ShardSleuth

ShardSleuth is a web app (with a lightweight CLI agent) that continuously validates correctness in distributed databases by detecting anomalies that normal monitoring misses: lost updates, write skew, stale reads, and broken invariants across shards/regions. It ingests database change streams and query traces, then runs “invariant checks” you define (e.g., account balance never negative, inventory never below zero) and flags violations with a minimal reproduction timeline. It also estimates which consistency level, retry policy, or transaction boundary likely caused the issue, and suggests safer patterns. This is a combination traditional + AI app: traditional pipelines do deterministic detection and correlation; AI helps generate candidate invariants from schema + code snippets and summarizes incident root-cause narratives. The goal is not to replace observability; it’s to make correctness observable in systems where eventual consistency and retries create silent data corruption.