Package 

Class MiscConfiguration

    • Method Detail

      • getSdkEnabled

         final Boolean getSdkEnabled()

        Server-driven master switch. When false, the SDK refuses to start a new session — startDeliveryByX(...) callbacks fire with com.doorstepai.sdks.tracking.internal.SdkError.SdkDisabled and no newSession request is made. Defaults to true so older config responses (without the field) keep working unchanged. Honored immediately on the next ensureLoaded().

      • getHoldWakeLock

         final Boolean getHoldWakeLock()

        Hold a continuous PARTIAL_WAKE_LOCK for the whole tracking lifetime.

        true (default) = legacy behavior. false = the Play-compliant mode: wake-up batched IMU sensors carry continuity through AP suspend, location callbacks arrive under the platform's brief exempt wake lock, and uploads use the short-duration uploadWakeLockTimeoutMs lock instead. Since March 1, 2026 Google Play enforces against non-exempt partial wake locks held >=2h average screen-off in >5% of sessions (wake locks under a foreground service COUNT), so fleets should be migrated to false once the device class passes the untethered screen-off yield gate (devices with no wake-up IMU variants — see the "CAP" capability log — should stay true).

      • getUploadWakeLockTimeoutMs

         final Long getUploadWakeLockTimeoutMs()

        Timeout (ms) for the short-duration wake lock wrapped around each upload tick when holdWakeLock is false, so an upload that started inside a location-callback wake window isn't truncated by re-suspend. Short bounded locks don't approach the Play enforcement metric. 0 disables the upload lock entirely.

      • getSessionWatchdogEnabled

         final Boolean getSessionWatchdogEnabled()

        Periodic (~15 min) WorkManager watchdog that re-asserts TrackingService when the durable session registry says deliveries are in flight but the service is dead (OEM task killers, FGS start denials). Checked at fire time, so a server flip takes effect without reschedule.

      • getDataPushDelayIntervalSeconds

         final Long getDataPushDelayIntervalSeconds()

        Sensor batch upload timer delay between upload loops, in seconds. Drives the period of DataManager.startUploadTimer — every N seconds the SDK snapshots all in-memory capture buffers into a BatchData, persists it to the pending-batches store, and (when online) drains the store FIFO to the server.

      • getDrainIntervalSeconds

         final Long getDrainIntervalSeconds()

        Minimum seconds between FIFO drains (radio bursts), DECOUPLED from the snapshot tick. Snapshot+persist keeps running every dataPushDelayIntervalSeconds (durability is unchanged — batches hit disk before any send), but the radio only wakes every drainIntervalSeconds (±20% jitter, so a fleet doesn't synchronize) and ships the accumulated batches in one burst. With one HTTPS request per 10 s and an RRC tail of ~10 s, the cellular radio never left the high-power state; batching sends 3-6 windows per burst and lets it idle between. <= dataPushDelayIntervalSeconds disables the decoupling (drain every tick, the old behavior). Byte-pressure and session-stop both force an immediate drain regardless of the interval.

      • getMaxBatchesDrainedPerTick

         final Integer getMaxBatchesDrainedPerTick()

        Maximum number of persisted batches the FIFO drain ships per upload tick. Caps radio time after a long offline window so one tick can't monopolise the network; the rest are picked up on subsequent ticks.

      • getMaxBatchRetries

         final Integer getMaxBatchRetries()

        Max retryable send failures for a single persisted batch before it is dead-lettered (dropped) instead of retried forever — prevents a poison batch from blocking the FIFO head indefinitely.

      • getRetryBaseDelayMs

         final Long getRetryBaseDelayMs()

        Base delay (ms) for the drain's full-jitter exponential backoff: the Nth retry of a row waits a random duration in 0, min(retryMaxDelayMs, base·2^N). A server Retry-After header overrides this when present.

      • getMaxCacheSizePerStream

         final Integer getMaxCacheSizePerStream()

        HARD per-stream cap (samples) on the in-memory capture buffers — drop-oldest ring semantics enforced in DataManager.appendBounded. Server-defined and read live on every append. It MUST cover one drain interval's worth of samples at the maximum expected rate (fastestHz × dataPushDelayIntervalSeconds); the buffer is drained once per push interval, so a value that only covers a shorter interval silently drops data on the fastest streams. At 130 Hz × 30 s ≈ 3900, hence the 4096 default (the old 1024 was sized for a 10 s interval and dropped ~59% of accel once the server moved to 30 s). CAUTION: raise this whenever dataPushDelayIntervalSeconds grows or a high-rate stream (accel/mag/ble) overflows — DataManager.applyConfig logs a "TOO SMALL" warning when this is under-provisioned. Under normal operation buffers never reach the cap; when upload ticks stall (CPU starvation, long suspend) it bounds memory so the SDK can't amplify the host's memory pressure into an OS process kill. Overflow drops are counted in the pipeline heartbeat (droppedOverflow). Also used as the ArrayList.ensureCapacity pre-size on every applyConfig. Floor 64.

      • getBinaryEncoding

         final Boolean getBinaryEncoding()

        When true, the SDK sends DSDP v2 binary payloads (S-011) instead of JSON. Default false so older server builds receive JSON unchanged.

        Legacy flag: prefer payloadFormat which supersedes it. Kept for backward compatibility — if payloadFormat is unset/"json" but binaryEncoding is true, the SDK treats it as PayloadFormat.DSDPV2.

      • getPayloadFormat

         final String getPayloadFormat()

        Selects the dataPush wire format: "json" (verbose JSON+deflate, the safe default), "dsdpv2" (the columnar float32 binary, S-011), or "v3" (the compact columnar quantize+delta+varint protobuf format — telemetry.proto). Unknown values fall back to "json". Roll out per fleet via group; the format is persisted per queued row so a mid-rollout app upgrade drains old JSON rows and new v3 rows correctly.

        A String (not an enum) so an unrecognized server value can't throw during config decode; parse via MiscConfiguration.resolvedPayloadFormat.

      • getDualTimestamp

         final Boolean getDualTimestamp()

        When true, every sample emits both wallClockTimestamp (capture-time unix epoch seconds) and sensorTimestamp (native sensor / boot-relative clock) (S-001). Default false so older server builds receive only the legacy timestamp field.

      • getPendingBatchesMaxBytes

         final Long getPendingBatchesMaxBytes()

        Total byte ceiling for the on-disk pending-batches store. Once exceeded, the pruner drops oldest batches first. Mirrors iOS's maxPendingBatchesTotalBytes = 16 MB but defaults higher on Android because we have private app storage rather than a sandbox file. Set to 0 to disable the byte budget.

      • getPendingBatchesMaxAgeMs

         final Long getPendingBatchesMaxAgeMs()

        Maximum age of a persisted batch before it's pruned. Sensor data older than the default (24 h) is rarely useful for delivery analysis. 0 disables the age budget.

      • getPendingBatchesMaxCount

         final Integer getPendingBatchesMaxCount()

        Hard cap on the number of persisted batches. Mirrors iOS's maxPendingBatchesCount = 240. 0 disables the count budget.