claude-evolve 1.3.14 → 1.3.16

package/README.md

@@ -12,7 +12,7 @@ Think of it like **genetic algorithms for code** - it handles the mutations and
 
 The system operates with specialized phases working together:
 
-- 🧠 **Ideation Phase**: Generates creative algorithm variations using Claude Opus
+- 🧠 **Ideation Phase**: Generates creative algorithm variations using codex o3-pro (if available) or Claude Opus
 - 🔬 **Development Phase**: Implements mutations using Claude Sonnet (with periodic Opus "megathinking")
 - 📊 **Evaluation Phase**: Tests performance against your custom evaluator
 - 📈 **Analysis Phase**: Tracks evolution progress and identifies top performers
@@ -180,6 +180,7 @@ Evolution experiments can fail for various reasons. The system tracks these fail
 - [Claude CLI](https://docs.anthropic.com/en/docs/claude-code) (`claude` command)
 
 ### Optional (but recommended)
+- [Codex CLI](https://github.com/aboutgaurav/codex) (`codex` command) - Uses o3-pro model for superior ideation when available
 - Scientific Python libraries (numpy, scipy, etc.) depending on your algorithms
 - Plotting libraries (matplotlib, plotly) for analyzing results
 

package/bin/claude-evolve-analyze

@@ -292,16 +292,18 @@ if command -v gnuplot >/dev/null 2>&1 && [[ $valid_performance_count -gt 0 ]]; t
   echo "# Row ID Performance Generation" >"$data_file"
   echo "# Generation AvgPerformance Color" >"$gen_avg_file"
 
-  # Get color by generation number (rotates through 5 colors)
+  # Get color by generation number (rotates through 7 colors)
   get_gen_color() {
     local gen_num="$1"
-    local color_index=$(( (gen_num - 1) % 5 + 1 ))
+    local color_index=$(( gen_num % 7 ))
     case $color_index in
-      1) echo "#1f77b4" ;;  # blue
-      2) echo "#ff7f0e" ;;  # orange  
-      3) echo "#2ca02c" ;;  # green
-      4) echo "#d62728" ;;  # red
-      5) echo "#9467bd" ;;  # purple
+      0) echo "#1f77b4" ;;  # blue
+      1) echo "#ff7f0e" ;;  # orange  
+      2) echo "#2ca02c" ;;  # green
+      3) echo "#d62728" ;;  # red
+      4) echo "#9467bd" ;;  # purple
+      5) echo "#8c564b" ;;  # brown
+      6) echo "#e377c2" ;;  # pink
     esac
   }
 
@@ -314,76 +316,101 @@ if command -v gnuplot >/dev/null 2>&1 && [[ $valid_performance_count -gt 0 ]]; t
   max_row=0
   max_id=""
   
-  # Use Python to generate chart data and capture output
-  python_output=$("$PYTHON_CMD" -c "
+  # Use Python to generate chart data
+  "$PYTHON_CMD" -c "
 import csv
 import re
 
-row_num = 0
-max_perf = 0
-max_row = 0
-max_id = ''
-
 with open('$csv_file', 'r') as f:
     reader = csv.reader(f)
     next(reader)  # Skip header
     
-    data_lines = []
-    gen_temp_lines = []
+    completed_order = 0  # Track order of completion
+    
+    with open('$data_file', 'w') as data_f:
+        data_f.write('# Order ID Performance Generation\\n')
+        
+    with open('$gen_data_temp', 'w') as gen_f:
+        pass  # Clear file
+    
+    max_perf = 0
+    max_id = ''
+    max_order = 0
     
     for row in reader:
         if len(row) < 5:
             continue
             
-        row_num += 1
         id, _, desc, perf, status = row[:5]
         
         # Extract generation from ID
         gen = 'gen01'  # default
-        match = re.match(r'^(gen[0-9]+)-', id)
+        gen_num = 1
+        match = re.match(r'^gen([0-9]+)-', id)
         if match:
-            gen = match.group(1)
+            gen = f'gen{match.group(1)}'
+            gen_num = int(match.group(1))
         
