claude-evolve 1.3.39 → 1.3.41

package/bin/claude-evolve-main

@@ -55,26 +55,30 @@ show_help() {
 claude-evolve - AI-powered algorithm evolution tool
 
 USAGE:
-    claude-evolve [--config=PATH] [COMMAND] [OPTIONS]
+    claude-evolve [--working-dir=PATH] [COMMAND] [OPTIONS]
 
 COMMANDS:
-    setup       Initialize evolution workspace
-    ideate      Generate new algorithm ideas
-    run         Execute evolution candidates
-    analyze     Analyze evolution results
-    help        Show this help message
+    setup              Initialize evolution workspace
+    ideate             Generate new algorithm ideas
+    run                Execute evolution candidates
+    analyze            Analyze evolution results
+    edit               Manage candidate statuses by generation
+    status             Show evolution progress and current leader
+    cleanup            Clean up unchanged algorithms and descendants
+    cleanup-duplicates Alias for cleanup (deprecated)
+    help               Show this help message
 
 GLOBAL OPTIONS:
-    --config=PATH  Use alternate config file (default: evolution/config.yaml)
-    -h, --help     Show help message
-    -v, --version  Show version
+    --working-dir=PATH  Use alternate working directory (default: evolution/)
+    -h, --help          Show help message
+    -v, --version       Show version
 
 EXAMPLES:
     claude-evolve setup
     claude-evolve ideate 5
     claude-evolve run --timeout 300
     claude-evolve analyze --open
-    claude-evolve --config=evolution-abc/config.yaml run
+    claude-evolve --working-dir=evolution-abc run
 
 For more information, visit: https://github.com/anthropics/claude-evolve
 EOF
@@ -91,9 +95,11 @@ show_menu() {
   echo "  2) ideate   - Generate new algorithm ideas"
   echo "  3) run      - Execute evolution candidates"
   echo "  4) analyze  - Analyze evolution results"
-  echo "  5) config   - Manage configuration settings"
-  echo "  6) help     - Show help message"
-  echo "  7) exit     - Exit"
+  echo "  5) edit     - Manage candidate statuses by generation"
+  echo "  6) status   - Show evolution progress and current leader"
+  echo "  7) config   - Manage configuration settings"
+  echo "  8) help     - Show help message"
+  echo "  9) exit     - Exit"
   echo
 
   # Show workspace status
@@ -105,19 +111,19 @@ show_menu() {
 }
 
 # Parse global options
-CONFIG_FILE=""
+WORKING_DIR=""
 while [[ $# -gt 0 ]] && [[ "$1" =~ ^-- ]]; do
   case "$1" in
-  --config)
+  --working-dir)
     if [[ -z ${2:-} ]]; then
-      echo -e "${RED}[ERROR] --config requires a file path${NC}" >&2
+      echo -e "${RED}[ERROR] --working-dir requires a directory path${NC}" >&2
       exit 1
     fi
-    CONFIG_FILE="$2"
+    WORKING_DIR="$2"
     shift 2
     ;;
-  --config=*)
-    CONFIG_FILE="${1#*=}"
+  --working-dir=*)
+    WORKING_DIR="${1#*=}"
     shift
     ;;
   *)
@@ -126,9 +132,11 @@ while [[ $# -gt 0 ]] && [[ "$1" =~ ^-- ]]; do
   esac
 done
 
-# Export CONFIG_FILE for subcommands
-if [[ -n $CONFIG_FILE ]]; then
-  export CLAUDE_EVOLVE_CONFIG="$CONFIG_FILE"
+# Export config file path for subcommands
+if [[ -n $WORKING_DIR ]]; then
+  # Remove trailing slash if present
+  WORKING_DIR="${WORKING_DIR%/}"
+  export CLAUDE_EVOLVE_CONFIG="$WORKING_DIR/config.yaml"
 fi
 
