llumo-inference

A telemetry SDK that initializes Traceloop with a buffered exporter and captures outgoing HTTP spans immediately at module load time.

Install

npm install llumo-inference

Usage

Initialize as early as possible in your app startup:

CommonJS

const { initTelemetry } = require("llumo-inference");

async function bootstrap() {
  await initTelemetry({
    token: process.env.LLUMO_API_KEY,
    flushDelayMs: 6000,
    playgroundName: "my-playground",
  });

  // start your app after telemetry is initialized
}

bootstrap();

ESM / TypeScript

import { initTelemetry } from "llumo-inference";

await initTelemetry({
  token: process.env.LLUMO_API_KEY,
  flushDelayMs: 6000,
  playgroundName: "my-playground",
});

API

initTelemetry(options?)

initTelemetry initializes Traceloop and binds the HTTP instrumentation to the live tracer provider.

| Option | Type | Default | Description | | --- | --- | --- | --- | | baseUrl | string | http://localhost:4455/api/v1/telemetry | Telemetry ingestion endpoint | | token | string | "" | Optional bearer token for telemetry + playground API calls | | flushDelayMs | number | 6000 | Per-trace debounce before export | | playgroundName | string | undefined | Optional playground name resolved via llumo API |

Notes

• HTTP/HTTPS patching is applied synchronously at module load to avoid missing early outgoing calls.
• Internal telemetry/llumo endpoints are ignored by outgoing request hook filters.
• Traces are buffered by traceId, deduplicated by span.id, and exported as grouped payloads.