         # Only include completed algorithms with non-zero performance
         if perf and perf != '' and status in ['complete', 'completed']:
             try:
                 perf_val = float(perf)
                 if perf_val > 0:
-                    # Assign generation number for coloring (1-based)
-                    gen_num = 1
-                    match = re.match(r'^gen([0-9]+)-', id)
-                    if match:
-                        gen_num = int(match.group(1))
+                    completed_order += 1
+                    
+                    # Write to data file
+                    with open('$data_file', 'a') as f:
+                        f.write(f'{completed_order} \"{id}\" {perf} {gen_num}\\n')
                     
-                    data_lines.append(f'{row_num} \\\"{id}\\\" {perf} {gen_num}')
-                    gen_temp_lines.append(f'{gen} {perf}')
+                    # Write to gen temp file
+                    with open('$gen_data_temp', 'a') as f:
+                        f.write(f'{gen} {perf}\\n')
                     
                     # Track the winner
                     if perf_val > max_perf:
                         max_perf = perf_val
-                        max_row = row_num
+                        max_order = completed_order
                         max_id = id
             except ValueError:
                 pass
 
-# Write data file
-with open('$data_file', 'a') as f:
-    for line in data_lines:
-        f.write(line + '\\\\n')
-
-# Write gen temp file
-with open('$gen_data_temp', 'a') as f:
-    for line in gen_temp_lines:
-        f.write(line + '\\\\n')
-
 # Output max values for shell
 print(f'max_perf={max_perf}')
-print(f'max_row={max_row}')
-print(f'max_id=\\\"{max_id}\\\"')
-")
+print(f'max_row={max_order}')
+print(f'max_id=\"{max_id}\"')
+"
   
-  # Evaluate the Python output
-  eval "$python_output"
+  # Capture the output properly
+  eval "$("$PYTHON_CMD" -c "
+import csv
+import re
+
+with open('$csv_file', 'r') as f:
+    reader = csv.reader(f)
+    next(reader)
+    
+    max_perf = 0
+    max_id = ''
+    max_order = 0
+    completed_order = 0
+    
+    for row in reader:
+        if len(row) >= 5 and row[3] and row[4] in ['complete', 'completed']:
+            try:
+                perf_val = float(row[3])
+                if perf_val > 0:
+                    completed_order += 1
+                    if perf_val > max_perf:
+                        max_perf = perf_val
+                        max_order = completed_order
+                        max_id = row[0]
+            except ValueError:
+                pass
+
+print(f'max_perf={max_perf}')
+print(f'max_row={max_order}')
+print(f'max_id=\"{max_id}\"')
+")"
 
   # Create generation averages file and track max generation
   gen_index=1
@@ -397,11 +424,11 @@ print(f'max_id=\\\"{max_id}\\\"')
         avg=$(echo "scale=4; $sum / $count" | bc -l 2>/dev/null || echo "0")
         gen_num=$(echo "$gen" | sed 's/gen0*//')
         # Track max generation number
-        if [[ $gen_num -gt $max_gen_num ]]; then
+        if [[ $gen_num =~ ^[0-9]+$ ]] && [[ $gen_num -gt $max_gen_num ]]; then
           max_gen_num=$gen_num
         fi
         color=$(get_gen_color "$gen_num")
-        echo "$gen_index \"$gen\" $avg \"$color\"" >>"$gen_avg_file"
+        echo "$gen_index \"Gen$gen_num\" $avg \"$color\"" >>"$gen_avg_file"
         ((gen_index++))
       fi
     fi
@@ -417,26 +444,58 @@ print(f'max_id=\\\"{max_id}\\\"')
 