 # Check for updates (quick, non-blocking)
@@ -137,21 +145,23 @@ check_for_updates
 # Main logic
 if [[ $# -eq 0 ]]; then
   show_menu
-  read -r -p "Enter your choice (1-7): " choice
+  read -r -p "Enter your choice (1-9): " choice
 
   case $choice in
   1) exec "$SCRIPT_DIR/claude-evolve-setup" ;;
   2) exec "$SCRIPT_DIR/claude-evolve-ideate" ;;
-  3) exec "$SCRIPT_DIR/claude-evolve-run-unified" ;;
+  3) exec "$SCRIPT_DIR/claude-evolve-run" ;;
   4) exec "$SCRIPT_DIR/claude-evolve-analyze" ;;
-  5) exec "$SCRIPT_DIR/claude-evolve-config" ;;
-  6) show_help ;;
-  7)
+  5) exec "$SCRIPT_DIR/claude-evolve-edit" ;;
+  6) exec "$SCRIPT_DIR/claude-evolve-status" ;;
+  7) exec "$SCRIPT_DIR/claude-evolve-config" ;;
+  8) show_help ;;
+  9)
     echo "Goodbye!"
     exit 0
     ;;
   *)
-    echo -e "${RED}Invalid choice. Please select 1-7.${NC}"
+    echo -e "${RED}Invalid choice. Please select 1-9.${NC}"
     exit 1
     ;;
   esac
@@ -174,12 +184,24 @@ ideate)
   ;;
 run)
   shift
-  exec "$SCRIPT_DIR/claude-evolve-run-unified" "$@"
+  exec "$SCRIPT_DIR/claude-evolve-run" "$@"
   ;;
 analyze)
   shift
   exec "$SCRIPT_DIR/claude-evolve-analyze" "$@"
   ;;
+edit)
+  shift
+  exec "$SCRIPT_DIR/claude-evolve-edit" "$@"
+  ;;
+status)
+  shift
+  exec "$SCRIPT_DIR/claude-evolve-status" "$@"
+  ;;
+cleanup-duplicates|cleanup)
+  shift
+  exec "$SCRIPT_DIR/claude-evolve-cleanup" "$@"
+  ;;
 config)
   shift
   exec "$SCRIPT_DIR/claude-evolve-config" "$@"

package/bin/{claude-evolve-run-unified → claude-evolve-run}

@@ -236,12 +236,32 @@ cleanup_workers() {
       # Worker finished
       if wait "$pid" 2>/dev/null; then
         echo "[DISPATCHER] Worker $pid completed successfully"
+        consecutive_failures=0  # Reset counter on success
       else
         local exit_code=$?
         if [[ $exit_code -eq 2 ]]; then
           echo "[DISPATCHER] Worker $pid hit rate limit, will retry later"
+          # Rate limits don't count as consecutive failures
         else
           echo "[DISPATCHER] Worker $pid failed with exit code $exit_code"
+          ((consecutive_failures++))
+          
+          # Check if we've hit the failure limit
+          if [[ $consecutive_failures -ge $MAX_CONSECUTIVE_FAILURES ]]; then
+            echo "" >&2
+            echo "⚠️  EVOLUTION PAUSED: Multiple consecutive failures detected" >&2
+            echo "⚠️  $consecutive_failures consecutive worker failures - indicates systemic issues" >&2
+            echo "⚠️  Possible causes: Claude API issues, evaluator bugs, configuration problems" >&2
+            echo "⚠️  Check recent worker logs in logs/ directory for specific error details" >&2
+            echo "⚠️  Fix issues before restarting evolution" >&2
+            echo "" >&2
+            
+            # Shutdown all workers and exit
+            shutdown_workers
+            exit 1
+          fi
+          
+          echo "[DISPATCHER] Consecutive failures: $consecutive_failures/$MAX_CONSECUTIVE_FAILURES"
         fi
       fi
     fi
@@ -269,9 +289,26 @@ get_csv_stats() {
   total_rows=$(wc -l < "$csv_path" | tr -d '[:space:]')
   complete_count=$(grep ',complete' "$csv_path" 2>/dev/null | wc -l | tr -d '[:space:]')
   
