Public API

This page lists the consumer-facing entrypoints. See configuration and naming for defaults and ports and telemetry records for exact boundary shapes.

Instrumentation

Metrics(name?, options?)

Standard and legacy TypeScript method decorator. It preserves this, arguments, return values, thrown errors, and rejected promises.

Without a name, it uses ClassName.methodName. An explicit name overrides the inferred value. Use Metrics(undefined, options) to infer the name while setting instrumentation options.

measure(name, operation, options?)

Runs and instruments one synchronous or asynchronous callback.

instrumentFunction(name, operation, options?)

Creates an instrumented function while preserving its TypeScript signature and this binding.

configureMetrics(configuration)

Configures the process-wide runtime and returns an idempotent restore function. Configuration supports metrics, logger, clock, and resource-usage ports plus a prefix, shared attributes, custom name formatter, and instrumentation error callback. NodeResourceUsageAdapter is used when no resource port is supplied.

MetricsInstrumenter

Injectable class containing the instrumentation behavior. It provides measure and instrument methods without relying on process-wide state.

resetMetrics()

Creates a fresh default in-memory runtime. It is mainly useful for isolated tests.

metrics

The default inspector exposes snapshot() and clear(). A snapshot contains bounded metric and log records, buffer capacities, and discarded-entry counts. Each default buffer retains 1,000 entries.

Options

InstrumentationOptions supports static attributes and these switches:

• recordCalls, recordDuration, and recordFailures default to enabled.
• logCalls and logFailures default to enabled when a logger is configured.
• captureStackTrace defaults to enabled for failure logs.
• recordCpu and recordMemory are opt-in.

MetricAttributes accepts only strings, numbers, and booleans at the type boundary. Choose bounded-cardinality values suitable for the target backend.

See the complete option and configuration tables in configuration and naming.

Public contract types

Type	Role
MetricsConfiguration	Process-wide or injected configuration
InstrumentationOptions	Per-operation signal switches
MetricAttributes	String, number, or boolean attributes
MetricKind	Closed set of emitted metric kinds
MetricUnit	Count, bytes, milliseconds, or microseconds
MetricMeasurementPrimitives	Serialized metric adapter input
InstrumentationLogPrimitives	Serialized structured log
ResourceUsagePrimitives	Serialized CPU and memory snapshot
MetricsInspectionSnapshot	Default in-memory inspector result
MetricsMethodDecorator	Standard and legacy decorator contract
LegacyMetricsMethodDecorator	Legacy decorator contract
StandardMetricsMethodDecorator	Standard decorator contract

ClockPort, LoggerPort, MetricsPort, and ResourceUsagePort define the outbound boundaries. Their exact methods and record shapes live in ports and telemetry records.

Telemetry objects

• MetricMeasurement carries one metric and serializes through toPrimitives().
• InstrumentationLog carries a call or failure log, exposes isFailure(), and serializes through toPrimitives().
• InstrumentationLogLevel models the called and failed lifecycle values.
• ResourceUsageSnapshot calculates CPU and memory values since an earlier snapshot.

Value Objects

The public technical Value Objects include BufferCapacity, OperationName, ElapsedMilliseconds, CpuMicroseconds, MemoryBytes, MemoryByteDelta, and StackTrace. MetricUnit is a unit type, while metric and resource snapshots are immutable boundary objects. Adapters serialize these objects at output boundaries.

The contracts reference also lists the public error classes and when consumers can receive them.

Dedicated entrypoints

• @haskou/metrics/testing
• @haskou/metrics/adapters/console
• @haskou/metrics/adapters/node
• @haskou/metrics/adapters/prometheus

The contracts reference lists their exported roles.