PerfGuard

PerfGuard is a web app (with optional CLI agent) that continuously detects SQL query performance regressions across releases. Instead of generic “slow query” dashboards, it focuses on change detection: it snapshots normalized queries, execution plans, and key runtime stats, then flags when a deploy causes plan flips, cardinality estimate shifts, or latency/IO spikes beyond a learned baseline. It integrates with CI/CD (GitHub Actions, GitLab CI) to run lightweight canary workloads or replay sampled production queries in a staging environment, producing a pass/fail “performance gate” and a short report engineers can act on. It supports PostgreSQL first (realistic scope), with a clear path to MySQL later. This is a combination traditional + AI app: traditional for collection/analysis, AI to summarize root-cause hypotheses and suggest safe next steps (e.g., stats refresh, query rewrite, index candidate) with confidence levels and links to evidence.