-  # Count pending: empty status or "pending"
-  # Handle potential Windows line endings by stripping carriage returns
-  pending_count=$(awk -F, 'NR>1 {gsub(/\r/, "", $5); if($5=="" || $5=="pending") count++} END {print count+0}' "$csv_path")
+  # Count pending using same logic as find_next_pending_with_lock
+  # This includes rows with <5 fields AND rows with empty/pending status
+  pending_count=$("$PYTHON_CMD" -c "
+import csv
+import sys
+
+pending_count = 0
+with open('$csv_path', 'r') as f:
+    reader = csv.reader(f)
+    rows = list(reader)
+    
+    for i in range(1, len(rows)):
+        # Same logic as find_next_pending_with_lock
+        if len(rows[i]) < 5:
+            pending_count += 1
+        elif len(rows[i]) >= 5 and (rows[i][4] == 'pending' or rows[i][4] == ''):
+            pending_count += 1
+
+print(pending_count)
+")
   
   echo "$total_rows $complete_count $pending_count"
 }
@@ -279,6 +316,100 @@ get_csv_stats() {
 echo "[DISPATCHER] Starting unified evolution engine"
 echo "[DISPATCHER] Configuration: max_workers=$MAX_WORKERS, timeout=${timeout_seconds:-none}"
 
+# Validate CSV and clean up stuck statuses
+if [[ -f "$FULL_CSV_PATH" ]]; then
+  echo "[DISPATCHER] Validating CSV and cleaning up..."
+  if ! "$PYTHON_CMD" -c "
+import csv
+import sys
+
+csv_file = '$FULL_CSV_PATH'
+
+try:
+    # Read CSV - let Python's csv module handle all the complexity
+    with open(csv_file, 'r') as f:
+        reader = csv.reader(f)
+        rows = list(reader)
+    
+    if not rows:
+        print('[ERROR] CSV is empty')
+        sys.exit(1)
+    
+    # Basic sanity checks
+    header = rows[0]
+    num_fields = len(header)
+    
+    if len(rows) == 1:
+        print('[INFO] CSV has no data rows (only header)')
+    
+    # Clean up any stuck 'running' statuses
+    changed = 0
+    for i in range(1, len(rows)):
+        if len(rows[i]) > 4 and rows[i][4] == 'running':
+            rows[i][4] = ''
+            changed += 1
+    
+    if changed > 0:
+        # Write back
+        with open(csv_file + '.tmp', 'w', newline='') as f:
+            writer = csv.writer(f)
+            writer.writerows(rows)
+        import os
+        os.rename(csv_file + '.tmp', csv_file)
+        print(f'[INFO] Reset {changed} stuck running candidates to pending')
+    
+    # Count pending candidates
+    pending = 0
+    for i in range(1, len(rows)):
+        # Row with < 5 fields or empty/pending status in field 5
+        if len(rows[i]) < 5 or (len(rows[i]) >= 5 and rows[i][4] in ['', 'pending']):
+            pending += 1
+    
+    print(f'[INFO] CSV loaded: {len(rows)-1} total candidates, {pending} pending')
+    
+except csv.Error as e:
+    print(f'[ERROR] CSV parsing error: {e}')
+    print('[ERROR] The CSV file appears to be malformed')
+    sys.exit(1)
+except Exception as e:
+    print(f'[ERROR] Failed to read CSV: {e}')
+    sys.exit(1)
+"; then
+    echo "[ERROR] CSV validation failed. Please check the error message above."
+    exit 1
+  fi
+fi
+
+# Automatic cleanup detection - check for unchanged algorithms and warn user
+echo "[DISPATCHER] Checking for duplicate/unchanged algorithms..."
+cleanup_output=$("$SCRIPT_DIR/claude-evolve-cleanup" --dry-run 2>&1)
+
+# Check if cleanup found any issues (look for "UNCHANGED:" in output)
+if echo "$cleanup_output" | grep -q "📋 UNCHANGED:"; then
+  echo "⚠️  WARNING: Issues detected that may need cleanup:"
+  echo "$cleanup_output"
+  echo ""
+  echo "🔧 RECOMMENDATION: Run 'claude-evolve cleanup --force' to fix these issues before continuing"
+  echo "   This will delete unchanged algorithms and reset their descendants to pending status"
+  echo ""
+  echo "⏰ Continuing in 10 seconds (Ctrl+C to abort and run cleanup)..."
+  
+  # Give user time to read and potentially abort
+  for i in {10..1}; do
+    echo -n "  $i..."
+    sleep 1
+  done
+  echo ""
+  echo "🚀 Proceeding with evolution run..."
+  echo ""
+else
+  echo "[DISPATCHER] No cleanup issues detected - proceeding with run"
+fi
+
+# Consecutive failure tracking
+consecutive_failures=0
+MAX_CONSECUTIVE_FAILURES=5
+
 # Main dispatch loop
 while true; do
   # Clean up finished workers
@@ -337,7 +468,7 @@ while true; do
   done
   
