POST to https://app.arize.com/graphql with an x-api-key header; see Forming calls.
Mutations in this page
createDriftMonitor: Creates a new custom drift monitor.createPerformanceMonitor: Creates a new custom performance monitor.createDataQualityMonitor: Create data quality monitor.patchDriftMonitor: Patch drift monitor.patchPerformanceMonitor: Patch performance monitor.patchDataQualityMonitor: Patch data quality monitor.patchMonitor: Patch monitor.triggerMonitor: Manually trigger a monitordeleteMonitor: Delete monitor.
Reference
createDriftMonitor
Creates a new custom drift monitor.createDriftMonitor(input: CreateDriftMonitorMutationInput!): CreateDriftMonitorMutationPayload
Arguments
CreateDriftMonitorMutationInput!
required
Show CreateDriftMonitorMutationInput fields
Show CreateDriftMonitorMutationInput fields
ID
The id of the space to look up the model name
String
The name of the model to create a monitor for. The use of this field also requires using spaceId.
ID
The model ID of the metric window.
DriftMetric!
required
The drift metric to monitor, ex. psi. One of:
psi, kl, js, ks, euclideanDistance, cosineSimilarity.String!
required
The name of the monitor.
Float
The threshold of the monitor. Used in conjunction with comparison operator to determine if alert condition has been met.
MonitorAutoThresholdInput
Show MonitorAutoThresholdInput fields
Show MonitorAutoThresholdInput fields
Float!
required
A coefficient by which to multiply the standard deviation in the threshold calculation: threshold = mean +/- stdDevMultiplier * stdDev
ComparisonOperator!
required
The operator of the monitor. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.Float
The time window that is used to calculate the metric on every evaluation. Should be able to be represented as an integer hour.
Float
The initial delay from the current time for evaluating the monitor in order to allow for ‘actuals’ to arrive. Generally not as applicable for non-actuals related monitors.
String
The name of the dimension (feature, prediction, actual, tag) to create the Drift Monitor for.
DimensionCategory!
required
The category of the feature label, ex. feature, prediction, actuals. One of:
featureLabel, prediction, actuals, actualScore, actualClass, predictionClass, predictionScore, tag, spanProperty, llmEval, annotation, userAnnotation, modelVersion, batchId.[String!]
The model versions by which to filter the primary dataset. An empty array will fetch all versions
[DimensionFilterInput!]
The filters to apply when calculating the primary dataset (e.g. current data)
Show DimensionFilterInput fields
Show DimensionFilterInput fields
DimensionFilterType!
required
The part of the model to filter on. One of:
featureLabel, tagLabel, predictionValue, actuals, predictionClass, predictionScore, actualClass, actualScore, spanProperty, llmEval.ComparisonOperator!
required
The operator of the filter. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.String
Required when specifying a tag or a feature label filter.
[String!]!
required
The values to match on.
DriftMonitorBaseline
Defines the reference data (e.g. baseline) to use to calculate drift
Show DriftMonitorBaseline fields
Show DriftMonitorBaseline fields
String
The feature or tag label of the comparison dataset. Required if
dimensionCategory is set to feature or tagDimensionCategory
The category of the model dimension, ex. feature, prediction, actuals. One of:
featureLabel, prediction, actuals, actualScore, actualClass, predictionClass, predictionScore, tag, spanProperty, llmEval, annotation, userAnnotation, modelVersion, batchId.Boolean!
required
If true, use the model’s primary baseline to build the comparison dataset.
Float
The moving evaluation window length of the custom baseline. E.g. how much data in time to use to compute the baseline
Float
How far back to look for data to use as the monitor’s baseline
DateTime
The start bookend of the custom baseline’s time range.
DateTime
The end bookend of the custom baseline’s time range.
[String!]
The model versions by which to filter the comparison dataset. An empty array will fetch all versions
[DimensionFilterInput!]
The filters on this comparison dataset. Only valid if using custom baselines.
Show DimensionFilterInput fields
Show DimensionFilterInput fields
DimensionFilterType!
required
The part of the model to filter on. One of:
featureLabel, tagLabel, predictionValue, actuals, predictionClass, predictionScore, actualClass, actualScore, spanProperty, llmEval.ComparisonOperator!
required
The operator of the filter. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.String
Required when specifying a tag or a feature label filter.
[String!]!
required
The values to match on.
[MonitorContactInput!]
How and who to contact when the monitor triggers
Show MonitorContactInput fields
Show MonitorContactInput fields
NotificationChannelType!
required
The notification channel type of the monitor contact. One of:
email, integration.EmailAddress
The email address of the monitor contact. Ex. steve@arize.com
ID
The integration key id of the integration
[WebhookSubscriptionInput!]
Webhooks to notify for this monitor and which events each is subscribed to.
Show WebhookSubscriptionInput fields
Show WebhookSubscriptionInput fields
ID!
required
Relay global ID of the webhook destination (webhooks_destinations).
[WebhookEventName!]!
required
Events to subscribe to (e.g. monitor.triggered, prompt.version.created). One of:
MONITOR_TRIGGERED, MONITOR_CLEARED, MONITOR_NO_DATA, PROMPT_VERSION_CREATED, PROMPT_VERSION_LABELED, PROMPT_VERSION_UNLABELED, EVALUATOR_VERSION_CREATED.String
Monitor notes. Notifications support {{monitor.name}}, {{monitor.id}}, {{resource.name}}, {{metric.name}}, {{metric.value}}, {{threshold.operator}}, {{threshold.value}}, {{evaluation.start}}, {{evaluation.end}}, {{monitor.url}}, {{traces.url}} (tracing projects only) (limited to 1000 characters)
DateTime
The start date of the monitor’s downtime if scheduled.
Int
Duration of downtime for a monitor in number of hours. Between 1 and 1344 hours
Int
Frequency of downtime for a monitor in number of days. Between 1 and 56 days
Boolean
If true, the monitor is scheduled to run at a specific time.
