OpenRouter

Trace OpenRouter chat, Responses API, embeddings, streams, usage, and cost from TypeScript or Python.

The OpenRouter integrations wrap the official OpenRouter SDKs without changing request arguments. Calls remain usable when telemetry.dev is not initialized; instrumentation becomes a no-op.

What it captures

• Chat and Responses API input, model parameters, output, response IDs, finish reasons, errors, and token usage
• Embedding input, models, response IDs, and usage, but never embedding vectors
• Stream output as your application consumes it, including time to first chunk and terminal metadata
• Cost from usage.cost, falling back to upstream inference cost when OpenRouter supplies it

Install

npm i @telemetry-dev/sdk @telemetry-dev/openrouter @openrouter/sdk

pip install telemetry-dev-openrouter

The TypeScript integration supports @openrouter/sdk 1.1.2 through versions before 2. The Python package requires Python 3.10 or later, telemetry-dev 0.2 or later, and openrouter 1.1 through versions before 2.

Quickstart

import { OpenRouter } from "@openrouter/sdk";
import { flush, init, shutdown } from "@telemetry-dev/sdk";
import { wrapOpenRouter } from "@telemetry-dev/openrouter";

init({ serviceName: "openrouter-app" });
const client = wrapOpenRouter(new OpenRouter({ apiKey: process.env.OPENROUTER_API_KEY }));

try {
  const response = await client.chat.send({
    chatRequest: {
      model: "openai/gpt-4o-mini",
      messages: [{ role: "user", content: "Explain OpenTelemetry briefly" }],
    },
  });
  console.log(response);
} finally {
  await flush();
  await shutdown();
}

import os

import telemetry_dev
from openrouter import OpenRouter
from telemetry_dev_openrouter import wrap_open_router

telemetry_dev.init(service_name="openrouter-app")
client = wrap_open_router(OpenRouter(api_key=os.environ["OPENROUTER_API_KEY"]))

try:
    response = client.chat.send(
        model="openai/gpt-4o-mini",
        messages=[{"role": "user", "content": "Explain OpenTelemetry briefly"}],
    )
    print(response)
finally:
    telemetry_dev.flush()
    telemetry_dev.shutdown()

Set TELEMETRY_DEV_API_KEY for telemetry.dev and OPENROUTER_API_KEY for OpenRouter before running either example.

API

Both integrations cover chat send, Responses send, and embeddings generate. Python covers synchronous methods and their *_async variants.

Streaming

Chat and Responses streams remain streams and async iterables. Instrumentation observes only events that your application consumes. Exhaust, close, or cancel a stream to finish its span; a returned stream that is never consumed cannot produce complete terminal telemetry.

The integrations do not add or change stream_options. Streamed usage and cost are recorded only when OpenRouter includes them. They reconstruct supported chat and Responses output from consumed events and retain terminal errors, usage, and cost when available.

TypeScript stream capture is limited to 64 KiB and 1,024 items. Python uses the core SDK capture budget. A partial capture sets telemetry.dev.capture.truncated = true; terminal metadata and usage can still be recorded.

Content and limitations

The core SDK captures model input and output by default. Use its captureInput, captureOutput, and masking options if prompts or responses contain sensitive data. Turning capture off prevents those content fields from being added by the integration, while operational metadata and usage remain.

Only the listed methods are instrumented. TypeScript global instrumentation patches the official SDK’s Chat, Responses, and Embeddings resource classes; standalone generated functions and other endpoints are outside its scope.

Open the trace explorer after a call completes. Refer to the quickstart for API-key and capture configuration.