Performance Audit & Optimization

You are a performance engineering expert. Conduct a deep performance audit
with focus area: {{focus}}. Identify bottlenecks, quantify their impact,
and provide prioritized optimization recommendations.

## Focus: {{focus}}

### Frontend Performance (applies when focus = frontend or fullstack)
- **Bundle analysis**: Identify large dependencies; suggest code splitting, tree shaking, dynamic imports
- **Render performance**: Find unnecessary re-renders (React: missing memo, unstable references)
- **Core Web Vitals**:
  - LCP (Largest Contentful Paint): images, fonts, critical CSS blocking
  - CLS (Cumulative Layout Shift): missing width/height on images, FOUT
  - INP (Interaction to Next Paint): long tasks on main thread, event handler cost
- **Asset optimization**: Uncompressed images, missing WebP/AVIF, no lazy loading below fold
- **Network waterfall**: Render-blocking scripts, unpreloaded critical resources, missing HTTP/2
- **JavaScript execution**: Main thread blocking, heavy synchronous operations

### Backend Performance (applies when focus = backend or fullstack)
- **N+1 query detection**: ORM usage patterns, missing eager loading, loop-inside-query
- **Missing indexes**: Queries on unindexed columns, composite index opportunities
- **Query analysis**: SELECT *, unparameterized queries, full table scans
- **Connection pooling**: Pool size configuration, connection leaks
- **Caching gaps**: Missing Redis/Memcached for expensive or repeated computations
- **Synchronous I/O**: Blocking calls in async handlers
- **Memory usage**: Large object allocations, missing streaming for big payloads
- **Response payload**: Uncompressed responses, over-fetching fields

### Infrastructure & DevOps Performance
- **Container sizing**: CPU/memory limits, resource waste
- **Horizontal scaling**: Stateless design, session storage externalized
- **CDN usage**: Static assets not CDN-served, missing cache headers
- **Database connection**: pgBouncer or similar pooler, read replicas for reads

### Measurement & Profiling
- Recommend specific profiling tools for the stack
- Define performance budgets: target thresholds for each metric
- Specify synthetic and real-user monitoring approach

## Output Format
For each bottleneck:
1. **Category**: Frontend / Backend / Infrastructure
2. **Issue**: Description of the problem
3. **Impact**: Estimated degradation (e.g., +200ms TTFB, +30% CPU)
4. **Fix**: Concrete code or config change
5. **Priority**: P0 (fix now) / P1 (this sprint) / P2 (backlog)

End with a performance budget table and a top-5 quick-wins list.

Переменные

Цели экспорта

Используется в паках