Int
The cadence of the monitor’s scheduled runtime in seconds.
[Int!]
The days of the week the monitor is scheduled to run.
MonitorThresholdMode
The threshold mode of the monitor, either single or range. One of:
single, range.Float
The second threshold of the monitor, if using a range threshold.
ComparisonOperator
The second operator of the monitor, if using a range threshold. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.Float
The second standard deviation multiplier, if using a range threshold. Used to determine the suggested threshold for the monitor.
Int
The lookback seconds for the auto threshold.
MonitorThresholdCalculationType
Calculation used for auto thresholds. Either mean + stdev or median + median absolute deviation (mad). One of:
stdev, mad.Boolean
Whether manual evaluation is enabled.
String
Returns
Monitor!
The created drift monitor.
Show Monitor fields
Show Monitor fields
Scalar fields:
id, name, uri, createdDate, monitorCategory, evaluationIntervalSeconds, evaluatedAt, threshold, thresholdMode, threshold2, latestComputedValue, autoThresholdEnabled, dynamicAutoThresholdEnabled, stdDevMultiplier, stdDevMultiplier2, topKPercentileValue, operator, operator2, driftMetric, dimensionCategory, performanceMetric, dataQualityMetric, positiveClassValue, metricAtRankingKValue, isTriggered, status, isManaged, notificationsEnabled, notes, updatedAt, downtimeStart, downtimeDurationHrs, downtimeFrequencyDays, scheduledRuntimeEnabled, scheduledRuntimeCadenceSeconds, scheduledRuntimeDaysOfWeek, autoThresholdLookbackSeconds, thresholdCalculationType, manualEvaluationEnabled.Object fields (select subfields): space, creator, calculationsWithinTimeRange, stateChangesWithinTimeRange, customMetric, primaryMetricWindow, comparisonMetricWindow, contacts, webhookSubscriptions, auditLog, metricHistory, tags.Example
Only required input fields are shown. Replace<ID> and <string> placeholders with real values; the object graph page shows how to look IDs up.
mutation CreateDriftMonitor($input: CreateDriftMonitorMutationInput!) {
createDriftMonitor(input: $input) {
monitor { id name status }
}
}
{
"input": {
"driftMetric": "psi",
"name": "<string>",
"operator": "greaterThan",
"dimensionCategory": "featureLabel"
}
}
createPerformanceMonitor
Creates a new custom performance monitor.createPerformanceMonitor(input: CreatePerformanceMonitorMutationInput!): CreatePerformanceMonitorMutationPayload
Arguments
CreatePerformanceMonitorMutationInput!
required
Show CreatePerformanceMonitorMutationInput fields
Show CreatePerformanceMonitorMutationInput fields
ID
The id of the space to look up the model name
String
The name of the model to create a monitor for. The use of this field also requires using spaceId.
ID
The id of the model to create the Performance Monitor for.
String!
required
A name that describes the performance monitor
PerformanceMetric
The performance metric to monitor. E.x. ‘accuracy’. The metric must be supported by the model type. One of:
accuracy, precision, recall, f_1, sensitivity, specificity, falsePositiveRate, falseNegativeRate, falseNegativeDensity, mse, rmse, mae, mape, predictionAverage, actualsAverage, auc, logLoss, rSquared, prAuc, meanError, calibration, ndcg, recallParity, falsePositiveRateParity, disparateImpact, wape, smape, mase, udf, groupAuc, mrr, rankingMap, rankingRecall, rankingPrecision, microAveragedPrecision, microAveragedRecall, multiClassPrecision, multiClassRecall, macroAveragedPrecision, macroAveragedRecall.ID
The id of the custom metric ID to create the Performance Monitor for.
String
The positive class used for classification metrics. This only applies to performance monitors for classification models.
String
The prediction class used for classification metrics. This only applies to performance monitors for multi-class models.
Int
The @K value for the performance metric. Model must be ranking.
ComparisonOperator!
required
The operator of the performance monitor. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.Float
The explicit threshold value of the monitor. If absent, the monitor will fallback to automatic threshold calculations.
MonitorAutoThresholdInput
Show MonitorAutoThresholdInput fields
Show MonitorAutoThresholdInput fields
Float!
required
A coefficient by which to multiply the standard deviation in the threshold calculation: threshold = mean +/- stdDevMultiplier * stdDev
[String!]
The model versions by which to filter the model’s inferences. An empty array will fetch all versions
[DimensionFilterInput!]
The filters on the model’s inferences. (ex. narrow the monitor to just the case where state=CA)
Show DimensionFilterInput fields
Show DimensionFilterInput fields
DimensionFilterType!
required
The part of the model to filter on. One of:
featureLabel, tagLabel, predictionValue, actuals, predictionClass, predictionScore, actualClass, actualScore, spanProperty, llmEval.ComparisonOperator!
required
The operator of the filter. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.String
Required when specifying a tag or a feature label filter.
[String!]!
required
The values to match on.
Float
The time window that is used to calculate the metric on every evaluation. Should be able to be represented as an integer hour.
Float
The initial delay from the current time for evaluating the monitor in order to allow for ‘actuals’ to arrive. Generally not as applicable for non-actuals related monitors.
[MonitorContactInput!]
How and who to contact when the monitor triggers
Show MonitorContactInput fields
Show MonitorContactInput fields
NotificationChannelType!
required
The notification channel type of the monitor contact. One of:
email, integration.EmailAddress
The email address of the monitor contact. Ex. steve@arize.com
ID
The integration key id of the integration
[WebhookSubscriptionInput!]
Webhooks to notify for this monitor and which events each is subscribed to.
Show WebhookSubscriptionInput fields
Show WebhookSubscriptionInput fields
ID!
required
Relay global ID of the webhook destination (webhooks_destinations).
