Garantia web/mobile
Low-end Device Check
Measure your app or site on real low-end Android hardware, not an emulator.
01 — O problema
O problema
Emulators don't faithfully reproduce behavior under the constrained memory/CPU of a real low-end phone.
Para quem e
Any company whose app/web must work on low-end Android phones (markets where most real users have limited hardware) and has no hardware of its own to test on.
02 — O que voce nos envia
O que voce nos envia
| Campo | Formato | Validação |
|---|---|---|
| url or package_name | http(s) URL, or Android package name | Exactly one of the two is required, never both |
| min_tier | low / mid / high | Defaults to low; compared against device_tier_observed as meets_min_tier |
| cold_start | Boolean | Defaults to true — measures am start -W TotalTime/WaitTime |
03 — O que roda
O que roda
- The worker runs the check on a real low-end Android device (never an emulator): device profile, load/cold-start timing, PSS memory, jank frames, and any FATAL EXCEPTION/ANR observed via a logcat window.
- Fault profiles covered by the harness: network_loss, adb_loss, sim_anomaly, reboot, ui_change (service_restart is not yet registered — apply()/restore() unimplemented in fault_injection.py, a pre-existing gap).
- No formal backtests have run against the real device yet (BLOCKED_HARDWARE) — the probe exists but has 0 measured runs today.
04 — O que voce recebe
O que voce recebe
- One MonitorRun per run
- Screenshots
- Timings
- Evidence
- Dashboard/API per plan
05 — Limites e cotas
Limites e cotas
Loading...
06 — Autonomia
Autonomia
Designed to run unattended end to end. Not yet measured: zero real backtest runs against the physical device today (BLOCKED_HARDWARE) — the check_type and probe exist, but autonomy has no measured figure yet.
07 — Conformidade
Conformidade
Evidência
Ainda não há alegação publicável para este serviço — os backtests e a evidência do mundo real ainda estão sendo medidos. Veja Limitações conhecidas abaixo.
08 — Limitações conhecidas
Limitações conhecidas
- No formal backtests measured yet.
- Measured today on a MID-range device until a real low-end device is added to the pool: every run honestly declares the observed device tier and whether it meets the minimum tier — a low-end result is never simulated.
- Depends on the internal_research/low-end physical pool, currently limited in size.
- Crash detection is a best-effort logcat-window scan, not a full stack-trace parser.
09 — FAQ
FAQ
Perguntas sobre este serviço
Not yet. No formal backtests or real-world samples exist yet, so it stays fail-closed until that changes.