โ† All workers

Anomaly Detection

Coming Soon

Score every observation against its own history. Know when something is unusual.

Data StreamStandard

What it does

Maintains a T-Digest per feed source per time window. Each new item gets scored against the model โ€” percentile and z-score-equivalent โ€” and the score attaches as an extension field. Constant space, constant time per update.

Example output

JSONFeed extension fields attached to each processed item.

{
  "_anomaly": {
    "metric": "cloud_cover_oktas",
    "value": 7,
    "percentile": 0.97,
    "z_score": 2.14,
    "anomaly": true,
    "severity": "high"
  }
}

Use cases

  • Solar energy: flag abnormal cloud cover that will choke production
  • Financial: detect volume spikes or price dislocations
  • IoT: sensor readings outside normal operating range
  • Parametric insurance: automatic trigger evaluation
Try it freeยทBrowse all workers