[WebhookEventName!]!
required
Events to subscribe to (e.g. monitor.triggered, prompt.version.created). One of:
MONITOR_TRIGGERED, MONITOR_CLEARED, MONITOR_NO_DATA, PROMPT_VERSION_CREATED, PROMPT_VERSION_LABELED, PROMPT_VERSION_UNLABELED, EVALUATOR_VERSION_CREATED.String
Monitor notes. Notifications support {{monitor.name}}, {{monitor.id}}, {{resource.name}}, {{metric.name}}, {{metric.value}}, {{threshold.operator}}, {{threshold.value}}, {{evaluation.start}}, {{evaluation.end}}, {{monitor.url}}, {{traces.url}} (tracing projects only) (limited to 1000 characters)
DateTime
The start date of the monitor’s downtime if scheduled.
Int
Duration of downtime for a monitor in number of hours. Between 1 and 1344 hours
Int
Frequency of downtime for a monitor in number of days. Between 1 and 56 days
Boolean
If true, the monitor is scheduled to run at a specific time.
Int
The cadence of the monitor’s scheduled runtime in seconds.
[Int!]
The days of the week the monitor is scheduled to run.
ModelEnvironmentName
The environment for which to retrieve the model schema for. One of:
production, validation, training, tracing.MonitorThresholdMode
The threshold mode of the monitor, either single or range. One of:
single, range.Float
The second threshold of the monitor, if using a range threshold.
ComparisonOperator
The second operator of the monitor, if using a range threshold. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.Float
The second standard deviation multiplier, if using a range threshold. Used to determine the suggested threshold for the monitor.
Int
The lookback seconds for the auto threshold.
MonitorThresholdCalculationType
Calculation used for auto thresholds. Either mean + stdev or median + median absolute deviation (mad). One of:
stdev, mad.Boolean
Whether manual evaluation is enabled.
String
Returns
Monitor!
The created performance monitor.
Show Monitor fields
Show Monitor fields
Scalar fields:
id, name, uri, createdDate, monitorCategory, evaluationIntervalSeconds, evaluatedAt, threshold, thresholdMode, threshold2, latestComputedValue, autoThresholdEnabled, dynamicAutoThresholdEnabled, stdDevMultiplier, stdDevMultiplier2, topKPercentileValue, operator, operator2, driftMetric, dimensionCategory, performanceMetric, dataQualityMetric, positiveClassValue, metricAtRankingKValue, isTriggered, status, isManaged, notificationsEnabled, notes, updatedAt, downtimeStart, downtimeDurationHrs, downtimeFrequencyDays, scheduledRuntimeEnabled, scheduledRuntimeCadenceSeconds, scheduledRuntimeDaysOfWeek, autoThresholdLookbackSeconds, thresholdCalculationType, manualEvaluationEnabled.Object fields (select subfields): space, creator, calculationsWithinTimeRange, stateChangesWithinTimeRange, customMetric, primaryMetricWindow, comparisonMetricWindow, contacts, webhookSubscriptions, auditLog, metricHistory, tags.Example
Only required input fields are shown. Replace<ID> and <string> placeholders with real values; the object graph page shows how to look IDs up.
mutation CreatePerformanceMonitor($input: CreatePerformanceMonitorMutationInput!) {
createPerformanceMonitor(input: $input) {
monitor { id name status }
}
}
{
"input": {
"name": "<string>",
"operator": "greaterThan"
}
}
createDataQualityMonitor
createDataQualityMonitor(input: CreateDataQualityMonitorMutationInput!): CreateDataQualityMonitorMutationPayload
Arguments
CreateDataQualityMonitorMutationInput!
required
Show CreateDataQualityMonitorMutationInput fields
Show CreateDataQualityMonitorMutationInput fields
ID
The id of the space to look up the model name
String
The name of the model to create a monitor for. The use of this field also requires using spaceId.
ID
The id of the model to create the Data Quality Monitor for.
String!
required
A name that describes the data quality monitor
String
The feature label to create the Data Quality Monitor on.
DimensionCategory!
required
The category of the feature label, ex. feature, prediction, actuals. One of:
featureLabel, prediction, actuals, actualScore, actualClass, predictionClass, predictionScore, tag, spanProperty, llmEval, annotation, userAnnotation, modelVersion, batchId.DataQualityMetric!
required
The data quality metric of the monitor, ex. count, avg, etc. One of:
avg, count, sum, percentEmpty, cardinality, standardDeviation, p50, p95, p99, p99_9, newValues, missingValues, averageStringListLength.Float
The time window that is used to calculate the metric on every evaluation. Should be able to be represented as an integer hour.
Float
The initial delay from the current time for evaluating the monitor in order to allow for ‘actuals’ to arrive. Generally not as applicable for non-actuals related monitors.
Float
The threshold of the monitor. Used in conjunction with comparison operator to determine if alert condition has been met.
MonitorAutoThresholdInput
Show MonitorAutoThresholdInput fields
Show MonitorAutoThresholdInput fields
Float!
required
A coefficient by which to multiply the standard deviation in the threshold calculation: threshold = mean +/- stdDevMultiplier * stdDev
ComparisonOperator!
required
The operator of the data quality monitor. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.[String!]
The model versions by which to filter the model’s inferences. An empty array will fetch all versions
[DimensionFilterInput!]
The filters on the model’s inferences. (ex. narrow the monitor to just the case where state=CA)
Show DimensionFilterInput fields
Show DimensionFilterInput fields
DimensionFilterType!
required
The part of the model to filter on. One of:
featureLabel, tagLabel, predictionValue, actuals, predictionClass, predictionScore, actualClass, actualScore, spanProperty, llmEval.ComparisonOperator!
required
The operator of the filter. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.String
Required when specifying a tag or a feature label filter.
[String!]!
required
The values to match on.
[MonitorContactInput!]
How and who to contact when the monitor triggers
Show MonitorContactInput fields
Show MonitorContactInput fields
NotificationChannelType!
required
The notification channel type of the monitor contact. One of:
email, integration.EmailAddress
The email address of the monitor contact. Ex. steve@arize.com
ID
The integration key id of the integration
[WebhookSubscriptionInput!]
