Been working a lot on this on the side for a few months, it's stable enough now that I wanted to put it in front of people who actually run OpenTelemetry. So it reads the OTel traces your services already emit and flags I/O anti-patterns: N+1 SQL and HTTP, redundant calls, slow queries, excessive fanout, chatty services between services and pool saturation. It works at the protocol/span level, so it doesn't need to know whether you're on Hibernate, EF Core, SQLAlchemy or a bare driver, it just sees the queries they end up sending. Two ways to run it: as a one-shot CI gate on captured traces (exits non-zero past a threshold, emits SARIF so findings land in GitHub/GitLab code scanning), or as a long-running daemon that ingests OTLP and exposes Prometheus metrics, a query API and a self-contained HTML dashboard. Single static binary, no agent to attach to your runtime.…