   # Brief pause to avoid busy waiting
-  sleep 2
+  sleep 5
 done
 
 # Clean shutdown

package/bin/claude-evolve-status

@@ -0,0 +1,220 @@
+#!/bin/bash
+
+set -e
+
+# Load configuration
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+# shellcheck source=../lib/config.sh
+source "$SCRIPT_DIR/../lib/config.sh"
+
+# Use CLAUDE_EVOLVE_CONFIG if set, otherwise default
+if [[ -n ${CLAUDE_EVOLVE_CONFIG:-} ]]; then
+  load_config "$CLAUDE_EVOLVE_CONFIG"
+else
+  load_config
+fi
+
+# Validate configuration
+if ! validate_config; then
+  echo "[ERROR] Configuration validation failed" >&2
+  exit 1
+fi
+
+# Function to show help
+show_help() {
+  cat <<EOF
+claude-evolve status - Show evolution progress and current leader
+
+USAGE:
+  claude-evolve status [OPTIONS]
+
+OPTIONS:
+  --brief     Show only summary stats (no per-generation breakdown)
+  --winner    Show only the current best performer
+  --help      Show this help message
+
+DESCRIPTION:
+  Displays evolution progress by generation including:
+  - Candidate counts by status (pending, complete, failed, running)
+  - Current best performer across all generations
+  - Generation-by-generation breakdown
+
+EXAMPLES:
+  claude-evolve status          # Full status report
+  claude-evolve status --brief  # Just totals and winner
+  claude-evolve status --winner # Just the best performer
+EOF
+}
+
+# Parse arguments
+SHOW_BRIEF=false
+SHOW_WINNER_ONLY=false
+
+while [[ $# -gt 0 ]]; do
+  case $1 in
+  --brief)
+    SHOW_BRIEF=true
+    shift
+    ;;
+  --winner)
+    SHOW_WINNER_ONLY=true
+    shift
+    ;;
+  --help)
+    show_help
+    exit 0
+    ;;
+  *)
+    echo "[ERROR] Unknown option: $1" >&2
+    exit 1
+    ;;
+  esac
+done
+
+# Check if CSV exists
+if [[ ! -f "$FULL_CSV_PATH" ]]; then
+  echo "[ERROR] Evolution CSV not found: $FULL_CSV_PATH" >&2
+  echo "Run 'claude-evolve setup' first or navigate to the correct directory" >&2
+  exit 1
+fi
+
+# Determine what we're evolving based on paths
+EVOLUTION_CONTEXT=""
+if [[ -n "$EVOLUTION_DIR" ]]; then
+  # Get the evolution directory name (e.g., "evolution-atr" -> "ATR")
+  EVOLUTION_NAME=$(basename "$EVOLUTION_DIR")
+  EVOLUTION_CONTEXT="${EVOLUTION_NAME#evolution-}"
+  EVOLUTION_CONTEXT=$(echo "$EVOLUTION_CONTEXT" | tr '[:lower:]' '[:upper:]')
+fi
+
+# If we can't determine from evolution dir, try from algorithm path
+if [[ -z "$EVOLUTION_CONTEXT" && -n "$ALGORITHM_PATH" ]]; then
+  # Get parent directory name or algorithm file name
+  if [[ -f "$FULL_ALGORITHM_PATH" ]]; then
+    ALGO_NAME=$(basename "$FULL_ALGORITHM_PATH" .py)
+    EVOLUTION_CONTEXT="$ALGO_NAME"
+  fi
+fi
+
+# Default if we still can't determine
+if [[ -z "$EVOLUTION_CONTEXT" ]]; then
+  EVOLUTION_CONTEXT="Algorithm"
+fi
+
+# Main status reporting using Python
+"$PYTHON_CMD" -c "
+import csv
+import sys
+
+csv_file = '$FULL_CSV_PATH'
+show_brief = '$SHOW_BRIEF' == 'true'
+show_winner_only = '$SHOW_WINNER_ONLY' == 'true'
+evolution_context = '$EVOLUTION_CONTEXT'
+
+try:
+    with open(csv_file, 'r') as f:
+        reader = csv.reader(f)
+        rows = list(reader)
+    
+    if len(rows) <= 1:
+        print('No evolution candidates found')
+        sys.exit(0)
+    
+    header = rows[0]
+    
+    # Collect all candidates with scores and statuses
+    all_candidates = []
+    stats_by_gen = {}
+    total_stats = {'pending': 0, 'complete': 0, 'failed': 0, 'running': 0}
+    
+    for row in rows[1:]:
+        if len(row) >= 1 and row[0]:  # Must have an ID
+            candidate_id = row[0]
+            