Webhooks to notify for this monitor and which events each is subscribed to.
Show WebhookSubscriptionInput fields
Show WebhookSubscriptionInput fields
ID!
required
Relay global ID of the webhook destination (webhooks_destinations).
[WebhookEventName!]!
required
Events to subscribe to (e.g. monitor.triggered, prompt.version.created). One of:
MONITOR_TRIGGERED, MONITOR_CLEARED, MONITOR_NO_DATA, PROMPT_VERSION_CREATED, PROMPT_VERSION_LABELED, PROMPT_VERSION_UNLABELED, EVALUATOR_VERSION_CREATED.String
Monitor notes. Notifications support {{monitor.name}}, {{monitor.id}}, {{resource.name}}, {{metric.name}}, {{metric.value}}, {{threshold.operator}}, {{threshold.value}}, {{evaluation.start}}, {{evaluation.end}}, {{monitor.url}}, {{traces.url}} (tracing projects only) (limited to 1000 characters)
DataQualityMonitorBaseline
Show DataQualityMonitorBaseline fields
Show DataQualityMonitorBaseline fields
String
The feature or tag label of the comparison dataset. Required if
dimensionCategory is set to feature or tagDimensionCategory
The category of the model dimension, ex. feature, prediction, actuals. One of:
featureLabel, prediction, actuals, actualScore, actualClass, predictionClass, predictionScore, tag, spanProperty, llmEval, annotation, userAnnotation, modelVersion, batchId.Boolean!
required
If true, use the model’s primary baseline to build the comparison dataset.
Float
The moving evaluation window length of the custom baseline. E.g. how much data in time to use to compute the baseline
Float
How far back to look for data to use as the monitor’s baseline
DateTime
The start bookend of the custom baseline’s time range.
DateTime
The end bookend of the custom baseline’s time range.
[String!]
The model versions by which to filter the comparison dataset. An empty array will fetch all versions
[DimensionFilterInput!]
The filters on this comparison dataset. Only valid if using custom baselines.
Show DimensionFilterInput fields
Show DimensionFilterInput fields
DimensionFilterType!
required
The part of the model to filter on. One of:
featureLabel, tagLabel, predictionValue, actuals, predictionClass, predictionScore, actualClass, actualScore, spanProperty, llmEval.ComparisonOperator!
required
The operator of the filter. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.String
Required when specifying a tag or a feature label filter.
[String!]!
required
The values to match on.
DateTime
The start date of the monitor’s downtime if scheduled.
Int
Duration of downtime for a monitor in number of hours. Between 1 and 1344 hours
Int
Frequency of downtime for a monitor in number of days. Between 1 and 56 days
Boolean
If true, the monitor is scheduled to run at a specific time.
Int
The cadence of the monitor’s scheduled runtime in seconds.
[Int!]
The days of the week the monitor is scheduled to run.
MonitorThresholdMode
The threshold mode of the monitor, either single or range. One of:
single, range.Float
The second threshold of the monitor, if using a range threshold.
Float
The second standard deviation multiplier, if using a range threshold. Used to determine the suggested threshold for the monitor.
ComparisonOperator
The second operator of the monitor, if using a range threshold. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.ModelEnvironmentName
The environment for which to retrieve the model schema for. One of:
production, validation, training, tracing.Int
The lookback seconds for the auto threshold.
MonitorThresholdCalculationType
Calculation used for auto thresholds. Either mean + stdev or median + median absolute deviation (mad). One of:
stdev, mad.Boolean
Whether manual evaluation is enabled.
String
Returns
Monitor
The created data quality monitor.
Show Monitor fields
Show Monitor fields
Scalar fields:
id, name, uri, createdDate, monitorCategory, evaluationIntervalSeconds, evaluatedAt, threshold, thresholdMode, threshold2, latestComputedValue, autoThresholdEnabled, dynamicAutoThresholdEnabled, stdDevMultiplier, stdDevMultiplier2, topKPercentileValue, operator, operator2, driftMetric, dimensionCategory, performanceMetric, dataQualityMetric, positiveClassValue, metricAtRankingKValue, isTriggered, status, isManaged, notificationsEnabled, notes, updatedAt, downtimeStart, downtimeDurationHrs, downtimeFrequencyDays, scheduledRuntimeEnabled, scheduledRuntimeCadenceSeconds, scheduledRuntimeDaysOfWeek, autoThresholdLookbackSeconds, thresholdCalculationType, manualEvaluationEnabled.Object fields (select subfields): space, creator, calculationsWithinTimeRange, stateChangesWithinTimeRange, customMetric, primaryMetricWindow, comparisonMetricWindow, contacts, webhookSubscriptions, auditLog, metricHistory, tags.Example
Only required input fields are shown. Replace<ID> and <string> placeholders with real values; the object graph page shows how to look IDs up.
mutation CreateDataQualityMonitor($input: CreateDataQualityMonitorMutationInput!) {
createDataQualityMonitor(input: $input) {
monitor { id name status }
}
}
{
"input": {
"name": "<string>",
"dimensionCategory": "featureLabel",
"dataQualityMetric": "avg",
"operator": "greaterThan"
}
}
patchDriftMonitor
patchDriftMonitor(input: PatchDriftMonitorInput!): PatchDriftMonitorPayload
Arguments
PatchDriftMonitorInput!
required
Show PatchDriftMonitorInput fields
Show PatchDriftMonitorInput fields
ID!
required
The ID of the monitor to patch
DriftMonitorPatchInput!
required
Patch of the monitor calculation
Show DriftMonitorPatchInput fields
Show DriftMonitorPatchInput fields
String
A name that describes the monitor
Float
Adjust the monitor’s alerting threshold
ComparisonOperator
Adjust the monitor’s alerting operator. Used in conjunction with the threshold to form an alert condition One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.Boolean
Whether the current monitor threshold value is re-calculated on a regular schedule and thus have a dynamic value that changes over time. Default is false
MonitorAutoThresholdInput
Show MonitorAutoThresholdInput fields
Show MonitorAutoThresholdInput fields
Float!
required
A coefficient by which to multiply the standard deviation in the threshold calculation: threshold = mean +/- stdDevMultiplier * stdDev
Boolean
Whether or not notifications are sent out when the monitor is triggered
[MonitorContactInput!]
