claude-evolve 1.3.30 → 1.3.37

package/bin/claude-evolve-run

@@ -1,6 +1,6 @@
 #!/bin/bash
 
-set -e
+# Removed 'set -e' to prevent silent exits on CSV helper failures
 
 # Load configuration
 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
@@ -328,7 +328,7 @@ echo "[INFO] Copied parent algorithm to: $output_file"
 # Check for claude CLI
 claude_cmd="${CLAUDE_CMD:-claude}"
 if ! command -v "$claude_cmd" >/dev/null 2>&1; then
-  echo "[ERROR] Claude CLI not found. Please install claude-cli." >&2
+  echo "[ERROR] Claude CLI not found. Please install 'claude' CLI tool or set CLAUDE_CMD environment variable." >&2
   update_csv_row "$row_num" "" "failed"
   exit 1
 fi
@@ -471,12 +471,45 @@ echo "----------------------------------------"
 
 # Process results
 if [[ $eval_exit_code -eq 0 ]]; then
-  # Extract score from JSON (simple grep approach)
-  if score=$(echo "$eval_output" | grep -o '"score"[[:space:]]*:[[:space:]]*[0-9.]*' | cut -d: -f2 | tr -d ' '); then
+  # DEBUG: Show raw evaluator output
+  echo "[DEBUG] Raw evaluator output:"
+  echo "----------------------------------------"
+  echo "$eval_output"
+  echo "----------------------------------------"
+  
+  # Extract the last valid JSON line and score
+  last_json=$(echo "$eval_output" | grep '^{.*}$' | tail -1)
+  
+  if [[ -n "$last_json" ]]; then
+    echo "[DEBUG] Found JSON: $last_json"
+    
+    # Extract score from JSON
+    score=$(echo "$last_json" | python3 -c "
+import sys, json
+try:
+    data = json.loads(sys.stdin.read().strip())
+    if 'performance' in data:
+        print(data['performance'])
+    elif 'score' in data:
+        print(data['score'])
+    else:
+        sys.exit(1)
+except:
+    sys.exit(1)
+" 2>/dev/null)
+  fi
+  
+  if [[ -n "$score" ]]; then
+    echo "[DEBUG] Extracted score via 'score' field: '$score'"
     if [[ -n $score ]]; then
       # Check if score is 0 and mark as failed
       if (( $(echo "$score == 0" | bc -l) )); then
-        update_csv_row "$row_num" "$score" "failed"
+        # Update CSV with full JSON data
+        if [[ -n "$last_json" ]]; then
+          python3 "${SCRIPT_DIR}/../lib/csv_helper.py" update_with_json "${FULL_CSV_PATH}" "$id" "$last_json"
+        else
+          update_csv_row "$row_num" "$score" "failed"
+        fi
         echo "[INFO] ✗ Evaluation failed with score 0"
         echo "[INFO] Performance score: $score"
         if should_continue_after_failure; then
@@ -485,16 +518,35 @@ if [[ $eval_exit_code -eq 0 ]]; then
           exit 1
         fi
       else
-        update_csv_row "$row_num" "$score" "complete"
+        # Update CSV with full JSON data
+        if [[ -n "$last_json" ]]; then
+          echo "[DEBUG] Updating CSV with full JSON data (branch 1)"
+          echo "[DEBUG] CSV path: ${FULL_CSV_PATH}"
+          echo "[DEBUG] Target ID: $id (row $row_num)"
+          echo "[DEBUG] JSON data: $last_json"
+          echo "[DEBUG] Running CSV helper..."
+          if python3 "${SCRIPT_DIR}/../lib/csv_helper.py" update_with_json "${FULL_CSV_PATH}" "$id" "$last_json"; then
+            echo "[DEBUG] CSV helper succeeded"
+          else
+            echo "[DEBUG] CSV helper failed with exit code $?"
+          fi
+        else
+          update_csv_row "$row_num" "$score" "complete"
+        fi
         echo "[INFO] ✓ Evaluation completed successfully"
         echo "[INFO] Performance score: $score"
       fi
     else
-      # Try "performance" field
-      if score=$(echo "$eval_output" | grep -o '"performance"[[:space:]]*:[[:space:]]*[0-9.]*' | cut -d: -f2 | tr -d ' '); then
+      # Try "performance" field - get only the last occurrence
+      if score=$(echo "$eval_output" | grep -o '"performance"[[:space:]]*:[[:space:]]*[0-9.]*' | cut -d: -f2 | tr -d ' ' | tail -1); then
+        echo "[DEBUG] Extracted score via 'performance' field: '$score'"
         # Check if score is 0 and mark as failed
         if [ "$(echo "$score == 0" | bc -l)" = "1" ]; then
-          update_csv_row "$row_num" "$score" "failed"
+          if [[ -n "$last_json" ]]; then
+            python3 "${SCRIPT_DIR}/../lib/csv_helper.py" update_with_json "${FULL_CSV_PATH}" "$id" "$last_json"
+          else
+            update_csv_row "$row_num" "$score" "failed"
+          fi
           echo "[INFO] ✗ Evaluation failed with score 0"
           echo "[INFO] Performance score: $score"
           if should_continue_after_failure; then
@@ -503,7 +555,21 @@ if [[ $eval_exit_code -eq 0 ]]; then
             exit 1
           fi
         else
-          update_csv_row "$row_num" "$score" "complete"
+          # Update CSV with full JSON data
+          if [[ -n "$last_json" ]]; then
+            echo "[DEBUG] Updating CSV with full JSON data (branch 2)"
+            echo "[DEBUG] CSV path: ${FULL_CSV_PATH}"
+            echo "[DEBUG] Target ID: $id (row $row_num)"
+            echo "[DEBUG] JSON data: $last_json"
+            echo "[DEBUG] Running CSV helper..."
+            if python3 "${SCRIPT_DIR}/../lib/csv_helper.py" update_with_json "${FULL_CSV_PATH}" "$id" "$last_json"; then
+              echo "[DEBUG] CSV helper succeeded"
+            else
+              echo "[DEBUG] CSV helper failed with exit code $?"
+            fi
+          else
+            update_csv_row "$row_num" "$score" "complete"
+          fi
           echo "[INFO] ✓ Evaluation completed successfully"
           echo "[INFO] Performance score: $score"
         fi