+            # Extract generation (e.g., 'gen03' from 'gen03-001')
+            if '-' in candidate_id:
+                gen = candidate_id.split('-')[0]
+                
+                # Get status and performance
+                status = row[4] if len(row) > 4 and row[4] else 'pending'
+                performance = row[3] if len(row) > 3 and row[3] else ''
+                
+                # Track by generation
+                if gen not in stats_by_gen:
+                    stats_by_gen[gen] = {'pending': 0, 'complete': 0, 'failed': 0, 'running': 0}
+                
+                if status in stats_by_gen[gen]:
+                    stats_by_gen[gen][status] += 1
+                    total_stats[status] += 1
+                else:
+                    stats_by_gen[gen]['pending'] += 1
+                    total_stats['pending'] += 1
+                
+                # Collect for winner analysis (only completed with valid scores)
+                if status == 'complete' and performance:
+                    try:
+                        score = float(performance)
+                        description = row[2] if len(row) > 2 else 'No description'
+                        all_candidates.append((candidate_id, description, score))
+                    except ValueError:
+                        pass
+    
+    # Find the winner
+    winner = None
+    if all_candidates:
+        winner = max(all_candidates, key=lambda x: x[2])
+    
+    # Show winner only
+    if show_winner_only:
+        if winner:
+            print(f'🏆 CURRENT LEADER: {winner[0]} (score: {winner[2]:.4f})')
+            print(f'   {winner[1]}')
+        else:
+            print('No completed candidates found')
+        sys.exit(0)
+    
+    # Show header
+    print(f'🧬 Evolution Status Report - {evolution_context}')
+    print('=' * 50)
+    
+    # Show overall stats
+    total_candidates = sum(total_stats.values())
+    if total_candidates > 0:
+        print(f'📊 OVERALL: {total_candidates} total candidates')
+        print(f'   • {total_stats[\"pending\"]} pending')
+        print(f'   • {total_stats[\"complete\"]} complete') 
+        print(f'   • {total_stats[\"failed\"]} failed')
+        print(f'   • {total_stats[\"running\"]} running')
+        print()
+    
+    # Show current winner
+    if winner:
+        print(f'🏆 CURRENT LEADER: {winner[0]} (score: {winner[2]:.4f})')
+        print(f'   {winner[1]}')
+        print()
+    else:
+        print('🏆 CURRENT LEADER: None (no completed candidates)')
+        print()
+    
+    # Show per-generation breakdown (unless brief mode)
+    if not show_brief and stats_by_gen:
+        print('📈 BY GENERATION:')
+        for gen in sorted(stats_by_gen.keys()):
+            data = stats_by_gen[gen]
+            total = sum(data.values())
+            
+            # Find best performer in this generation
+            gen_candidates = [c for c in all_candidates if c[0].startswith(gen + '-')]
+            gen_best = max(gen_candidates, key=lambda x: x[2]) if gen_candidates else None
+            
+            status_str = f'{data[\"pending\"]}p {data[\"complete\"]}c {data[\"failed\"]}f {data[\"running\"]}r'
+            
+            if gen_best:
+                print(f'   {gen}: {total} total ({status_str}) - best: {gen_best[0]} ({gen_best[2]:.4f})')
+            else:
+                print(f'   {gen}: {total} total ({status_str}) - best: none')
+
+except Exception as e:
+    print(f'Error reading evolution status: {e}')
+    sys.exit(1)
+"