metrics

Documentation

Overview

Package metrics provides in-memory metrics for the proxy.

Index

• type Metrics
  • func New() *Metrics
  • func (m *Metrics) GetSnapshot() Snapshot
  • func (m *Metrics) RecordDeduplicated()
  • func (m *Metrics) RecordFailure()
  • func (m *Metrics) RecordRateLimited()
  • func (m *Metrics) RecordRequest(streaming bool)
  • func (m *Metrics) RecordSuccess(model string, latency time.Duration)
• type Snapshot
  • func (s Snapshot) CalculateP95() time.Duration
  • func (s Snapshot) CalculateP99() time.Duration

Types

type Metrics

type Metrics struct {
	// contains filtered or unexported fields
}

Metrics holds in-memory metrics for the proxy.

func New

func New() *Metrics

New creates a new metrics instance.

func (*Metrics) GetSnapshot

func (m *Metrics) GetSnapshot() Snapshot

GetSnapshot returns a snapshot of current metrics.

func (*Metrics) RecordDeduplicated

func (m *Metrics) RecordDeduplicated()

RecordDeduplicated records a deduplicated request.

func (*Metrics) RecordFailure

func (m *Metrics) RecordFailure()

RecordFailure records a failed request.

func (*Metrics) RecordRateLimited

func (m *Metrics) RecordRateLimited()

RecordRateLimited records a rate-limited request.

func (*Metrics) RecordRequest

func (m *Metrics) RecordRequest(streaming bool)

RecordRequest records an incoming request.

func (*Metrics) RecordSuccess

func (m *Metrics) RecordSuccess(model string, latency time.Duration)

RecordSuccess records a successful request.

type Snapshot

type Snapshot struct {
	RequestsReceived int64
	RequestsStreamed int64
	RequestsSuccess  int64
	RequestsFailed   int64
	UpstreamCalls    int64
	RateLimited      int64
	Deduplicated     int64
	Latencies        []time.Duration
	ModelCounts      map[string]int64
}

Snapshot represents a point-in-time view of metrics.

func (Snapshot) CalculateP95

func (s Snapshot) CalculateP95() time.Duration

CalculateP95 calculates the p95 latency from the snapshot.

func (Snapshot) CalculateP99

func (s Snapshot) CalculateP99() time.Duration

CalculateP99 calculates the p99 latency from the snapshot.