Langfuse Go SDK

NOTE: This is a fork of the original repository at https://github.com/henomis/langfuse-go.

This is Langfuse's unofficial Go client, designed to enable you to use Langfuse's services easily from your own applications.

Langfuse

Langfuse traces, evals, prompt management and metrics to debug and improve your LLM application.

API support

Index Operations	Status
Trace	🟢
Generation	🟢
Span	🟢
Event	🟢
Score	🟢

Getting started

Installation

You can load langfuse-go into your project by using:

go get github.com/jwfriese/langfuse-go

Configuration

Just like the official Python SDK, these three environment variables will be used to configure the Langfuse client:

LANGFUSE_HOST: The host of the Langfuse service.
LANGFUSE_PUBLIC_KEY: Your public key for the Langfuse service.
LANGFUSE_SECRET_KEY: Your secret key for the Langfuse service.

Usage

Please refer to the examples folder to see how to use the SDK.

Here below a simple usage example:

package main

import (
	"context"

	"github.com/jwfriese/langfuse-go"
	"github.com/jwfriese/langfuse-go/model"
)

func main() {
	l := langfuse.New()

	err := l.Trace(&model.Trace{Name: "test-trace"})
	if err != nil {
		panic(err)
	}

	err = l.Span(&model.Span{Name: "test-span"})
	if err != nil {
		panic(err)
	}

	err = l.Generation(
		&model.Generation{
			Name:  "test-generation",
			Model: "gpt-3.5-turbo",
			ModelParameters: model.M{
				"maxTokens":   "1000",
				"temperature": "0.9",
			},
			Input: []model.M{
				{
					"role":    "system",
					"content": "You are a helpful assistant.",
				},
				{
					"role":    "user",
					"content": "Please generate a summary of the following documents \nThe engineering department defined the following OKR goals...\nThe marketing department defined the following OKR goals...",
				},
			},
			Metadata: model.M{
				"key": "value",
			},
		},
	)
	if err != nil {
		panic(err)
	}

	err = l.Event(
		&model.Event{
			Name: "test-event",
			Metadata: model.M{
				"key": "value",
			},
			Input: model.M{
				"key": "value",
			},
			Output: model.M{
				"key": "value",
			},
		},
	)
	if err != nil {
		panic(err)
	}

	err = l.GenerationEnd(
		&model.Generation{
			Output: model.M{
				"completion": "The Q3 OKRs contain goals for multiple teams...",
			},
		},
	)
	if err != nil {
		panic(err)
	}

	err = l.Score(
		&model.Score{
			Name:  "test-score",
			Value: 0.9,
		},
	)
	if err != nil {
		panic(err)
	}

	err = l.SpanEnd(&model.Span{})
	if err != nil {
		panic(err)
	}

	l.Flush(context.Background())

}

Who uses langfuse-go?

LinGoose Go framework for building awesome LLM apps