README

A PHP SDK for interacting with the Langfuse API. This package provides tools for tracing, scores, annotation queues, dataset runs and prompt management to enable comprehensive LLM observability.

Requirements

• PHP >= 8.1

Installation

composer require mtyrtov/langfuse-php:dev-master

Quick Start

use Langfuse\LangfuseClient;
use Langfuse\LangfuseProfiler;

require_once "vendor/autoload.php";

# Initialize the client
$client = new LangfuseClient(
    'your-public-key',
    'your-secret-key',
    'https://your-langfuse-host.com' # Optional, defaults to http://127.0.0.1:3000
);

$profiler = new LangfuseProfiler($client);

Core Features

Tracing

Traces represent full user sessions or requests in your application.

# Create a trace
$trace = $profiler->trace('user-query')
    ->setSessionId() // Automatically generates a session ID
    ->setUserId('user-123');

# Set input and output
$trace->setInput('Hello, how are you?');
$trace->setOutput("I'm fine, thank you!");

Spans

Spans represent logical sections within a trace, such as specific processing steps.

# Create a span within a trace
$classificationSpan = $trace->span('classification');

# End the span when the operation is complete
$classificationSpan->end();

Generations

Generations track individual LLM calls and their results.

# Create a generation within a span
$generation = $classificationSpan->generation('model-call')
    ->withModel('gpt-4.1-mini')
    ->withModelParameters(['temperature' => 0])
    ->setInput(['role' => 'user', 'content' => 'Hello there'])
    ->setOutput('AI response here');

Flushing Data

Ensure all data is sent to Langfuse before your application terminates.

$result = $profiler->flush();

Laravel

class LangfuseProvider extends ServiceProvider
{
    public function register(): void
    {
        $this->app->singleton(LangfuseClient::class, function ($app) {
            return new LangfuseClient(
                # Your variables from .env
                config('services.langfuse.public_key'),
                config('services.langfuse.secret_key'),
                config('services.langfuse.base_uri')
            );
        });

        $this->app->singleton(LangfuseProfiler::class, function ($app) {
            return new LangfuseProfiler($app->make(LangfuseClient::class));
        });
    }

    public function boot(): void
    {
        //
    }
}

Advanced Usage

Working with Multiple Generations

# Create multiple generations within a span
$generation1 = $trace->span('classification')->generation('generation-1')
    ->setModel('gpt-4.1-mini')
    ->setModelParameters(['temperature' => 0.5])
    ->setPrompt($prompt)
    ->setInput($messages)
    ->setOutput('RESULT_1');

$generation2 = $trace->span('classification')->generation('generation-2')
    ->setModel('gpt-4.1-nano')
    ->setModelParameters(['temperature' => 0])
    ->setPrompt($prompt)
    ->setInput($messages)
    ->setOutput('RESULT_2');