Class LMKitTelemetry

Namespace: LMKit.Telemetry

Assembly: LM-Kit.NET.dll

OpenTelemetry identifiers for LM-Kit instrumentation. Use these constants when configuring your OpenTelemetry pipeline (e.g., AddSource/AddMeter).

public static class LMKitTelemetry

Inheritance: object

LMKitTelemetry

Inherited Members: object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Fields

ActivitySourceName: The ActivitySource name used by LM-Kit for GenAI spans.

DecodePathClassic: Per-conversation context, decoded by the classic controller.

DecodePathParallel: Leased slot in the shared parallel engine's KV pool.

DecodePathTag: Tag naming which decoder served a request. Two decoders exist and they own their KV cache differently: the classic controller keeps one context per conversation, the parallel engine leases a slot from a shared pool. Which one ran decides how a workload scales with concurrency, so it has to be readable from the metric stream rather than inferred by the caller from request shape.

FallbackReasonTag: Tag naming why a request could not use the parallel engine. Drawn from a small fixed set of sentences, so it is safe to use as a metric dimension.

InstrumentationVersion: LM-Kit instrumentation version emitted on spans/metrics.

MeterName: The Meter name used by LM-Kit for GenAI metrics.

OperationEmbeddings: GenAI operation name for embeddings.

OperationExecuteTool: GenAI operation name for tool execution.

OperationInvokeAgent: GenAI operation name for agent invocation.

OperationTextCompletion: GenAI operation name for text completion.

ProviderName: GenAI provider name used in telemetry.

TokenTypeInput: Token type for input tokens.

TokenTypeOutput: Token type for output tokens.

Methods

EnsureInstrumentsCreated(): Forces eager creation of the LM-Kit Meter and its instruments. Call this once at host startup (e.g. before the first HTTP request is served) so observability consumers using MeterListener see the full instrument catalog immediately instead of only after the first inference. Calling this repeatedly is safe; the instruments are created exactly once.