Implementation audit — tickets #24–#37

Audit date: 2026-09-19. Current application build is 0.14.0-trip-data (cd429ac). This audit checks the published ticket criteria against implementation, automated checks and physical TECNO evidence; a closed issue alone is not treated as proof. #37 measurement evidence is diagnostic accounting, with the physical-energy limitation stated explicitly below.

Ticket Criteria and authoritative evidence
#24 C1: bundled HTTPS assets, no PC backend/reverse, device origins and build path. C2: separate category/source/date/unknown-power UI. C3: snapshot counts and invalid coordinates, stable category IDs. C4: public snapshot adapter contains no PLN authentication; availability labelled unknown. C5: catalogue boundary tests and real TECNO map/error/detail evidence. Validation, catalogue.test.mjs, issue24-device.json.
#25 C1: nominal Smart/Long Range selection. C2: charger/payload inputs and editable solo/tandem. C3: finite/range validation. C4: committed native profile with timestamp/manual source. C5: invalid input and force-stop persistence. Validation, journey.test.mjs, issue25-device.json.
#26 C1: GPS denial plus map selection and general search. C2: provider/licence/attribution/limits decision. C3: motorcycle/toll/motorway validation and recorded detour regression. C4: provider geometry/time/source/elevation and unavailable states. C5: controlled adapter/native cancellation tests and actual search/route contracts. Validation, detour comparison, routing.test.mjs, RoutingTransportTest, issue26-device.json, issue26-gps-denied.json.
#27 C1: public Journey application seam and minimal scenario extraction. C2: pp/model/terrain/payload assumptions exposed. C3: immutable actual SOC, 65−60=5 and strict reserve boundaries. C4: missing elevation vs insufficient energy; no regen credit. C5: independent versioned prediction persistence and restart. Validation, energy.test.mjs, issue27-device.json.
#28 C1: partial/90/100 target search and unchanged starting SOC. C2: direct skip connections, travel/charge/setup ranking, bounded search disclosure. C3: supply/profile caps and explicit unknown-supply assumptions. C4: avoidance and conditional infrastructure. C5: destination exit energy and alternative/unreachable scenarios. C6: unverified OMOWAY dealer exclusion. Validation, charge-plan.test.mjs, issue28-device.json.
#29 C1: default >20 with full-charge search first. C2: explicit >10 then >5 choices. C3: persisted plan-scoped consent, new-plan reset and normal alternative. C4: exact10/exact5 fail. C5: stale-response/offer protections and reopening. Validation, reserve-plan.test.mjs, issue29-device.json.
#30 C1: ordered destinations/departure/stays/return and day-grouped itinerary. C2: continuous SOC/time across midnight and reordered recalculation. C3: lodging permission/connection/power required. C4: return or exit energy checked. C5: full leg/charge/exit table and scenario tests. C6: synthetic fixtures, no personal vacation plan embedded as product data. Validation, itinerary.test.mjs, issue30-device.json.
#31 C1: native permission-aware foreground recording lifecycle. C2: actual SOC/profile/legs/GPS/charge event data. C3: immutable pre-leg copies. C4: late/duplicate fix rejection, interruption and explicit resume. C5: TECNO screen-off/lost-location/restart replay and permanent synthetic flag. Validation, TripRecorderTest, issue31-device.json.
#32 C1: position-driven route matching, progress and TTS. C2: native screen-off lifecycle and independent end cleanup. C3: permission, GPS, audio focus, reopening and unavailable voice behavior. C4: explicit off-route uncertainty. C5: deterministic replay plus physical TTS completion/deduplication. Validation, NavigationSessionTest, issue32-device.json, issue32-device-advanced.json.
#33 C1: actual SOC/charge/off-route triggers from current state with reserve retained. C2: charger/target/feasibility notifications; explicit charge completion. C3: cadence, revision checks and atomic publication. C4: offline local actuals and stale result timestamp. C5: shared single location listener and independent owners on TECNO. Validation, live-replan.test.mjs, JourneyTraceTest, issue33-{location,automatic,offroute,offline,worker}-device.json.
#34 C1: real-quality/profile/payload/terrain filters exclude simulations. C2: versioned future prediction with immutable old records and uncalibrated state. C3: actual/predicted errors separate from reserve performance. C4: chronological held-out evaluation, corrupt/incomplete data and restart coverage; visible support/assumptions. C5: insufficient real evidence explicitly unverified; extreme errors remain failures. Validation, calibration.test.mjs, issue34-device.json; revision-bound native publication additionally verified by #35.
#35 C1: selected JSON with units/provenance/time and credential-field exclusion. C2: scoped Android file sharing and failure preservation. C3: explicit confirmation, transactional deletion and prior model invalidation protected against stale publication. C4: actual file parsing, cancel/delete/restart and identical retained trip on TECNO. Validation, TripRecorderTest, journey-native-storage.test.mjs, issue35-device.json.
#36 C1: source/snapshot/attempt timestamps and non-live state. C2: structure/ID/coordinate/completeness checks before durable replacement. C3: legitimate count changes accepted, invalid points excluded. C4: stable source IDs for avoidance, no invented verification. C5: timeout/corruption/interrupted restart/success evidence. Validation, catalogue-store.test.mjs, issue36-device.json.
#37 C1: counterbalanced equal-duration scenarios, device/network/accuracy and USB/modeled-accounting boundary. C2: idle, screen-off active and real replan results, with no unsupported cadence change. C3: separate hardened-tool repeats and sample/freshness reliability disclosures, no physical savings claim. C4: zero post-stop listeners/consumers and no app wake locks on TECNO. Protocol and results.

Evidence scope

Device JSON captures are local ignored artifacts under apps/omox/research/runtime/; the validation documents explain their setup and findings. They have been located and parsed during this audit. They are not substitutes for current automated regression checks. The latest completed suite before #37 has 66 Node and 24 JVM tests, with Android assembly/lint passing.

The dedicated TECNO runs Android 11; Android 13 notification-denial and Android 14–15 foreground-launch behavior have SDK/manifest/lint coverage but no physical-device claim. APK distribution remains a local debug build; release signing or store publication is not part of these tickets. Current commits have not been pushed.

Road prediction accuracy and battery savings remain separate empirical claims. Synthetic replay and modeled energy estimates must not be presented as measured riding accuracy or physical battery savings. Online demo routing has no production SLA; station snapshots do not certify live availability, access, compatibility or delivered power. Full offline navigation and direct vehicle telemetry are outside this ticket range.