   # Generate dual plot
   if [[ -s "$data_file" ]]; then
-    # Build dynamic plot command for generations
+    # Debug: show data file content
+    # echo "DEBUG: Data file content:"
+    # cat "$data_file"
+    # echo "DEBUG: max_gen_num=$max_gen_num"
+    
+    # Plot all algorithms in order of completion, colored by generation
     plot_cmd=""
-    for ((i=1; i<=max_gen_num; i++)); do
-      color=$(get_gen_color "$i")
-      if [[ -n $plot_cmd ]]; then
-        plot_cmd="$plot_cmd, \\"$'\n'
+    gen_plots_added=0
+    
+    # Find all generations that have data
+    generations=($(awk '{if(NR>1) print $4}' "$data_file" | sort -n | uniq))
+    
+    for gen_num in "${generations[@]}"; do
+      if [[ -n $gen_num ]]; then
+        color=$(get_gen_color "$gen_num")
+        if [[ $gen_plots_added -gt 0 ]]; then
+          plot_cmd="$plot_cmd, \\"$'\n'
+        fi
+        plot_cmd="${plot_cmd}     \"$data_file\" using (\$4==$gen_num?\$1:1/0):3 with linespoints linewidth 2 linecolor rgb \"$color\" pointsize 1.2 title \"Gen $gen_num\""
+        ((gen_plots_added++))
       fi
-      plot_cmd="${plot_cmd}     \"$data_file\" using (\$4==$i?\$1:1/0):3 with linespoints linewidth 2 linecolor rgb \"$color\" pointsize 0.8 title \"Gen $i\""
     done
+    
     # Add winner point
-    plot_cmd="$plot_cmd, \\"$'\n'
-    plot_cmd="${plot_cmd}     \"$winner_file\" using 1:3 with points pointtype 7 pointsize 2 linecolor rgb \"#0066cc\" title \"Winner\""
+    if [[ -n $max_id && -s "$winner_file" ]]; then
+      if [[ $gen_plots_added -gt 0 ]]; then
+        plot_cmd="$plot_cmd, \\"$'\n'
+      fi
+      plot_cmd="${plot_cmd}     \"$winner_file\" using 1:3 with points pointtype 7 pointsize 3 linecolor rgb \"gold\" title \"Best ($max_id)\""
+    fi
     
-    # Build x-axis labels for generation chart
+    # Fallback if no generation-specific plots
+    if [[ $gen_plots_added -eq 0 ]]; then
+      plot_cmd="\"$data_file\" using 1:3 with linespoints linewidth 2 linecolor rgb \"#1f77b4\" pointsize 1.2 title \"Evolution Progress\""
+      if [[ -n $max_id && -s "$winner_file" ]]; then
+        plot_cmd="$plot_cmd, \\"$'\n'
+        plot_cmd="${plot_cmd}     \"$winner_file\" using 1:3 with points pointtype 7 pointsize 3 linecolor rgb \"gold\" title \"Best ($max_id)\""
+      fi
+    fi
+    
+    # Build x-axis labels for generation chart (include all generations from data)
     xtics_labels=""
-    for ((i=1; i<=max_gen_num; i++)); do
-      if [[ -n $xtics_labels ]]; then
-        xtics_labels="$xtics_labels, "
+    label_index=1
+    for gen in $(cut -d' ' -f1 "$gen_data_temp" | sort -u); do
+      if [[ -n $gen ]]; then
+        gen_display=$(echo "$gen" | sed 's/gen0*//')
+        if [[ -n $xtics_labels ]]; then
+          xtics_labels="$xtics_labels, "
+        fi
+        xtics_labels="${xtics_labels}\"Gen$gen_display\" $label_index"
+        ((label_index++))
       fi
-      xtics_labels="${xtics_labels}\"Gen$i\" $i"
     done
     
     gnuplot <<EOF
@@ -448,11 +507,12 @@ set multiplot layout 2,1 margins 0.08,0.82,0.15,0.95 spacing 0.1,0.15
 
 #=================== TOP PLOT: Performance Over Time ===================
 set title "Algorithm Evolution Performance Over Time" font ",14"
-set xlabel "Evolution Run"
+set xlabel "Completion Order (Algorithm #)"
 set ylabel "Performance Score"
 set grid
 set key outside right
-set xtics auto
+set xtics 1
+set autoscale y
 
