Monitors

Overview

The monitors command group manages Arize monitors — automated checks that track performance, drift, and data-quality metrics and alert when thresholds are breached.

Command	Description	Client Method
monitors list	List monitors for a model	get_all_monitors
monitors get	Get a monitor by name	get_monitor
monitors create-performance	Create a performance monitor	create_performance_monitor
monitors create-drift	Create a drift monitor	create_drift_monitor
monitors create-data-quality	Create a data quality monitor	create_data_quality_monitor
monitors delete	Delete a monitor	delete_monitor
monitors copy	Copy a monitor to another model	copy_monitor
monitors values	Get metric values over time	get_monitor_metric_values
monitors latest-value	Get the latest metric value	get_latest_monitor_value

---

monitors list

arize_toolkit monitors list [--model-name NAME] [--model-id ID] [--category CATEGORY]

Lists all monitors for a model. Optionally filter by category.

Options

• --model-name (optional) — Model name.
• --model-id (optional) — Model ID.
• --category (optional) — Filter: drift, dataQuality, or performance.

Example

arize_toolkit monitors list --model-name "fraud-detection-v3"
arize_toolkit monitors list --model-name "fraud-detection-v3" --category performance

---

monitors get

arize_toolkit monitors get NAME --model MODEL

Retrieves detailed information for a single monitor.

Arguments

• NAME — The monitor name.

Options

• --model (required) — The model name.

Example

arize_toolkit --json monitors get "accuracy-check" --model "fraud-detection-v3"

---

monitors create-performance

arize_toolkit monitors create-performance NAME --model MODEL --environment ENV [OPTIONS]

Creates a new performance monitor.

Arguments

• NAME — Name for the monitor.

Required Options

• --model — Model name.
• --environment — Environment: tracing, production, validation, or training.

Key Options

• --performance-metric — Metric name (e.g. accuracy, auc). Either this or --custom-metric-id is required.
• --custom-metric-id — Custom metric ID to monitor.
• --operator — Comparison operator: greaterThan, lessThan, greaterThanOrEqual, lessThanOrEqual. Defaults to greaterThan.
• --threshold — Alert threshold value.
• --std-dev-multiplier — Standard deviation multiplier. Defaults to 2.0.
• --evaluation-window — Evaluation window in seconds. Defaults to 259200 (3 days).
• --delay — Delay in seconds before evaluation. Defaults to 0.
• --threshold-mode — single or double. Defaults to single.
• --threshold2 / --operator2 — Second threshold settings (for double mode).
• --email — Email addresses for notifications (repeatable).
• --integration-key-id — Integration key IDs (repeatable).
• --notes — Notes for the monitor.

Example

arize_toolkit monitors create-performance "accuracy-alert" \
    --model "fraud-detection-v3" \
    --environment production \
    --performance-metric accuracy \
    --operator lessThan \
    --threshold 0.95 \
    --email "alerts@example.com"

---

monitors create-drift

arize_toolkit monitors create-drift NAME --model MODEL [OPTIONS]

Creates a new drift monitor.

Arguments

• NAME — Name for the monitor.

Required Options

• --model — Model name.

Key Options

• --drift-metric — Drift metric: psi, js, kl, ks, euclideanDistance, or cosineSimilarity. Defaults to psi.
• --dimension-category — Category to monitor (e.g. prediction, featureLabel). Defaults to prediction.
• --dimension-name — Specific dimension name.
• --operator, --threshold, --std-dev-multiplier, --evaluation-window, --delay, --threshold-mode, --threshold2, --operator2 — Same as performance monitors.
• --email, --integration-key-id, --notes — Notification and documentation options.

Example

arize_toolkit monitors create-drift "prediction-drift" \
    --model "fraud-detection-v3" \
    --drift-metric psi \
    --threshold 0.2

---

monitors create-data-quality

arize_toolkit monitors create-data-quality NAME --model MODEL --data-quality-metric METRIC --environment ENV [OPTIONS]

Creates a new data quality monitor.

Arguments

• NAME — Name for the monitor.

Required Options

• --model — Model name.
• --data-quality-metric — Metric name (e.g. percentEmpty, cardinality, avg).
• --environment — Environment: tracing, production, validation, or training.

Key Options

• --dimension-category — Category to monitor. Defaults to prediction.
• --dimension-name — Specific dimension name.
• --operator, --threshold, --std-dev-multiplier, --evaluation-window, --delay, --threshold-mode, --threshold2, --operator2 — Threshold settings.
• --email, --integration-key-id, --notes — Notification and documentation options.

Example

arize_toolkit monitors create-data-quality "null-check" \
    --model "fraud-detection-v3" \
    --data-quality-metric percentEmpty \
    --environment production \
    --operator greaterThan \
    --threshold 0.05

---

monitors delete

arize_toolkit monitors delete NAME --model MODEL [--yes]

Deletes a monitor. Prompts for confirmation unless --yes is passed.

Arguments

• NAME — The monitor name.

Options

• --model (required) — The model name.
• --yes — Skip confirmation.

Example

arize_toolkit monitors delete "old-monitor" --model "fraud-detection-v3" --yes

---

monitors copy

arize_toolkit monitors copy MONITOR_NAME --model MODEL [--new-name NAME] [--new-model MODEL] [--new-space-id ID]

Copies a monitor to the same or a different model/space.

Arguments

• MONITOR_NAME — The source monitor name.

Options

• --model (required) — Source model name.
• --new-name (optional) — Name for the copied monitor.
• --new-model (optional) — Target model name.
• --new-space-id (optional) — Target space ID.

Example

arize_toolkit monitors copy "accuracy-alert" \
    --model "fraud-detection-v3" \
    --new-model "fraud-detection-v4" \
    --new-name "accuracy-alert-v4"

---

monitors values

arize_toolkit monitors values MONITOR_NAME --model MODEL [--granularity GRANULARITY] [--start-time TIME] [--end-time TIME]

Fetches the monitor's metric values over time.

Arguments

• MONITOR_NAME — The monitor name.

Options

• --model (required) — Model name.
• --granularity — Time series granularity: hour, day, week, month. Defaults to hour.
• --start-time / --end-time (optional) — Time range in ISO format.

Example

arize_toolkit --json monitors values "accuracy-alert" \
    --model "fraud-detection-v3" \
    --granularity day \
    --start-time 2025-01-01

---

monitors latest-value

arize_toolkit monitors latest-value MONITOR_NAME --model MODEL [--granularity GRANULARITY]

Returns the most recent metric value, threshold, and timestamp for a monitor.

Arguments

• MONITOR_NAME — The monitor name.

Options

• --model (required) — Model name.
• --granularity — Granularity: hour, day, week, month. Defaults to hour.

Example

$ arize_toolkit --json monitors latest-value "accuracy-alert" --model "fraud-detection-v3"
{
  "timestamp": "2025-02-24T12:00:00Z",
  "metric_value": 0.97,
  "threshold_value": 0.95
}