Mini Data Engine Runtime Copilot

A runtime diagnostics layer for local data engineering labs that wraps PostgreSQL-like and Databricks-like toy implementations with MCP tooling. Exposes 47 operations including explain_run for traced execution timelines, project_run_regression for full test suites with explainability, health_check and benchmark_calls for SLO tracking, and memory_search for incident retrieval. Built for teams running onboarding workshops or teaching storage internals who want runtime operations accessible through structured tooling instead of shell scripts. The explain layer treats both successful runs and controlled failure scenarios as first-class traced events, so you can inspect idempotency conflicts or concurrency storms the same way you'd review a prod incident.