The contacts of the monitor.
Show MonitorContactInput fields
Show MonitorContactInput fields
NotificationChannelType!
required
The notification channel type of the monitor contact. One of:
email, integration.EmailAddress
The email address of the monitor contact. Ex. steve@arize.com
ID
The integration key id of the integration
[String!]
An array of the model versions by which to filter the dataset. An empty array will fetch all versions
[DimensionFilterInput!]
The filters on the model’s inferences. (ex. narrow the monitor to just the case where state=CA). If this property is provided it will replace any existing filters.
Show DimensionFilterInput fields
Show DimensionFilterInput fields
DimensionFilterType!
required
The part of the model to filter on. One of:
featureLabel, tagLabel, predictionValue, actuals, predictionClass, predictionScore, actualClass, actualScore, spanProperty, llmEval.ComparisonOperator!
required
The operator of the filter. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.String
Required when specifying a tag or a feature label filter.
[String!]!
required
The values to match on.
String
Notes on the monitor
PositiveInt
The number of seconds to wait before starting the monitor. Between 0 and 15552000 (180 days)
DateTime
The start date and hour of the monitor’s downtime if scheduled (ex: 2024-12-01T10:00:00Z).
PositiveInt
Duration of downtime for a monitor in number of hours. Between 1 and 1344 hours
PositiveInt
Frequency of downtime for a monitor in number of days. Between 1 and 56 days
Boolean
If true, the monitor is scheduled to run at a specific time.
Int
The cadence of the monitor’s scheduled runtime in seconds.
[Int!]
The days of the week the monitor is scheduled to run.
Float
The second threshold of the monitor, if using a range threshold.
ComparisonOperator
The second operator of the monitor, if using a range threshold. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.Float
The second standard deviation multiplier, if using a range threshold. Used to determine the suggested threshold for the monitor.
MonitorThresholdMode
The threshold mode of the monitor, either single or range. One of:
single, range.Int
The lookback seconds for the auto threshold.
MonitorThresholdCalculationType
Calculation used for auto thresholds. Either mean + stdev or median + median absolute deviation (mad). One of:
stdev, mad.Boolean
Whether manual evaluation is enabled.
Float
The evaluation window length of the monitor in seconds.
Float
The moving window length of the monitor in seconds.
Float
The moving window delay of the monitor in seconds.
String
Returns
Monitor!
The updated drift monitor.
Show Monitor fields
Show Monitor fields
Scalar fields:
id, name, uri, createdDate, monitorCategory, evaluationIntervalSeconds, evaluatedAt, threshold, thresholdMode, threshold2, latestComputedValue, autoThresholdEnabled, dynamicAutoThresholdEnabled, stdDevMultiplier, stdDevMultiplier2, topKPercentileValue, operator, operator2, driftMetric, dimensionCategory, performanceMetric, dataQualityMetric, positiveClassValue, metricAtRankingKValue, isTriggered, status, isManaged, notificationsEnabled, notes, updatedAt, downtimeStart, downtimeDurationHrs, downtimeFrequencyDays, scheduledRuntimeEnabled, scheduledRuntimeCadenceSeconds, scheduledRuntimeDaysOfWeek, autoThresholdLookbackSeconds, thresholdCalculationType, manualEvaluationEnabled.Object fields (select subfields): space, creator, calculationsWithinTimeRange, stateChangesWithinTimeRange, customMetric, primaryMetricWindow, comparisonMetricWindow, contacts, webhookSubscriptions, auditLog, metricHistory, tags.Example
Only required input fields are shown. Replace<ID> and <string> placeholders with real values; the object graph page shows how to look IDs up.
mutation PatchDriftMonitor($input: PatchDriftMonitorInput!) {
patchDriftMonitor(input: $input) {
monitor { id name status }
}
}
{
"input": {
"monitorId": "<ID>",
"set": {}
}
}
patchPerformanceMonitor
patchPerformanceMonitor(input: PatchPerformanceMonitorInput!): PatchPerformanceMonitorPayload
Arguments
PatchPerformanceMonitorInput!
required
Show PatchPerformanceMonitorInput fields
Show PatchPerformanceMonitorInput fields
ID!
required
The ID of the monitor to patch
PerformanceMonitorPatchInput!
required
Patch of the monitor calculation
Show PerformanceMonitorPatchInput fields
Show PerformanceMonitorPatchInput fields
String
A name that describes the monitor
Float
Adjust the monitor’s alerting threshold
ComparisonOperator
Adjust the monitor’s alerting operator. Used in conjunction with the threshold to form an alert condition One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.Boolean
Whether the current monitor threshold value is re-calculated on a regular schedule and thus have a dynamic value that changes over time. Default is false
MonitorAutoThresholdInput
Show MonitorAutoThresholdInput fields
Show MonitorAutoThresholdInput fields
Float!
required
A coefficient by which to multiply the standard deviation in the threshold calculation: threshold = mean +/- stdDevMultiplier * stdDev
Boolean
Whether or not notifications are sent out when the monitor is triggered
[MonitorContactInput!]
The contacts of the monitor.
Show MonitorContactInput fields
Show MonitorContactInput fields
NotificationChannelType!
required
The notification channel type of the monitor contact. One of:
email, integration.EmailAddress
The email address of the monitor contact. Ex. steve@arize.com
ID
The integration key id of the integration
[String!]
An array of the model versions by which to filter the dataset. An empty array will fetch all versions
[DimensionFilterInput!]
The filters on the model’s inferences. (ex. narrow the monitor to just the case where state=CA). If this property is provided it will replace any existing filters.
