Custom Metrics

Overview

The custom-metrics command group manages custom metrics — user-defined metric expressions evaluated against model data.

Command	Description	Client Method
`custom-metrics list`	List custom metrics	`get_all_custom_metrics`
`custom-metrics get`	Get a metric by name	`get_custom_metric`
`custom-metrics create`	Create a custom metric	`create_custom_metric`
`custom-metrics update`	Update a custom metric	`update_custom_metric`
`custom-metrics delete`	Delete a custom metric	`delete_custom_metric`
`custom-metrics copy`	Copy a metric to another model	`copy_custom_metric`

`custom-metrics list`

arize_toolkit custom-metrics list [--model-name NAME]

Lists custom metrics. If --model-name is provided, lists metrics for that model. If omitted, lists metrics across all models grouped by model name.

Options

--model-name (optional) — Filter to a specific model.

Example

# All metrics for a specific model
arize_toolkit custom-metrics list --model-name "fraud-detection-v3"

# All metrics across all models (grouped output)
arize_toolkit custom-metrics list

`custom-metrics get`

arize_toolkit custom-metrics get METRIC_NAME --model MODEL

Retrieves details for a single custom metric.

Arguments

METRIC_NAME — The metric name.

Options

--model (required) — The model name.

Example

arize_toolkit --json custom-metrics get "avg-confidence" --model "fraud-detection-v3"

`custom-metrics create`

arize_toolkit custom-metrics create METRIC_NAME --metric EXPRESSION --model MODEL [OPTIONS]

Creates a new custom metric for a model.

Arguments

METRIC_NAME — Name for the metric.

Required Options

--metric — The metric expression/formula (e.g. "select avg(prediction) from model").
--model — Model name.

Optional Options

--environment — Environment: tracing, production, staging, development. Defaults to production. For generative use cases, use tracing.
--description — Metric description.

Example

arize_toolkit custom-metrics create "avg-confidence" \
    --metric "select avg(prediction_score) from model" \
    --model "fraud-detection-v3" \
    --environment production \
    --description "Average prediction confidence score"

`custom-metrics update`

arize_toolkit custom-metrics update METRIC_NAME --model MODEL [OPTIONS]

Updates an existing custom metric.

Arguments

METRIC_NAME — Current metric name.

Required Options

--model — Model name.

Optional Options

--new-name — New metric name.
--metric — Updated expression.
--description — Updated description.
--environment — Updated environment.

Example

arize_toolkit custom-metrics update "avg-confidence" \
    --model "fraud-detection-v3" \
    --new-name "avg-pred-confidence" \
    --description "Updated description"

`custom-metrics delete`

arize_toolkit custom-metrics delete METRIC_NAME --model MODEL [--yes]

Deletes a custom metric. Prompts for confirmation unless --yes is passed.

Arguments

METRIC_NAME — The metric name.

Options

--model (required) — Model name.
--yes — Skip confirmation.

Example

arize_toolkit custom-metrics delete "old-metric" --model "fraud-detection-v3" --yes

`custom-metrics copy`

arize_toolkit custom-metrics copy METRIC_NAME --model MODEL [OPTIONS]

Copies a custom metric to another model, optionally renaming it.

Arguments

METRIC_NAME — The source metric name.

Required Options

--model — Source model name.

Optional Options

--new-model — Target model name.
--new-name — Name for the copied metric.
--new-description — Description for the copy.
--new-environment — Environment for the copy. Defaults to production.

Example

arize_toolkit custom-metrics copy "avg-confidence" \
    --model "fraud-detection-v3" \
    --new-model "fraud-detection-v4" \
    --new-name "avg-confidence-v4"