PerfDiff

PerfDiff is a web app (with a lightweight CLI agent) that automatically detects performance regressions in CI by comparing load-test runs against a baseline and explaining what changed. It ingests results from common tools (k6/JMeter/Locust), stores time-series metrics, and produces a “diff report” per build: p95/p99 latency shifts, error-rate deltas, throughput changes, and resource anomalies. The AI layer summarizes likely causes (e.g., new endpoint, N+1 query, GC pressure) and points to the exact commit range, endpoints, and metrics that moved. It also enforces performance budgets (SLO-style gates) so teams stop shipping slow builds. Brutal reality: this only works if teams already run repeatable tests and have stable environments; otherwise noise will drown signal. The MVP should focus on deterministic comparison, clear gating, and minimal setup—not fancy dashboards.