feat: added a script to summarize token usage based on a log folder

1 files changed, 117 insertions, 0 deletions

diff --git a/scripts/summarize_tokens.py b/scripts/summarize_tokens.py
new file mode 100755
index 0000000..02387a2
--- /dev/null
+++ b/scripts/summarize_tokens.py
@@ -0,0 +1,117 @@
+#!/usr/bin/env python3
+import os
+import json
+import argparse
+from collections import defaultdict
+from pathlib import Path
+
+def summarize_tokens(log_dir):
+    """
+    Summarizes token usage from JSON logs in the specified directory.
+    Groups the results by model and then by scenario.
+    """
+    # Grouped by model, then scenario
+    # Totals: prompt, completion, total, run_count
+    summary = defaultdict(lambda: defaultdict(lambda: {
+        'prompt_tokens': 0, 
+        'completion_tokens': 0, 
+        'total_tokens': 0, 
+        'run_count': 0
+    }))
+    
+    overall_totals = {
+        'prompt_tokens': 0, 
+        'completion_tokens': 0, 
+        'total_tokens': 0, 
+        'run_count': 0
+    }
+
+    path = Path(log_dir)
+    if not path.exists() or not path.is_dir():
+        print(f"Error: '{log_dir}' is not a valid directory.")
+        return
+
+    # Iterate through all JSON files in the log directory
+    files_processed = 0
+    for json_file in path.rglob('*.json'):
+        try:
+            with open(json_file, 'r', encoding='utf-8') as f:
+                data = json.load(f)
+                
+                # Check if it has the required fields
+                if all(k in data for k in ['prompt_tokens', 'completion_tokens', 'total_tokens', 'model', 'scenario']):
+                    model = data['model']
+                    scenario = data['scenario']
+                    
+                    p_tokens = data.get('prompt_tokens', 0)
+                    c_tokens = data.get('completion_tokens', 0)
+                    t_tokens = data.get('total_tokens', 0)
+
+                    # Update model/scenario totals
+                    summary[model][scenario]['prompt_tokens'] += p_tokens
+                    summary[model][scenario]['completion_tokens'] += c_tokens
+                    summary[model][scenario]['total_tokens'] += t_tokens
+                    summary[model][scenario]['run_count'] += 1
+                    
+                    # Update overall totals
+                    overall_totals['prompt_tokens'] += p_tokens
+                    overall_totals['completion_tokens'] += c_tokens
+                    overall_totals['total_tokens'] += t_tokens
+                    overall_totals['run_count'] += 1
+                    
+                    files_processed += 1
+        except json.JSONDecodeError:
+            # We skip files that are not valid JSON or completely empty 
+            print(f"Warning: Could not parse {json_file}")
+            pass
+        except Exception as e:
+            print(f"Warning: Error processing {json_file}: {e}")
+
+    if files_processed == 0:
+        print(f"No valid log files with token metrics found in '{log_dir}'.")
+        return
+
+    # Print summary
+    print("=" * 90)
+    print(f"{'Token Usage Summary':^90}")
+    print("=" * 90)
+    print()
+
+    for model, scenarios in sorted(summary.items()):
+        print(f"Model: {model}")
+        print("-" * 90)
+        print(f"{'Scenario':<45} | {'Runs':<5} | {'Prompt':<10} | {'Completion':<10} | {'Total':<10}")
+        print("-" * 90)
+        
+        model_p = 0
+        model_c = 0
+        model_t = 0
+        model_runs = 0
+        
+        for scenario, counts in sorted(scenarios.items()):
+            p = counts['prompt_tokens']
+            c = counts['completion_tokens']
+            t = counts['total_tokens']
+            r = counts['run_count']
+            
+            model_p += p
+            model_c += c
+            model_t += t
+            model_runs += r
+            
+            print(f"{scenario:<45} | {r:<5} | {p:<10,d} | {c:<10,d} | {t:<10,d}")
+            
+        print("-" * 90)
+        print(f"{'Model Total':<45} | {model_runs:<5} | {model_p:<10,d} | {model_c:<10,d} | {model_t:<10,d}")
+        print("=" * 90)
+        print()
+
+    print(f"{'OVERALL TOTALS':<45} | {overall_totals['run_count']:<5} | {overall_totals['prompt_tokens']:<10,d} | {overall_totals['completion_tokens']:<10,d} | {overall_totals['total_tokens']:<10,d}")
+    print("=" * 90)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Summarize token usage from JSON logs.")
+    parser.add_argument("log_dir", type=str, help="Path to the directory containing JSON log files")
+    args = parser.parse_args()
+    
+    summarize_tokens(args.log_dir)