package/bin/claude-evolve-run-clean

@@ -0,0 +1,94 @@
+#!/bin/bash
+
+# Simple, robust version - fail fast on malformed evaluator output
+
+# Extract the score parsing section only
+parse_evaluator_output() {
+  local eval_output="$1"
+  local score=""
+  
+  echo "[DEBUG] Raw evaluator output:"
+  echo "----------------------------------------"
+  echo "$eval_output"
+  echo "----------------------------------------"
+  
+  # Try to extract the last valid JSON line
+  last_json=$(echo "$eval_output" | grep '^{.*}$' | tail -1)
+  
+  if [[ -n "$last_json" ]]; then
+    echo "[DEBUG] Found JSON line: $last_json"
+    
+    # Parse with Python for robustness
+    score=$(echo "$last_json" | python3 -c "
+import sys, json
+try:
+    data = json.loads(sys.stdin.read().strip())
+    if 'performance' in data:
+        print(data['performance'])
+    elif 'score' in data:
+        print(data['score'])
+    else:
+        sys.exit(1)
+except:
+    sys.exit(1)
+" 2>/dev/null)
+    
+    if [[ -n "$score" ]]; then
+      echo "[DEBUG] Extracted score: $score"
+      
+      # Validate it's a clean number
+      if [[ "$score" =~ ^[0-9]+\.?[0-9]*$ ]]; then
+        echo "$score"
+        return 0
+      else
+        echo "[ERROR] Score is not a valid number: '$score'" >&2
+        return 1
+      fi
+    else
+      echo "[ERROR] Could not extract score from JSON" >&2
+      return 1
+    fi
+  else
+    echo "[ERROR] No valid JSON found in evaluator output" >&2
+    echo "[ERROR] Evaluator must output clean JSON with 'score' or 'performance' field" >&2
+    echo "[ERROR] Mixed logging/output is not allowed" >&2
+    return 1
+  fi
+}
+
+# Test with your sample
+test_output='Evaluating algorithm: /Users/willer/Documents/GitHub/trading-strategies/evolution-mats/evolution_gen01-001.py
+Copied /Users/willer/Documents/GitHub/trading-strategies/evolution-mats/evolution_gen01-001.py -> algorithm.py for evaluation
+Starting MATS evaluation...
+Running backtest from 2015-01-01 to 2025-06-01...
+
+Evaluation completed successfully!
+Performance: 1.0775
+Total Return: 2839.30%
+Sharpe: 1.310, Sortino: 1.661
+Max Drawdown: -20.87%
+Total Trades: 2604, Win Rate: 64.0%
+
+Full results JSON:
+{
+  "performance": 1.077506371224117,
+  "total_return": 28.392963641910857,
+  "yearly_return": 0.24689362518462943,
+  "sharpe": 1.3100255343786216,
+  "sortino": 1.6612254920789753,
+  "max_drawdown": -0.2086643676866254,
+  "volatility": 0.18846474263704896,
+  "total_trades": 2604,
+  "win_rate": 0.6401689708141322,
+  "profit_factor": 1.9441283380055399,
+  "final_value": 2939296.3641910856
+}
+Restored original algorithm.py
+{"performance": 1.077506371224117, "total_return": 28.392963641910857, "yearly_return": 0.24689362518462943, "sharpe": 1.3100255343786216, "sortino": 1.6612254920789753, "max_drawdown": -0.2086643676866254, "volatility": 0.18846474263704896, "total_trades": 2604, "win_rate": 0.6401689708141322, "profit_factor": 1.9441283380055399, "final_value": 2939296.3641910856}'
+
+echo "Testing parser..."
+if result=$(parse_evaluator_output "$test_output"); then
+  echo "SUCCESS: Got score: $result"
+else
+  echo "FAILED: Could not parse"
+fi

package/lib/csv_helper.py

@@ -60,6 +60,7 @@ def update_row_with_fields(headers: list[str], rows: list[list[str]], target_id:
         if row[0] == target_id:
             for field, value in fields.items():
                 if field in col_indices:
+                    print(f"[DEBUG] Updating field '{field}' with value: {repr(value)}", file=sys.stderr)
                     row[col_indices[field]] = str(value)
             break
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-evolve",
-  "version": "1.3.30",
+  "version": "1.3.37",
   "bin": {
     "claude-evolve": "./bin/claude-evolve",
     "claude-evolve-main": "./bin/claude-evolve-main",