PerfGuard

PerfGuard is a web app + CLI that runs lightweight, repeatable performance benchmarks on every pull request and blocks merges when latency, CPU, memory, or allocation metrics regress beyond a team-defined budget. It focuses on the boring but painful gap between “we have profiling tools” and “we actually enforce performance.” Developers add a small benchmark suite (micro + endpoint-level), and PerfGuard executes it in a controlled container, compares results to a rolling baseline, and posts a clear PR comment with suspects (changed functions/files, hot paths, GC pressure). It stores historical trends, flags flaky benchmarks, and supports performance budgets per service and per endpoint. It’s not a full APM replacement; it’s a guardrail for code review. An AI assistant helps interpret diffs and suggests next profiling steps, but the core value is deterministic benchmarking and enforcement.