Axon Levant agro-node telemetry customer2-wl-5102 · wildland

How one sensor reading gets from a hillside to a web page

A node in the field measures its surroundings every 30 minutes and sends the result home. Nine things have to happen for that reading to end up on a screen. Press start and watch each one, then scroll the whole dataset — all 3,000 readings — underneath.

Speed

Messages published0

Readings extracted0

Rows in database0

Duplicates skipped0

Fire alarms seen0

The signal chain

Each step is a separate piece of software doing one job. The lamp lights when that step has handled its first reading.

Every reading in the file

All 3,000 readings from customer2_wildland_3000.ndjson, 30 minutes apart, covering 31 July to 1 October 2026. Rows appear as the pipeline ingests them. Click any row to see exactly what the sensor sent.

# sampled_time air °C hum % soil % fire risk wildland batt %

0 of 3,000 readings ingested

Raw payload select a row

Nothing selected yet. Click a row above to see the JSON the node actually published — this is the same text that travels over the network.

What was measured on the real system

The page above replays the dataset in your browser using the same transformation logic the live ingest service runs. These figures come from the actual stack — a real broker, a real Node-RED instance, and a real managed PostgreSQL database.

Rows ingested 3,000 per customer · 6,000 total Both datasets, complete, in one pass.

Duplicates created 0 Forced replays from the top produced no duplicate rows and no errors.

Database PostgreSQL 17.10, 98 columns, 0 extensions Plain Postgres. No time-series extension.

Transport security TLS 1.3 The database refuses unencrypted connections; the broker refuses anonymous clients.

Alarm window preserved 2026-08-25 05:00Z state_code 3, wildland state alarm — matching the source event guide to the minute.

Data lost to the schema None Every field in both datasets maps to a real column, including the wildland model and soil channel 2.

This page carries the dataset inside it, so nothing here is querying the live database — that would need a network connection back to the server. What you are watching is the real data and the real logic; the numbers in this last section are the results recorded when the same run happened against the real stack.