Langfuse Hello World

Overview

Minimal working example demonstrating core Langfuse tracing functionality.

Prerequisites

• Completed

  langfuse-install-auth

  setup
• Valid API credentials configured
• Development environment ready

Instructions

Step 1: Create Entry File

Create a new file for your hello world trace.

Step 2: Import and Initialize Client

import { Langfuse } from "langfuse";

const langfuse = new Langfuse({
  publicKey: process.env.LANGFUSE_PUBLIC_KEY!,
  secretKey: process.env.LANGFUSE_SECRET_KEY!,
  baseUrl: process.env.LANGFUSE_HOST,
});

Step 3: Create Your First Trace

async function helloLangfuse() {
  // Create a trace (top-level operation)
  const trace = langfuse.trace({
    name: "hello-world",
    userId: "demo-user",
    metadata: { source: "hello-world-example" },
    tags: ["demo", "getting-started"],
  });

  // Add a span (child operation)
  const span = trace.span({
    name: "process-input",
    input: { message: "Hello, Langfuse!" },
  });

  // Simulate some processing
  await new Promise((resolve) => setTimeout(resolve, 100));

  // End the span with output
  span.end({
    output: { result: "Processed successfully!" },
  });

  // Add a generation (LLM call tracking)
  trace.generation({
    name: "llm-response",
    model: "gpt-4",
    input: [{ role: "user", content: "Say hello" }],
    output: { content: "Hello! How can I help you today?" },
    usage: {
      promptTokens: 5,
      completionTokens: 10,
      totalTokens: 15,
    },
  });

  // Flush to ensure data is sent
  await langfuse.flushAsync();

  console.log("Trace created! View at:", trace.getTraceUrl());
}

helloLangfuse().catch(console.error);

Output

• Working code file with Langfuse client initialization
• A trace visible in Langfuse dashboard containing:
  • One span with input/output
  • One generation with mock LLM data
• Console output showing:

Trace created! View at: https://cloud.langfuse.com/trace/abc123...

Error Handling

npm list langfuse

flushAsync()

Examples

TypeScript Complete Example

import { Langfuse } from "langfuse";

const langfuse = new Langfuse();

async function main() {
  // Create trace
  const trace = langfuse.trace({
    name: "hello-world",
    input: { query: "What is Langfuse?" },
  });

  // Simulate LLM call
  const generation = trace.generation({
    name: "answer-query",
    model: "gpt-4",
    modelParameters: { temperature: 0.7 },
    input: [
      { role: "system", content: "You are a helpful assistant." },
      { role: "user", content: "What is Langfuse?" },
    ],
  });

  // Simulate response
  await new Promise((r) => setTimeout(r, 500));

  // End generation with output
  generation.end({
    output: "Langfuse is an open-source LLM observability platform...",
    usage: { promptTokens: 25, completionTokens: 50 },
  });

  // Update trace with final output
  trace.update({
    output: { answer: "Langfuse is an LLM observability platform." },
  });

  // Flush and get URL
  await langfuse.flushAsync();
  console.log("View trace:", trace.getTraceUrl());
}

main();

Python Complete Example

from langfuse import Langfuse
import time

langfuse = Langfuse()

def main():
    # Create trace
    trace = langfuse.trace(
        name="hello-world",
        input={"query": "What is Langfuse?"},
        user_id="demo-user",
    )

    # Add a span for processing
    span = trace.span(
        name="process-query",
        input={"query": "What is Langfuse?"},
    )

    # Simulate processing
    time.sleep(0.1)

    span.end(output={"processed": True})

    # Add LLM generation
    generation = trace.generation(
        name="answer-query",
        model="gpt-4",
        model_parameters={"temperature": 0.7},
        input=[
            {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user", "content": "What is Langfuse?"},
        ],
    )

    # Simulate LLM response
    time.sleep(0.5)

    generation.end(
        output="Langfuse is an open-source LLM observability platform...",
        usage={"prompt_tokens": 25, "completion_tokens": 50},
    )

    # Update trace with final output
    trace.update(
        output={"answer": "Langfuse is an LLM observability platform."}
    )

    # Flush data
    langfuse.flush()
    print(f"View trace: {trace.get_trace_url()}")

if __name__ == "__main__":
    main()

With Decorators (Python)

from langfuse.decorators import observe, langfuse_context

@observe()
def process_query(query: str) -> str:
    # This function is automatically traced
    return f"Processed: {query}"

@observe(as_type="generation")
def generate_response(messages: list) -> str:
    # This is tracked as an LLM generation
    langfuse_context.update_current_observation(
        model="gpt-4",
        usage={"prompt_tokens": 10, "completion_tokens": 20},
    )
    return "Hello from Langfuse!"

@observe()
def main():
    result = process_query("Hello!")
    response = generate_response([{"role": "user", "content": "Hi"}])
    return response

main()

Resources

• Langfuse Tracing Concepts
• Langfuse SDK Reference
• Langfuse Examples

Next Steps

Proceed to

langfuse-local-dev-loop