claude-evolve 1.3.3 → 1.3.5

package/README.md

@@ -196,6 +196,27 @@ your-project/
 └── (your main project files)
 ```
 
+## Environment Variables for Evaluators
+
+When your evaluator.py runs, it has access to the `EXPERIMENT_ID` environment variable containing the current experiment's ID (e.g., `gen07-001`). This allows evaluators to:
+
+- Save experiment-specific output files
+- Log metrics with experiment identifiers
+- Implement experiment-aware logic
+- Track which algorithm variant is being evaluated
+
+Example usage in evaluator.py:
+```python
+import os
+
+# Get the current experiment ID
+experiment_id = os.environ.get('EXPERIMENT_ID', 'unknown')
+
+# Use it for logging or file naming
+output_file = f"results_{experiment_id}.json"
+print(f"Evaluating experiment: {experiment_id}")
+```
+
 ## Configuration
 
 Edit `evolution/config.yaml` to customize:

package/bin/claude-evolve-analyze

@@ -222,21 +222,16 @@ 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
+  # Get color by generation number (rotates through 5 colors)
   get_gen_color() {
     local gen_num="$1"
-    case $gen_num in
+    local color_index=$(( (gen_num - 1) % 5 + 1 ))
+    case $color_index in
       1) echo "#1f77b4" ;;  # blue
       2) echo "#ff7f0e" ;;  # orange  
       3) echo "#2ca02c" ;;  # green
       4) echo "#d62728" ;;  # red
       5) echo "#9467bd" ;;  # purple
-      6) echo "#8c564b" ;;  # brown
-      7) echo "#e377c2" ;;  # pink
-      8) echo "#7f7f7f" ;;  # gray
-      9) echo "#bcbd22" ;;  # olive
-      10) echo "#17becf" ;; # cyan
-      *) echo "#cccccc" ;;  # default gray
     esac
   }
 
@@ -284,8 +279,9 @@ if command -v gnuplot >/dev/null 2>&1 && [[ $valid_performance_count -gt 0 ]]; t
     fi
   done <"$csv_file"
 
-  # Create generation averages file
+  # Create generation averages file and track max generation
   gen_index=1
+  max_gen_num=0
   for gen in $(cut -d' ' -f1 "$gen_data_temp" | sort -u); do
     if grep -q "^$gen " "$gen_data_temp"; then
       # Calculate average for this generation
@@ -294,6 +290,10 @@ if command -v gnuplot >/dev/null 2>&1 && [[ $valid_performance_count -gt 0 ]]; t
       if [[ $count -gt 0 ]]; then
         avg=$(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 $max_gen_num ]]; then
+          max_gen_num=$gen_num
+        fi
         color=$(get_gen_color "$gen_num")
         echo "$gen_index \"$gen\" $avg \"$color\"" >>"$gen_avg_file"
         ((gen_index++))
@@ -308,6 +308,28 @@ if command -v gnuplot >/dev/null 2>&1 && [[ $valid_performance_count -gt 0 ]]; t
 
   # Generate dual plot
   if [[ -s "$data_file" ]]; then
+    # Build dynamic plot command for generations
+    plot_cmd=""
+    for ((i=1; i<=max_gen_num; i++)); do
+      color=$(get_gen_color "$i")
+      if [[ -n $plot_cmd ]]; then
+        plot_cmd="$plot_cmd, \\"$'\n'
+      fi
+      plot_cmd="${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\""
+    
+    # Build x-axis labels for generation chart
+    xtics_labels=""
+    for ((i=1; i<=max_gen_num; i++)); do
+      if [[ -n $xtics_labels ]]; then
+        xtics_labels="$xtics_labels, "
+      fi
+      xtics_labels="${xtics_labels}\"Gen$i\" $i"
+    done
+    
     gnuplot <<EOF
 set terminal png size 1200,800
 set output "$output_file"
@@ -324,12 +346,7 @@ set key outside right
 set xtics auto
 
 # Define colors for generations
-plot "$data_file" using (\$4==1?\$1:1/0):3 with linespoints linewidth 2 linecolor rgb "#1f77b4" pointsize 0.8 title "Gen 1", \\
-     "$data_file" using (\$4==2?\$1:1/0):3 with linespoints linewidth 2 linecolor rgb "#ff7f0e" pointsize 0.8 title "Gen 2", \\
-     "$data_file" using (\$4==3?\$1:1/0):3 with linespoints linewidth 2 linecolor rgb "#2ca02c" pointsize 0.8 title "Gen 3", \\
-     "$data_file" using (\$4==4?\$1:1/0):3 with linespoints linewidth 2 linecolor rgb "#d62728" pointsize 0.8 title "Gen 4", \\
-     "$data_file" using (\$4==5?\$1:1/0):3 with linespoints linewidth 2 linecolor rgb "#9467bd" pointsize 0.8 title "Gen 5", \\
-     "$winner_file" using 1:3 with points pointtype 7 pointsize 2 linecolor rgb "#0066cc" title "Winner"
+plot $plot_cmd
 
 #=================== BOTTOM PLOT: Generation Averages ===================
 set title "Average Performance by Generation" font ",14"
@@ -341,7 +358,7 @@ unset key
 set grid y
 
 # Set custom x-axis labels
-set xtics ("Gen1" 1, "Gen2" 2, "Gen3" 3, "Gen4" 4, "Gen5" 5)
+set xtics ($xtics_labels)
 
 plot "$gen_avg_file" using 1:3 with boxes linecolor rgb "#4CAF50" notitle
 

package/bin/claude-evolve-run

@@ -434,13 +434,13 @@ echo "[INFO] Algorithm ready at: $output_file"
 
 # Run evaluator
 echo "[INFO] Running evaluation..."
