llmmeter-otel

OpenTelemetry sink for llmmeter. Emits Gen-AI semantic-convention spans + metrics for every recorded LLM call, so they show up alongside the rest of your traces in Jaeger, Tempo, Honeycomb, Datadog, New Relic — anything that speaks OTLP.

import { meter, multiSink, sqliteSink } from "@amit641/llmmeter";
import { otelSink } from "llmmeter-otel";
import { trace } from "@opentelemetry/api";

const ai = meter(openai, {
  sink: multiSink(
    sqliteSink({ filePath: "./.llmmeter/llmmeter.db" }),
    otelSink({ tracer: trace.getTracer("my-app") }),
  ),
});

Set up the OTel SDK as you normally would (@opentelemetry/sdk-node + an OTLP exporter); this sink just adds Gen-AI spans on top.

Span attributes

| Attribute | Source | | --- | --- | | gen_ai.system | record.provider | | gen_ai.operation.name | record.operation | | gen_ai.request.model | record.model | | gen_ai.usage.input_tokens | record.tokens.input | | gen_ai.usage.output_tokens | record.tokens.output | | gen_ai.usage.cached_input_tokens | record.tokens.cachedInput (if present) | | gen_ai.conversation.id | record.conversationId | | enduser.id | record.userId | | llmmeter.cost_usd | record.costUsd | | llmmeter.feature | record.feature | | llmmeter.ttft_ms | record.ttftMs |

Plus metrics: gen_ai.client.token.usage and gen_ai.client.operation.duration histograms.