Instruction file imported from albertcastineira/copilot-guidelines (.github/instructions/performance-budget-and-profiling.instructions.md). Copyright stays with the author.

Performance Budget And Profiling

Performance changes must be measured, not guessed.

Performance budget policy

For critical paths, define budgets before implementation:

• latency budget (p50/p95/p99),
• throughput target,
• error-rate threshold,
• CPU/memory budget.

Required profiling workflow

1. Capture baseline metrics.
2. Profile hot paths under representative load.
3. Implement optimizations.
4. Re-measure and compare against baseline.
5. Document results and remaining risks.

Backend checks

• Avoid N+1 query patterns.
• Verify index usage for high-frequency queries.
• Ensure serialization/parsing overhead is bounded.

Frontend checks

• Validate LCP/INP/CLS impact.
• Avoid render thrashing and unnecessary re-renders.
• Verify bundle and route-level payload budgets.

Output expectation

Responses must include:

• baseline vs after metrics,
• bottlenecks found,
• optimization decisions,
• rollback plan if performance regresses.