 # Define colors for generations
 plot $plot_cmd

package/bin/claude-evolve-ideate

@@ -14,14 +14,34 @@ else
   load_config
 fi
 
-# Helper function to call Claude with usage limit detection
-call_claude_with_limit_check() {
+# Helper function to call AI model (codex o3-pro if available, else Claude)
+call_ai_with_limit_check() {
   local prompt="$1"
-  local model="${2:-opus}"
+  local fallback_model="${2:-opus}"
+  
+  # Check if codex is available
+  if command -v codex >/dev/null 2>&1; then
+    echo "[INFO] Using codex o3-pro for ideation (smartest available model)" >&2
+    
+    # Call codex with o3-pro model
+    local ai_output
+    ai_output=$(echo "$prompt" | codex -m o3-pro 2>&1)
+    local ai_exit_code=$?
+    
+    if [[ $ai_exit_code -eq 0 ]]; then
+      echo "$ai_output"
+      return 0
+    else
+      echo "[WARN] Codex failed, falling back to Claude" >&2
+    fi
+  fi
+  
+  # Fall back to Claude
+  echo "[INFO] Using Claude $fallback_model for ideation" >&2
   
   # Call Claude and capture output
   local claude_output
-  claude_output=$(echo "$prompt" | claude --dangerously-skip-permissions --model "$model" -p 2>&1)
+  claude_output=$(echo "$prompt" | claude --dangerously-skip-permissions --model "$fallback_model" -p 2>&1)
   local claude_exit_code=$?
   
   # Check for usage limit
@@ -51,6 +71,11 @@ call_claude_with_limit_check() {
   return $claude_exit_code
 }
 
+# Backward compatibility alias
+call_claude_with_limit_check() {
+  call_ai_with_limit_check "$@"
+}
+
 # Parse arguments
 use_strategies=true
 no_ai=false
@@ -253,9 +278,9 @@ ideate_manual() {
 
 # Generate ideas using AI with multi-strategy approach
 ideate_ai_strategies() {
-  # Check for claude CLI
-  if ! command -v claude >/dev/null 2>&1; then
-    echo "[WARN] Claude CLI not found. Falling back to manual entry."
+  # Check for AI CLI (codex or claude)
+  if ! command -v codex >/dev/null 2>&1 && ! command -v claude >/dev/null 2>&1; then
+    echo "[WARN] No AI CLI found (codex or claude). Falling back to manual entry."
     return 1
   fi
 
@@ -329,9 +354,9 @@ Example descriptions:
 
 Add exactly $count rows to the CSV file now."
 
-  echo "[INFO] Calling Claude Opus to generate $count novel exploration ideas..."
-  if ! call_claude_with_limit_check "$prompt" "opus"; then
-    echo "[WARN] Claude failed to generate novel ideas" >&2
+  echo "[INFO] Generating $count novel exploration ideas..."
+  if ! call_ai_with_limit_check "$prompt" "opus"; then
+    echo "[WARN] AI failed to generate novel ideas" >&2
     return 1
   fi
   echo "[INFO] Novel exploration ideas generated"
@@ -382,9 +407,9 @@ Example descriptions:
 
 Add exactly $count parameter tuning rows to the CSV file now."
 
-  echo "[INFO] Calling Claude Opus to generate $count hill climbing ideas..."
-  if ! call_claude_with_limit_check "$prompt" "opus"; then
-    echo "[WARN] Claude failed to generate hill climbing ideas" >&2
+  echo "[INFO] Generating $count hill climbing ideas..."
+  if ! call_ai_with_limit_check "$prompt" "opus"; then
+    echo "[WARN] AI failed to generate hill climbing ideas" >&2
     return 1
   fi
   echo "[INFO] Hill climbing ideas generated"
@@ -435,9 +460,9 @@ Example descriptions:
 
 Add exactly $count structural modification rows to the CSV file now."
 
-  echo "[INFO] Calling Claude Opus to generate $count structural mutation ideas..."
-  if ! call_claude_with_limit_check "$prompt" "opus"; then
-    echo "[WARN] Claude failed to generate structural mutation ideas" >&2
+  echo "[INFO] Generating $count structural mutation ideas..."
+  if ! call_ai_with_limit_check "$prompt" "opus"; then
+    echo "[WARN] AI failed to generate structural mutation ideas" >&2
     return 1
   fi
   echo "[INFO] Structural mutation ideas generated"
@@ -488,9 +513,9 @@ Example descriptions:
 
 Add exactly $count hybrid combination rows to the CSV file now."
 