Show DimensionFilterInput fields
Show DimensionFilterInput fields
DimensionFilterType!
required
The part of the model to filter on. One of:
featureLabel, tagLabel, predictionValue, actuals, predictionClass, predictionScore, actualClass, actualScore, spanProperty, llmEval.ComparisonOperator!
required
The operator of the filter. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.String
Required when specifying a tag or a feature label filter.
[String!]!
required
The values to match on.
String
Notes on the monitor
PositiveInt
The number of seconds to wait before starting the monitor. Between 0 and 15552000 (180 days)
DateTime
The start date and hour of the monitor’s downtime if scheduled (ex: 2024-12-01T10:00:00Z).
PositiveInt
Duration of downtime for a monitor in number of hours. Between 1 and 1344 hours
PositiveInt
Frequency of downtime for a monitor in number of days. Between 1 and 56 days
Boolean
If true, the monitor is scheduled to run at a specific time.
Int
The cadence of the monitor’s scheduled runtime in seconds.
[Int!]
The days of the week the monitor is scheduled to run.
Float
The second threshold of the monitor, if using a range threshold.
ComparisonOperator
The second operator of the monitor, if using a range threshold. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.Float
The second standard deviation multiplier, if using a range threshold. Used to determine the suggested threshold for the monitor.
MonitorThresholdMode
The threshold mode of the monitor, either single or range. One of:
single, range.Int
The lookback seconds for the auto threshold.
MonitorThresholdCalculationType
Calculation used for auto thresholds. Either mean + stdev or median + median absolute deviation (mad). One of:
stdev, mad.Boolean
Whether manual evaluation is enabled.
Float
The evaluation window length of the monitor in seconds.
Float
The moving window length of the monitor in seconds.
Float
The moving window delay of the monitor in seconds.
String
Returns
Monitor!
The updated performance monitor.
Show Monitor fields
Show Monitor fields
Scalar fields:
id, name, uri, createdDate, monitorCategory, evaluationIntervalSeconds, evaluatedAt, threshold, thresholdMode, threshold2, latestComputedValue, autoThresholdEnabled, dynamicAutoThresholdEnabled, stdDevMultiplier, stdDevMultiplier2, topKPercentileValue, operator, operator2, driftMetric, dimensionCategory, performanceMetric, dataQualityMetric, positiveClassValue, metricAtRankingKValue, isTriggered, status, isManaged, notificationsEnabled, notes, updatedAt, downtimeStart, downtimeDurationHrs, downtimeFrequencyDays, scheduledRuntimeEnabled, scheduledRuntimeCadenceSeconds, scheduledRuntimeDaysOfWeek, autoThresholdLookbackSeconds, thresholdCalculationType, manualEvaluationEnabled.Object fields (select subfields): space, creator, calculationsWithinTimeRange, stateChangesWithinTimeRange, customMetric, primaryMetricWindow, comparisonMetricWindow, contacts, webhookSubscriptions, auditLog, metricHistory, tags.Example
Only required input fields are shown. Replace<ID> and <string> placeholders with real values; the object graph page shows how to look IDs up.
mutation PatchPerformanceMonitor($input: PatchPerformanceMonitorInput!) {
patchPerformanceMonitor(input: $input) {
monitor { id name status }
}
}
{
"input": {
"monitorId": "<ID>",
"set": {}
}
}
patchDataQualityMonitor
patchDataQualityMonitor(input: PatchDataQualityMonitorInput!): PatchDataQualityMonitorPayload
Arguments
PatchDataQualityMonitorInput!
required
Show PatchDataQualityMonitorInput fields
Show PatchDataQualityMonitorInput fields
ID!
required
The ID of the monitor to patch
DataQualityMonitorPatchInput!
required
Patch of the monitor calculation
Show DataQualityMonitorPatchInput fields
Show DataQualityMonitorPatchInput fields
String
A name that describes the monitor
Float
Adjust the monitor’s alerting threshold
ComparisonOperator
Adjust the monitor’s alerting operator. Used in conjunction with the threshold to form an alert condition One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.Boolean
Whether the current monitor threshold value is re-calculated on a regular schedule and thus have a dynamic value that changes over time. Default is false
MonitorAutoThresholdInput
Show MonitorAutoThresholdInput fields
Show MonitorAutoThresholdInput fields
Float!
required
A coefficient by which to multiply the standard deviation in the threshold calculation: threshold = mean +/- stdDevMultiplier * stdDev
Boolean
Whether or not notifications are sent out when the monitor is triggered
[MonitorContactInput!]
The contacts of the monitor.
Show MonitorContactInput fields
Show MonitorContactInput fields
NotificationChannelType!
required
The notification channel type of the monitor contact. One of:
email, integration.EmailAddress
The email address of the monitor contact. Ex. steve@arize.com
ID
The integration key id of the integration
[String!]
An array of the model versions by which to filter the dataset. An empty array will fetch all versions
[DimensionFilterInput!]
The filters on the model’s inferences. (ex. narrow the monitor to just the case where state=CA). If this property is provided it will replace any existing filters.
Show DimensionFilterInput fields
Show DimensionFilterInput fields
DimensionFilterType!
required
The part of the model to filter on. One of:
featureLabel, tagLabel, predictionValue, actuals, predictionClass, predictionScore, actualClass, actualScore, spanProperty, llmEval.ComparisonOperator!
required
The operator of the filter. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.String
Required when specifying a tag or a feature label filter.
[String!]!
required
The values to match on.
String
Notes on the monitor
PositiveInt
The number of seconds to wait before starting the monitor. Between 0 and 15552000 (180 days)
DateTime
The start date and hour of the monitor’s downtime if scheduled (ex: 2024-12-01T10:00:00Z).
PositiveInt
Duration of downtime for a monitor in number of hours. Between 1 and 1344 hours
PositiveInt
Frequency of downtime for a monitor in number of days. Between 1 and 56 days
Boolean
If true, the monitor is scheduled to run at a specific time.
