@eka-care/usage-sdk (TypeScript / Node.js)

Non-blocking SDK for pushing usage metrics to Kafka.

One client per process, many workspaces per client.

Install

npm install @eka-care/usage-sdk

Configure

Kafka connection and tuning come from environment variables:

| Env var | Default | Applies to | | ---------------------------- | ------- | ---------------------- | | EKA_KAFKA_BROKERS | — | broker list (required) | | EKA_KAFKA_COMPRESSION_TYPE | lz4 | lz4/gzip/snappy/zstd/none | | EKA_KAFKA_ACKS | 1 | -1 (or all), 0, 1 | | EKA_KAFKA_RETRIES | 5 | producer retry count |

linger.ms and batch.size are librdkafka-only and do not apply to kafkajs. Use flushIntervalMs and maxQueueSize options for TS-side batching tuning.

Usage

import { EkaClient } from "@eka-care/usage-sdk";

const client = new EkaClient({
  serviceName: "scribe-api",
  onError: (e, ctx) => console.error("sdk err", e.message, ctx),
});

await client.connect();

// workspaceId is per-call — one client serves all tenants
client.record("ws_123", "ekascribe", "transcription_minute", 8.2);

// with unit cost and error metadata
client.record("ws_123", "api", "api_error", 1, "error", undefined,
              { error: "upstream timeout", endpoint: "/v1/records" });

process.on("SIGTERM", () => client.shutdown());

Tests

npm install
npm test

Build

npm run build