-  echo "[INFO] Calling Claude Opus to generate $count crossover hybrid ideas..."
-  if ! call_claude_with_limit_check "$prompt" "opus"; then
-    echo "[WARN] Claude failed to generate crossover ideas" >&2
+  echo "[INFO] Generating $count crossover hybrid ideas..."
+  if ! call_ai_with_limit_check "$prompt" "opus"; then
+    echo "[WARN] AI failed to generate crossover ideas" >&2
     return 1
   fi
   echo "[INFO] Crossover hybrid ideas generated"
@@ -498,9 +523,9 @@ Add exactly $count hybrid combination rows to the CSV file now."
 
 # Legacy AI generation mode (for backward compatibility)
 ideate_ai_legacy() {
-  # Check for claude CLI
-  if ! command -v claude >/dev/null 2>&1; then
-    echo "[WARN] Claude CLI not found. Falling back to manual entry."
+  # Check for AI CLI (codex or claude)
+  if ! command -v codex >/dev/null 2>&1 && ! command -v claude >/dev/null 2>&1; then
+    echo "[WARN] No AI CLI found (codex or claude). Falling back to manual entry."
     return 1
   fi
 
@@ -553,9 +578,9 @@ Requirements for new CSV rows:
 
 Add exactly $TOTAL_IDEAS algorithm variation rows to the CSV file now."
 
-  echo "[INFO] Calling Claude Opus to generate $TOTAL_IDEAS ideas (legacy mode)..."
-  if ! call_claude_with_limit_check "$prompt" "opus"; then
-    echo "[WARN] Claude failed to generate ideas" >&2
+  echo "[INFO] Generating $TOTAL_IDEAS ideas (legacy mode)..."
+  if ! call_ai_with_limit_check "$prompt" "opus"; then
+    echo "[WARN] AI failed to generate ideas" >&2
     return 1
   fi
   echo "[INFO] Legacy ideas generated"

package/bin/claude-evolve-run-parallel

@@ -71,7 +71,10 @@ reader = csv.reader(sys.stdin)
 next(reader)  # Skip header
 count = 0
 for row in reader:
-    if len(row) >= 5 and (row[4] == 'pending' or row[4] == ''):
+    # If row has fewer than 5 fields, treat as pending
+    if len(row) < 5:
+        count += 1
+    elif len(row) >= 5 and (row[4] == 'pending' or row[4] == ''):
         count += 1
 print(count)
 "

package/lib/csv-lock.sh

@@ -162,7 +162,15 @@ with open('$csv_file', 'r') as f:
 # Find first pending candidate
 candidate_id = None
 for i in range(1, len(rows)):
-    if len(rows[i]) >= 5 and (rows[i][4] == 'pending' or rows[i][4] == ''):
+    # If row has fewer than 5 fields, it's pending
+    if len(rows[i]) < 5:
+        candidate_id = rows[i][0]
+        # Ensure row has 5 fields before setting status
+        while len(rows[i]) < 5:
+            rows[i].append('')
+        rows[i][4] = 'running'  # Update status
+        break
+    elif len(rows[i]) >= 5 and (rows[i][4] == 'pending' or rows[i][4] == ''):
         candidate_id = rows[i][0]
         rows[i][4] = 'running'  # Update status
         break

package/lib/csv_helper.py

@@ -12,7 +12,11 @@ def find_pending_row(csv_path):
         reader = csv.reader(f)
         next(reader)  # Skip header
         for row_num, row in enumerate(reader, start=2):
-            # Ensure row has at least 5 fields
+            # If row has fewer than 5 fields, it's pending
+            if len(row) < 5:
+                return row_num
+            
+            # Ensure row has at least 5 fields for status check
             while len(row) < 5:
                 row.append('')
             

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-evolve",
-  "version": "1.3.14",
+  "version": "1.3.16",
   "bin": {
     "claude-evolve": "./bin/claude-evolve",
     "claude-evolve-main": "./bin/claude-evolve-main",