URurbanfox
Metrics

Get fraud-model performance history

Changed on

Retrieve every training-run confusion-matrix snapshot for the platform's fraud-detection models -- full history plus the latest run, not only the latest. Unlike most other /metrics data, this is not tenant-scoped: model quality is measured once at training time and is identical for every tenant. Counts, not precomputed rates: callers derive precision/recall/F1 themselves.

get/v2/{tenant_slug}/metrics/model-performance

Request

Path parameters

tenant_slugstring required

Tenant DNS slug (lowercase letters, digits, and hyphens; 3-100 characters).

Example:demo-retail

Query parameters

modelstring nullable

Filter to a single model's performance history (ml-core project id, e.g. ato_global_sess_clf). Not a fixed enum -- the model roster is defined in ml-core, not here. Omit to return every model's history.

Response

Successful Response

Example response

{
  "data": {
    "model_performance": [
      {
        "class_label": "ato",
        "exported_at": "2026-08-04T12:00:00Z",
        "false_negatives": 5,
        "false_positives": 10,
        "model": "ato_global_sess_clf",
        "model_version": "3",
        "threshold": 0.777,
        "true_negatives": 90,
        "true_positives": 20
      }
    ]
  }
}

Changes