OpenAI SDK (JS/TS)

Wrap the OpenAI JavaScript/TypeScript SDK to automatically trace prompts, completions, latencies, and token usage in Muster.

The Muster JS/TS integration ships a wrapper around the OpenAI SDK that adds observability with a single function call. Wrap your OpenAI() instance once and every call routed through it is captured automatically.

Key Features

The integration automatically tracks:

All prompts and completions (including streaming and function calling)
Latencies and time-to-first-token metrics
API errors
Token usage and costs

Setup

Installation

npm install @langfuse/openai openai

Environment Variables

Add credentials to your .env:

LANGFUSE_SECRET_KEY="sk-lf-..."
LANGFUSE_PUBLIC_KEY="pk-lf-..."
LANGFUSE_BASE_URL="https://app.getmuster.io"   # or your self-hosted Muster URL

OpenTelemetry Initialisation

import { NodeSDK } from "@opentelemetry/sdk-node";
import { LangfuseSpanProcessor } from "@langfuse/otel";

const sdk = new NodeSDK({
  spanProcessors: [new LangfuseSpanProcessor()],
});

sdk.start();

Basic Usage

import OpenAI from "openai";
import { observeOpenAI } from "@langfuse/openai";

const openai = observeOpenAI(new OpenAI());

const res = await openai.chat.completions.create({
  messages: [{ role: "system", content: "Tell me a story about a dog." }],
  model: "gpt-4o",
});

Advanced Features

The langfuseConfig parameter on observeOpenAI supports:

generationName — identify specific generation types
langfusePrompt — link to managed prompts
generationMetadata — add custom metadata
sessionId — track user sessions
userId — identify users
tags — categorise traces

const openai = observeOpenAI(new OpenAI(), {
  generationName: "summary",
  sessionId: "session_abc",
  userId: "user_123",
  tags: ["prod"],
});

Important Notes

Compatible with OpenAI SDK versions ≥4.0.0.
Short-lived applications require manual flushing: await langfuseSpanProcessor.forceFlush().
The OpenAI Assistants API is not supported.
Streamed responses require stream_options: { include_usage: true } to capture token usage.