOR_THRESHOLD == 5000

test/utils/test_misc.py

@@ -1,8 +1,10 @@
+import tempfile
 import unittest
+from pathlib import Path
 
 import pytest
 
-from rdagent.core.utils import SingletonBaseClass
+from rdagent.core.utils import SingletonBaseClass, import_class, safe_resolve_path
 
 
 class A(SingletonBaseClass):
@@ -70,5 +72,76 @@ class MiscTest(unittest.TestCase):
         # print(a1.kwargs)  # a1 will be changed.
 
 
+class TestSafeResolvePath:
+    """Tests for safe_resolve_path — path traversal prevention."""
+
+    def test_inside_root_returns_absolute(self):
+        with tempfile.TemporaryDirectory() as tmpdir:
+            root = Path(tmpdir)
+            result = safe_resolve_path(root / "subdir" / "file.txt", safe_root=root)
+            assert result.is_absolute()
+            assert str(result).startswith(str(root.resolve()))
+
+    def test_no_safe_root_just_resolves(self):
+        result = safe_resolve_path(Path("/tmp/nonexistent_test"), safe_root=None)
+        assert result.is_absolute()
+
+    def test_path_traversal_raises(self):
+        with tempfile.TemporaryDirectory() as tmpdir:
+            root = Path(tmpdir)
+            with pytest.raises(ValueError, match="outside allowed root"):
+                safe_resolve_path(root / ".." / "etc" / "passwd", safe_root=root)
+
+    def test_symlink_outside_root_raises(self):
+        with tempfile.TemporaryDirectory() as tmpdir:
+            root = Path(tmpdir)
+            inside = root / "inside"
+            inside.mkdir()
+            link = inside / "escape"
+            link.symlink_to("/etc/passwd")
+            with pytest.raises(ValueError, match="outside allowed root"):
+                safe_resolve_path(link, safe_root=root)
+
+    def test_root_itself_is_valid(self):
+        with tempfile.TemporaryDirectory() as tmpdir:
+            root = Path(tmpdir)
+            result = safe_resolve_path(root, safe_root=root)
+            assert result == root.resolve()
+
+    def test_expanduser_resolves_home(self):
+        result = safe_resolve_path(Path("~/nonexistent_test"), safe_root=None)
+        assert str(result).startswith(str(Path.home()))
+
+
+class TestImportClass:
+    """Tests for import_class — dynamic class loading."""
+
+    def test_valid_class_import(self):
+        cls = import_class("pathlib.Path")
+        assert cls is Path
+
+    def test_builtin_class_import(self):
+        cls = import_class("collections.OrderedDict")
+        from collections import OrderedDict
+        assert cls is OrderedDict
+
+    def test_invalid_module_raises_import_error(self):
+        with pytest.raises(ImportError, match="Module not found"):
+            import_class("nonexistent.module.ClassName")
+
+    def test_missing_class_raises_import_error(self):
+        with pytest.raises(ImportError, match="Class not found"):
+            import_class("pathlib.NonExistentClass")
+
+    def test_invalid_format_raises_import_error(self):
+        with pytest.raises(ImportError, match="Invalid class path"):
+            import_class("no_dots_at_all")
+
+    def test_pandas_class_import(self):
+        cls = import_class("pandas.DataFrame")
+        import pandas as pd
+        assert cls is pd.DataFrame
+
+
 if __name__ == "__main__":
     unittest.main()