Int
The cadence of the monitor’s scheduled runtime in seconds.
[Int!]
The days of the week the monitor is scheduled to run.
Float
The second threshold of the monitor, if using a range threshold.
ComparisonOperator
The second operator of the monitor, if using a range threshold. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.Float
The second standard deviation multiplier, if using a range threshold. Used to determine the suggested threshold for the monitor.
MonitorThresholdMode
The threshold mode of the monitor, either single or range. One of:
single, range.Int
The lookback seconds for the auto threshold.
MonitorThresholdCalculationType
Calculation used for auto thresholds. Either mean + stdev or median + median absolute deviation (mad). One of:
stdev, mad.Boolean
Whether manual evaluation is enabled.
Float
The evaluation window length of the monitor in seconds.
Float
The moving window length of the monitor in seconds.
Float
The moving window delay of the monitor in seconds.
String
Returns
Monitor!
The updated data quality monitor.
Show Monitor fields
Show Monitor fields
Scalar fields:
id, name, uri, createdDate, monitorCategory, evaluationIntervalSeconds, evaluatedAt, threshold, thresholdMode, threshold2, latestComputedValue, autoThresholdEnabled, dynamicAutoThresholdEnabled, stdDevMultiplier, stdDevMultiplier2, topKPercentileValue, operator, operator2, driftMetric, dimensionCategory, performanceMetric, dataQualityMetric, positiveClassValue, metricAtRankingKValue, isTriggered, status, isManaged, notificationsEnabled, notes, updatedAt, downtimeStart, downtimeDurationHrs, downtimeFrequencyDays, scheduledRuntimeEnabled, scheduledRuntimeCadenceSeconds, scheduledRuntimeDaysOfWeek, autoThresholdLookbackSeconds, thresholdCalculationType, manualEvaluationEnabled.Object fields (select subfields): space, creator, calculationsWithinTimeRange, stateChangesWithinTimeRange, customMetric, primaryMetricWindow, comparisonMetricWindow, contacts, webhookSubscriptions, auditLog, metricHistory, tags.Example
Only required input fields are shown. Replace<ID> and <string> placeholders with real values; the object graph page shows how to look IDs up.
mutation PatchDataQualityMonitor($input: PatchDataQualityMonitorInput!) {
patchDataQualityMonitor(input: $input) {
monitor { id name status }
}
}
{
"input": {
"monitorId": "<ID>",
"set": {}
}
}
patchMonitor
patchMonitor(input: PatchMonitorInput!): PatchMonitorPayload
Arguments
PatchMonitorInput!
required
Show PatchMonitorInput fields
Show PatchMonitorInput fields
ID!
required
The ID of the monitor to patch
MonitorPatchInput!
required
Patch of the monitor calculation
Show MonitorPatchInput fields
Show MonitorPatchInput fields
String
A name that describes the monitor
Float
Adjust the monitor’s alerting threshold
ComparisonOperator
Adjust the monitor’s alerting operator. Used in conjunction with the threshold to form an alert condition One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.Boolean
Whether the current monitor threshold value is re-calculated on a regular schedule and thus have a dynamic value that changes over time. Default is false
MonitorAutoThresholdInput
Show MonitorAutoThresholdInput fields
Show MonitorAutoThresholdInput fields
Float!
required
A coefficient by which to multiply the standard deviation in the threshold calculation: threshold = mean +/- stdDevMultiplier * stdDev
Boolean
Whether or not notifications are sent out when the monitor is triggered
[MonitorContactInput!]
The contacts of the monitor.
Show MonitorContactInput fields
Show MonitorContactInput fields
NotificationChannelType!
required
The notification channel type of the monitor contact. One of:
email, integration.EmailAddress
The email address of the monitor contact. Ex. steve@arize.com
ID
The integration key id of the integration
[String!]
An array of the model versions by which to filter the dataset. An empty array will fetch all versions
[DimensionFilterInput!]
The filters on the model’s inferences. (ex. narrow the monitor to just the case where state=CA). If this property is provided it will replace any existing filters.
Show DimensionFilterInput fields
Show DimensionFilterInput fields
DimensionFilterType!
required
The part of the model to filter on. One of:
featureLabel, tagLabel, predictionValue, actuals, predictionClass, predictionScore, actualClass, actualScore, spanProperty, llmEval.ComparisonOperator!
required
The operator of the filter. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.String
Required when specifying a tag or a feature label filter.
[String!]!
required
The values to match on.
String
Notes on the monitor
PositiveInt
The number of seconds to wait before starting the monitor. Between 0 and 15552000 (180 days)
DateTime
The start date and hour of the monitor’s downtime if scheduled (ex: 2024-12-01T10:00:00Z).
PositiveInt
Duration of downtime for a monitor in number of hours. Between 1 and 1344 hours
PositiveInt
Frequency of downtime for a monitor in number of days. Between 1 and 56 days
Boolean
If true, the monitor is scheduled to run at a specific time.
Int
The cadence of the monitor’s scheduled runtime in seconds.
[Int!]
The days of the week the monitor is scheduled to run.
Float
The second threshold of the monitor, if using a range threshold.
ComparisonOperator
The second operator of the monitor, if using a range threshold. One of:
greaterThan, lessThan, equals, notEquals, greaterThanOrEqual, lessThanOrEqual, topN, contains, containsString, similarTo.Float
The second standard deviation multiplier, if using a range threshold. Used to determine the suggested threshold for the monitor.
MonitorThresholdMode
The threshold mode of the monitor, either single or range. One of:
single, range.Int
The lookback seconds for the auto threshold.
MonitorThresholdCalculationType
Calculation used for auto thresholds. Either mean + stdev or median + median absolute deviation (mad). One of:
stdev, mad.Boolean
Whether manual evaluation is enabled.
Float
The evaluation window length of the monitor in seconds.
Float
The moving window length of the monitor in seconds.
Float
The moving window delay of the monitor in seconds.
String
Returns
Monitor!
