Make fleet data quality
measurable.
A reproducible lab for inspecting telemetry freshness, replay integrity, offset continuity and vehicle-level health — built around deterministic synthetic event streams.
Scenario explorer
Switch profiles to compare data-quality signals from deterministic simulations
Observed quality signals
Integrity checks
Separate signal from assumption.
FleetScope focuses on telemetry behaviors that are easy to misclassify when inspecting event streams. It reports evidence without silently turning every irregularity into a data-loss claim.
p50/p95/mean/max age, late and future timestamps, explicit resend copies and duplicate event identifiers.
Offset gaps, out-of-order arrivals, repeated offsets and per-partition coverage over observed windows.
Unchanged values remain observable. Conflicting replays are distinguished from exact retry deliveries.
Benchmarked locally, not production-claimed.
Three-repeat microbenchmarks on Python 3.13.5 measured deterministic generation plus batch analysis with traced memory. Results below describe this local environment only.
| Events | Median runtime | Slowest run | Median traced memory |
|---|---|---|---|
| 1,035 | 0.234 s | 0.236 s | 2.32 MiB |
| 5,175 | 1.245 s | 1.292 s | 11.45 MiB |
| 10,350 | 2.443 s | 2.621 s | 22.62 MiB |
Built to be inspected and extended.
FastAPI service, typed event schema, deterministic profile generator, batch analyzer, SQLite run history, event explorer, CSV export, Prometheus endpoint, CLI, Docker setup, CI workflow and test suite.
Verification: 16 pytest checks passed; Python compilation and JavaScript syntax checks passed; FastAPI TestClient smoke checks returned HTTP 200 across dashboard assets and key API routes. Browser screenshot QA was not completed in the build environment.