-echo "[INFO] Executing: $PYTHON_CMD $FULL_EVALUATOR_PATH $output_file"
+echo "[INFO] Executing: EXPERIMENT_ID=$id $PYTHON_CMD $FULL_EVALUATOR_PATH $output_file"
 eval_output=""
 eval_exit_code=0
 
 if [[ -n $timeout_seconds ]]; then
   echo "[INFO] Evaluation timeout: ${timeout_seconds}s"
-  if eval_output=$(timeout "$timeout_seconds" "$PYTHON_CMD" "$FULL_EVALUATOR_PATH" "$output_file" 2>&1); then
+  if eval_output=$(EXPERIMENT_ID="$id" timeout "$timeout_seconds" "$PYTHON_CMD" "$FULL_EVALUATOR_PATH" "$output_file" 2>&1); then
     eval_exit_code=0
   else
     eval_exit_code=$?
@@ -455,7 +455,7 @@ if [[ -n $timeout_seconds ]]; then
     fi
   fi
 else
-  if eval_output=$("$PYTHON_CMD" "$FULL_EVALUATOR_PATH" "$output_file" 2>&1); then
+  if eval_output=$(EXPERIMENT_ID="$id" "$PYTHON_CMD" "$FULL_EVALUATOR_PATH" "$output_file" 2>&1); then
     eval_exit_code=0
   else
     eval_exit_code=$?
@@ -473,7 +473,7 @@ echo "----------------------------------------"
   echo "=== EVALUATOR EXECUTION ==="
   echo "ID: $id"
   echo "Algorithm: $output_file"
-  echo "Command: $PYTHON_CMD $FULL_EVALUATOR_PATH $output_file"
+  echo "Command: EXPERIMENT_ID=$id $PYTHON_CMD $FULL_EVALUATOR_PATH $output_file"
   echo "Exit code: $eval_exit_code"
   echo "Timestamp: $(date)"
   echo
@@ -554,7 +554,7 @@ else
       recovery_attempted=true
       # Retry the evaluation
       echo "[INFO] Retrying evaluation after recovery attempt..."
-      if eval_output=$("$PYTHON_CMD" "$FULL_EVALUATOR_PATH" "$output_file" 2>&1); then
+      if eval_output=$(EXPERIMENT_ID="$id" "$PYTHON_CMD" "$FULL_EVALUATOR_PATH" "$output_file" 2>&1); then
         # Re-process the successful result
         if score=$(echo "$eval_output" | grep -o '"score"[[:space:]]*:[[:space:]]*[0-9.]*' | cut -d: -f2 | tr -d ' '); then
           if [[ -n $score ]]; then

package/bin/claude-evolve-worker

@@ -188,7 +188,7 @@ eval_exit_code=0
 if [[ -n $timeout_seconds ]]; then
   echo "[WORKER-$$] Evaluation timeout: ${timeout_seconds}s"
   # For Modal compatibility, don't capture stderr
-  if eval_output=$(timeout "$timeout_seconds" "$PYTHON_CMD" "$FULL_EVALUATOR_PATH" "$output_file"); then
+  if eval_output=$(EXPERIMENT_ID="$id" timeout "$timeout_seconds" "$PYTHON_CMD" "$FULL_EVALUATOR_PATH" "$output_file"); then
     eval_exit_code=0
   else
     eval_exit_code=$?
@@ -200,7 +200,7 @@ if [[ -n $timeout_seconds ]]; then
   fi
 else
   # For Modal compatibility, don't capture stderr
-  if eval_output=$("$PYTHON_CMD" "$FULL_EVALUATOR_PATH" "$output_file"); then
+  if eval_output=$(EXPERIMENT_ID="$id" "$PYTHON_CMD" "$FULL_EVALUATOR_PATH" "$output_file"); then
     eval_exit_code=0
   else
     eval_exit_code=$?

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-evolve",
-  "version": "1.3.3",
+  "version": "1.3.5",
   "bin": {
     "claude-evolve": "./bin/claude-evolve",
     "claude-evolve-main": "./bin/claude-evolve-main",