Sensor Data → Millisecond Search

Ingest. Index. Search. Pattern. Alert.

Industrial log and vibration search, built by Slaven Bajac — powered by the Pop DataEngine. Feed sensor data via HTTP batch or MQTT live. The engine processes, indexes, and searches in milliseconds. Your data and index never leave the server.

HTTP batch ingest MQTT live streaming Cross-channel patterns Anomaly detection Spectral analysis Rolling baseline

How it works

Three steps

1

Register

Define your device channels and quantization profile. One API call.

2

Ingest

POST batches or stream via MQTT. The engine indexes every sample.

3

Search

Query by level range, direction, or cross-channel pattern. Millisecond response.

API

Endpoints

REST API. All requests and responses are JSON.

POST /v1/schema — register device + channels
POST /v1/ingest — batch sensor values
POST /v1/search — level range + direction query
POST /v1/pattern — cross-channel pattern match
GET /v1/anomalies?device= — rare state detection
GET /v1/raw?device=&ts= — trigger-retained raw window
GET /v1/health — liveness + stats
DELETE /v1/device/{id} — data deletion (GDPR)

Features

Built for the edge

Dual spectral mode

Frequency-domain fault detection. Two modes: generic (strongest spectral lines) or targeted (pre-declared frequencies). Designed for edge deployment.

📈

Rolling baseline

Per-channel rolling z-score against its own recent history. Regime-robust, drift-sensitive.

🔌

Privacy by design

API-only black box. Client gets positions, timestamps, and their own values. Index and code never leave the server.

Multi-resolution rollup

Level histograms at 5 resolutions. 20 KB vs 130 MB (6415:1) for 120 s @ 20 kHz.

🔥 Patent Pending

The engine layout algorithm is the core IP. It maps multi-channel sensor data into a compact state key, enabling millisecond exact-match search across arbitrary time ranges and cross-channel pattern detection.

Feedback

Report a bug or request a feature

Something not working? Want a new feature? Let us know.