Production Harness Engineering

#!/usr/bin/env python3
"""harness-roi.py — 하네스 ROI 측정 대시보드"""
import subprocess
import json
from pathlib import Path
from datetime import date, timedelta
import re

def get_git_stats(days: int = 14) -> dict:
    """git 히스토리에서 코드 변경률(churn) 계산"""
    since = (date.today() - timedelta(days=days)).isoformat()
    
    result = subprocess.run(
        ['git', 'log', '--since', since, '--oneline', '--stat'],
        capture_output=True, text=True
    )
    
    commits = result.stdout.count('commit') if 'commit' in result.stdout else result.stdout.count('\n\n')
    insertions = sum(int(m) for m in re.findall(r'(\d+) insertion', result.stdout))
    deletions = sum(int(m) for m in re.findall(r'(\d+) deletion', result.stdout))
    
    churn_rate = deletions / max(insertions, 1)
    
    return {
        'period_days': days,
        'commits': commits,
        'insertions': insertions,
        'deletions': deletions,
        'churn_rate': round(churn_rate, 2),
        'churn_status': '✅' if churn_rate < 0.15 else '⚠️' if churn_rate < 0.30 else '❌'
    }

def analyze_audit_logs(days: int = 7) -> dict:
    """JSONL 감사 로그에서 도구 사용 패턴 분석"""
    audit_dir = Path.home() / '.claude' / 'audit'
    if not audit_dir.exists():
        return {'error': '감사 로그 없음 — PostToolUse JSONL 훅을 먼저 설정하세요'}
    
    total_tool_calls = 0
    tool_counts = {}
    blocked_calls = 0
    
    for i in range(days):
        log_date = (date.today() - timedelta(days=i)).isoformat()
        log_file = audit_dir / f'{log_date}.jsonl'
        if not log_file.exists():
            continue
        
        for line in log_file.read_text().split('\n'):
            if not line.strip():
                continue
            try:
                entry = json.loads(line)
                payload = entry.get('event', {})
                tool_name = payload.get('tool_name', 'unknown')
                total_tool_calls += 1
                tool_counts[tool_name] = tool_counts.get(tool_name, 0) + 1
                if payload.get('exit_code') == 2:
                    blocked_calls += 1
            except json.JSONDecodeError:
                pass
    
    top_tools = sorted(tool_counts.items(), key=lambda x: x[1], reverse=True)[:5]
    
    return {
        'period_days': days,
        'total_tool_calls': total_tool_calls,
        'blocked_calls': blocked_calls,
        'block_rate': round(blocked_calls / max(total_tool_calls, 1) * 100, 1),
        'top_tools': top_tools,
        'calls_per_day': round(total_tool_calls / days, 1)
    }

def generate_dashboard():
    print("=" * 50)
    print("HARNESS ROI DASHBOARD")
    print(f"Date: {date.today().isoformat()}")
    print("=" * 50)
    
    print("\n📊 Code Churn (14일):")
    git_stats = get_git_stats(14)
    if 'error' not in git_stats:
        print(f"  Insertions: {git_stats['insertions']:,} lines")
        print(f"  Deletions:  {git_stats['deletions']:,} lines")
        print(f"  Churn Rate: {git_stats['churn_rate']} {git_stats['churn_status']} (목표: <0.15)")
        print(f"  Commits:    {git_stats['commits']}")
    
    print("\n🔧 Tool Usage (7일):")
    audit_stats = analyze_audit_logs(7)
    if 'error' not in audit_stats:
        print(f"  Total calls: {audit_stats['total_tool_calls']:,}")
        print(f"  Blocked:     {audit_stats['blocked_calls']} ({audit_stats['block_rate']}% block rate)")
        print(f"  Daily avg:   {audit_stats['calls_per_day']} calls/day")
        print("  Top tools:")
        for tool, count in audit_stats['top_tools']:
            print(f"    {tool}: {count}")
    else:
        print(f"  {audit_stats['error']}")
    
    print("\n📋 Next Actions:")
    if 'churn_rate' in git_stats and git_stats['churn_rate'] > 0.15:
        print("  ⚠️  높은 코드 변경률 — CLAUDE.md 규칙 강화 또는 더 구체적인 태스크 분해")
    if 'block_rate' in audit_stats and audit_stats['block_rate'] > 5:
        print("  ⚠️  높은 블록률 — 허용 규칙 범위 확인 (너무 제한적)")
    print("  💡 Verification Tax 측정: PR 첫 커밋→승인 시간 추적 시작 권장")

if __name__ == '__main__':
    generate_dashboard()