Skip to main content

metrics

Workspace API


Workspace API / services/granite/metrics

services/granite/metrics

Interfaces

BaseMetricLabels

Defined in: services/granite/metrics.ts:33

Base labels for all Granite metrics

Extended by

Indexable

[key: string]: string | number | boolean | undefined

Properties

provider

provider: string

Defined in: services/granite/metrics.ts:34


LLMRequestLabels

Defined in: services/granite/metrics.ts:39

Labels for LLM request metrics

Extends

Indexable

[key: string]: string | number | boolean | undefined

Properties

provider

provider: string

Defined in: services/granite/metrics.ts:34

Inherited from

BaseMetricLabels.provider

model

model: string

Defined in: services/granite/metrics.ts:40

status

status: "timeout" | "error" | "success"

Defined in: services/granite/metrics.ts:41

streaming

streaming: boolean

Defined in: services/granite/metrics.ts:42


LLMTokensLabels

Defined in: services/granite/metrics.ts:46

Labels for LLM token metrics

Extends

Indexable

[key: string]: string | number | boolean | undefined

Properties

provider

provider: string

Defined in: services/granite/metrics.ts:34

Inherited from

BaseMetricLabels.provider

model

model: string

Defined in: services/granite/metrics.ts:47


RetryLabels

Defined in: services/granite/metrics.ts:51

Labels for retry metrics

Extends

Indexable

[key: string]: string | number | boolean | undefined

Properties

provider

provider: string

Defined in: services/granite/metrics.ts:34

Inherited from

BaseMetricLabels.provider

statusCode

statusCode: number

Defined in: services/granite/metrics.ts:52


EmbeddingLabels

Defined in: services/granite/metrics.ts:56

Labels for embedding request metrics

Extends

Indexable

[key: string]: string | number | boolean | undefined

Properties

provider

provider: string

Defined in: services/granite/metrics.ts:34

Inherited from

BaseMetricLabels.provider

model

model: string

Defined in: services/granite/metrics.ts:57


RerankerLabels

Defined in: services/granite/metrics.ts:61

Labels for reranker request metrics

Extends

Indexable

[key: string]: string | number | boolean | undefined

Properties

provider

provider: string

Defined in: services/granite/metrics.ts:34

Inherited from

BaseMetricLabels.provider

model?

optional model: string

Defined in: services/granite/metrics.ts:62

status?

optional status: "timeout" | "error" | "success"

Defined in: services/granite/metrics.ts:63


HealthCheckLabels

Defined in: services/granite/metrics.ts:67

Labels for health check metrics

Extends

  • Record<string, string | number | boolean | undefined>

Indexable

[key: string]: string | number | boolean | undefined

Properties

endpoint

endpoint: string

Defined in: services/granite/metrics.ts:68

Functions

trackLLMRequest()

trackLLMRequest(labels, durationMs): void

Defined in: services/granite/metrics.ts:120

Track LLM request completion

Parameters

labels

LLMRequestLabels

durationMs

number

Returns

void


trackLLMTokensCapped()

trackLLMTokensCapped(labels): void

Defined in: services/granite/metrics.ts:135

Track context token capping (input too large)

Parameters

labels

LLMTokensLabels

Returns

void


trackLLMResponseTokensCapped()

trackLLMResponseTokensCapped(labels): void

Defined in: services/granite/metrics.ts:142

Track response token capping (requested > max)

Parameters

labels

LLMTokensLabels

Returns

void


trackLLMRetry()

trackLLMRetry(labels): void

Defined in: services/granite/metrics.ts:149

Track retry events

Parameters

labels

RetryLabels

Returns

void


trackLLMInputTokens()

trackLLMInputTokens(labels, tokens): void

Defined in: services/granite/metrics.ts:156

Track input token count

Parameters

labels

BaseMetricLabels

tokens

number

Returns

void


trackEmbeddingRequest()

trackEmbeddingRequest(labels, textCount): void

Defined in: services/granite/metrics.ts:167

Track embedding request completion

Parameters

labels

EmbeddingLabels

textCount

number

Returns

void


trackRerankerRequest()

trackRerankerRequest(labels): void

Defined in: services/granite/metrics.ts:181

Track reranker request completion

Parameters

labels

RerankerLabels

Returns

void


trackRerankerTopKCapped()

trackRerankerTopKCapped(labels, requested, capped): void

Defined in: services/granite/metrics.ts:188

Track top-K capping events

Parameters

labels

RerankerLabels

requested

number

capped

number

Returns

void


trackHealthCheckFailure()

trackHealthCheckFailure(labels): void

Defined in: services/granite/metrics.ts:207

Track health check failures

Parameters

labels

HealthCheckLabels

Returns

void


logRequestMetrics()

logRequestMetrics(params): void

Defined in: services/granite/metrics.ts:219

Log a comprehensive metrics summary for a request Use this for detailed observability of individual requests

Parameters

params
operation

"generateText" | "generateTextStream" | "embed" | "rerank" | "chunk"

provider

string

model

string

endpoint

string

inputTokens?

number

maxTokens?

number

maxContextLength?

number

safetyMarginPercent?

number

latencyMs

number

status

"timeout" | "error" | "success"

retries?

number

capped?

boolean

streaming?

boolean

Returns

void