Okyanus/pomona-sensor-data-checker

Pomona Sensor Data Checker

Upload a sensor log from a greenhouse, grow tent or hydroponic system and see which readings you shouldn't trust:

• —missing values (empty cells, null, or never logged)
• —impossible values (pH outside 3–11, humidity outside 0–100 %, EC outside 0–12 mS/cm…)
• —broken values (err, NaN, text where a number should be)
• —badly timed readings (no timezone, future-dated, stale latest reading, logging gaps, reboots, time going backwards)
• —stuck, flat, noisy or drifting sensors (compared with the previous 11 readings of the same device)

Everything runs in your browser. Files are never uploaded.

Companion Space: Pomona pH Calibration Checker turns pH buffer readings into a calibration and checks that the probe is not wearing out.

Formats

• —One row per reading time: time, temperature, humidity, pH, EC, … (common column names are recognised; optional device_id / zone_id).
• —One row per measurement: timestamp_utc, measurement, value plus optional device_id, zone_id, boot_id, the format written by Pomona's ESP32 SD logger.

Where the rules come from

The checks are the deterministic sensor-quality rules from Pomona (services/model-router/app/sensor_quality.py), ported to JavaScript. The port is tested against the Python rules on 8,638 cases, including Python 3.11's timestamp parsing, so the page gives the same answers as the platform.

Why rules instead of an AI model? We measured it on the Agri Telemetry Sanity Bench: fine-tuned small models scored 0.34–0.68, the rules 1.00.

Advisory only. This tool checks whether data is trustworthy; it does not judge whether growing conditions are good, and it never controls equipment.

Source: okyanu/pomona/spaces/sensor-data-checker