MQTT · Rooms & BLE

Connecting…
No gateway data yet
Door monitor is the first panel on Overview when doors are tracked. Then Live MQTT, floor map, rooms, tags. Connection log last.
Messages
0
Rate /min
0
Topics
0
Rooms (topics)
0

Live MQTT

Buffer 0 / 800 Session total 0 Topics 0 Gateways 0 Rate —/min Last msg Tag estimate
Waiting for MQTT messages. Subscribed to gateway topics.

Floor Map

Each card is a gateway room. Tags listed use smoothed RSSI room assignment. After guided 4-corner calibration, the mini map on each calibrated room shows tag dots at estimated in-room position (RSSI vs corner references).

No gateways detected yet.

Room fingerprints (RSSI)

One profile per room (up to 50). Easiest: use guided calibrationstep 1 tap where the BLE gateway is on the room map (each outlet is different), then four corners: we highlight each corner, you move there and tap one button; we capture 2.5s per corner (same as manual) and merge all four into one room vector. Advanced: single Sample 2.5s still works. With two or more calibrated rooms, room pick uses a joint z-score across rooms (shared μ/σ per gateway) plus a pair margin so two similar rooms do not flip-flop. Reject threshold below still applies to mean-centered dB fit. Server: data/fingerprints.json or DASHBOARD_FINGERPRINTS_PATH.

Uses the room and reference tag you selected below. Step 1 you tap the diagram where this room’s gateway is plugged in (each room differs). Then four corners — one capture button each. The map is a local room frame (not geographic north).

Pick a tracked tag you will wear or hold at each corner during the wizard (same hardware you care about later is best). RSSI shape differs by device; one profile is shared by all tags. You need live packets from that MAC before starting.

Reject if mean-centered RMS > dB

    Rooms (by gateway)

    Room numbers follow the server gateway list order (same MAC → same room index every load). Tag estimated room = smoothed RSSI winner with sticky hysteresis across those gateways.

    No gateways seen yet.
    Connection log (broker & MQTT client)
    No events yet.
    No events yet.

    Threshold fall detection (accelerometer)

    Live signal vector magnitude (SVM) from MOKO-style x_axis_data / y_axis_data / z_axis_data (tracked tags only). Pipeline: optional near–free-fall dip → impact spike (high SVM) → post-impact settle (SVM returns close to 1g and stays tight for a hold time), then a confidence gate from strong impact, free-fall hint, or orientation change. W6 docs say the wristband can broadcast real-time 3-axis accelerometer data for fall recognition; these defaults are a wrist-oriented starting point, not a waist/chest model.

    Tag MAC SVM (g) Phase Peak (g) Alerts
    Waiting for accelerometer rows from tracked tags…

    Event log

    Ingestor, allowlists & forwarding

    Live snapshot from gateway-ingestor (/health via this dashboard). Prod / dev show allowlist versions and MAC counts pushed or reconciled on the ingestor. Forwarding reflects ingestor env (not your API process). Batch and allowlist reconcile URLs below come from the ingestor container env (same as the worker uses for POST / GET). With INGEST_QUEUE_MODE=redis, redis_enqueued means batches reached Redis; this panel also shows stream length, pending entries, and dead-letter count.

    Loading…

    Queue flush removes all pending stream entries and recreates worker consumer groups.

    Prod allowlist

    Dev allowlist

    Forwarding & queue

    Recent ingestor pipeline events

    Time Route Action Count Batch / detail
    Open this tab to load events…

    Tip: to verify HTTP delivery to your API, watch ingest-worker container logs or add metrics on your batch endpoint.