The updated monitor.
Show Monitor fields
Show Monitor fields
Scalar fields:
id, name, uri, createdDate, monitorCategory, evaluationIntervalSeconds, evaluatedAt, threshold, thresholdMode, threshold2, latestComputedValue, autoThresholdEnabled, dynamicAutoThresholdEnabled, stdDevMultiplier, stdDevMultiplier2, topKPercentileValue, operator, operator2, driftMetric, dimensionCategory, performanceMetric, dataQualityMetric, positiveClassValue, metricAtRankingKValue, isTriggered, status, isManaged, notificationsEnabled, notes, updatedAt, downtimeStart, downtimeDurationHrs, downtimeFrequencyDays, scheduledRuntimeEnabled, scheduledRuntimeCadenceSeconds, scheduledRuntimeDaysOfWeek, autoThresholdLookbackSeconds, thresholdCalculationType, manualEvaluationEnabled.Object fields (select subfields): space, creator, calculationsWithinTimeRange, stateChangesWithinTimeRange, customMetric, primaryMetricWindow, comparisonMetricWindow, contacts, webhookSubscriptions, auditLog, metricHistory, tags.Example
Only required input fields are shown. Replace<ID> and <string> placeholders with real values; the object graph page shows how to look IDs up.
mutation PatchMonitor($input: PatchMonitorInput!) {
patchMonitor(input: $input) {
monitor { id name status }
}
}
{
"input": {
"monitorId": "<ID>",
"set": {}
}
}
triggerMonitor
Manually trigger a monitortriggerMonitor(input: TriggerMonitorInput!): TriggerMonitorPayload
Arguments
Returns
Boolean!
Whether the monitor was successfully triggered
Monitor
The monitor the evaluation was run on
Show Monitor fields
Show Monitor fields
Scalar fields:
id, name, uri, createdDate, monitorCategory, evaluationIntervalSeconds, evaluatedAt, threshold, thresholdMode, threshold2, latestComputedValue, autoThresholdEnabled, dynamicAutoThresholdEnabled, stdDevMultiplier, stdDevMultiplier2, topKPercentileValue, operator, operator2, driftMetric, dimensionCategory, performanceMetric, dataQualityMetric, positiveClassValue, metricAtRankingKValue, isTriggered, status, isManaged, notificationsEnabled, notes, updatedAt, downtimeStart, downtimeDurationHrs, downtimeFrequencyDays, scheduledRuntimeEnabled, scheduledRuntimeCadenceSeconds, scheduledRuntimeDaysOfWeek, autoThresholdLookbackSeconds, thresholdCalculationType, manualEvaluationEnabled.Object fields (select subfields): space, creator, calculationsWithinTimeRange, stateChangesWithinTimeRange, customMetric, primaryMetricWindow, comparisonMetricWindow, contacts, webhookSubscriptions, auditLog, metricHistory, tags.Example
Only required input fields are shown. Replace<ID> and <string> placeholders with real values; the object graph page shows how to look IDs up.
mutation TriggerMonitor($input: TriggerMonitorInput!) {
triggerMonitor(input: $input) {
success
monitor { id name status }
}
}
{
"input": {
"monitorId": "<ID>"
}
}
deleteMonitor
deleteMonitor(input: DeleteMonitorMutationInput!): DeleteMonitorMutationPayload
Arguments
Returns
Monitor
Show Monitor fields
Show Monitor fields
Scalar fields:
id, name, uri, createdDate, monitorCategory, evaluationIntervalSeconds, evaluatedAt, threshold, thresholdMode, threshold2, latestComputedValue, autoThresholdEnabled, dynamicAutoThresholdEnabled, stdDevMultiplier, stdDevMultiplier2, topKPercentileValue, operator, operator2, driftMetric, dimensionCategory, performanceMetric, dataQualityMetric, positiveClassValue, metricAtRankingKValue, isTriggered, status, isManaged, notificationsEnabled, notes, updatedAt, downtimeStart, downtimeDurationHrs, downtimeFrequencyDays, scheduledRuntimeEnabled, scheduledRuntimeCadenceSeconds, scheduledRuntimeDaysOfWeek, autoThresholdLookbackSeconds, thresholdCalculationType, manualEvaluationEnabled.Object fields (select subfields): space, creator, calculationsWithinTimeRange, stateChangesWithinTimeRange, customMetric, primaryMetricWindow, comparisonMetricWindow, contacts, webhookSubscriptions, auditLog, metricHistory, tags.Model
Show Model fields
Show Model fields
Scalar fields:
id, name, uri, createdAt, isDemoModel, isRestricted, modelType, modelEnvironmentNames, averagePerformanceMetric, totalTraffic, taskVariables, annotationsEnabled, isAgentExperimentTraceProject, projectType, annotationsDisabled, defaultAnnotationConfigIds.Object fields (select subfields): projectRoleBindings, space, volumeGateVisibility, modelVersions, modelBatches, modelPredictionVolume, modelEnvironments, dimension, dimensionValues, tracingSchema, modelSchema, modelPrimaryBaseline, modelReferenceDataset, modelDashboards, monitors, signalStatus, signalIssues, signalIssueActivityBounds, signalIssueCountsByEntity, signalIssueTimeSeries, monitorAggregations, driftConfigs, performanceConfigs, dataQualityConfigs, modelColumnMappingConfigs, projectSourceMappingConfigs, performanceMetrics, performanceStatus, performanceMetricOverTime, customMetrics, driftStatus, dataQualityStatus, UMAPPoints, spanRecordsPublic, costOverTime, tokensOverTime, costByModel, llmTracingStats, network, tags, _access.Example
Only required input fields are shown. Replace<ID> and <string> placeholders with real values; the object graph page shows how to look IDs up.
mutation DeleteMonitor($input: DeleteMonitorMutationInput!) {
deleteMonitor(input: $input) {
monitor { id name status }
model { id name }
}
}
{
"input": {
"monitorId": "<ID>"
}
}