Langfuse

langfuse.com
На карте связей Открыть сайт

LLM engineering platform for observability and metrics

Описание

Open Source Agent Evals &Observability · Trace, evaluate, and improve AI agents with one open platform. Use production data to understand behavior, collaborate on fixes, and ship better quality at lower cost and latency. · Launch, observe, improve — repeat. · All the tools, oneintegrated platform. · Works with any stack. · Open platform. Open source.

Langfuse is an open-source LLM engineering platform designed to help developers build, monitor, and optimize large language model applications. It provides tools for observability, metrics tracking, prompt management, and evaluations, enabling teams to iterate on their LLM workflows efficiently. Users can self-host or use a cloud version, making it accessible for both startups and larger enterprises.

Возможности

Complete LLM Application Tracing
Failure Debugging & Inspection
Evaluation Dataset Generation
OpenTelemetry & Library Compatibility
Multi-language SDKs
Open Source & Self-Hosting
Drop-in LLM Integrations
Enterprise SSO & RBAC
Audit Logs & SCIM API
Industry Compliance & Certifications

Сценарии использования

LLM APPLICATION DEVELOPMENT, AI AGENT OPTIMIZATION, PERFORMANCE MONITORING, USER FEEDBACK LOOP, CONTINUOUS INTEGRATION

Частые вопросы

Langfuse provides built-in capabilities to measure and optimize crucial KPIs such as response times, token usage, and cost efficiency.

Langfuse offers straightforward configuration options for managing tracing in your application. Refer to the documentation for specific setup instructions based on your framework.

Langfuse supports deployment in serverless environments including AWS Lambda, Vercel, and Cloudflare Workers.

Langfuse allows you to create evaluation datasets and track scores over time, perform A/B tests between different agent versions, and collect user feedback directly through the platform.

You can retrieve experiment scores via the UI or through the API/SDK.

This is a common issue addressed in the troubleshooting documentation. If you encounter this, check your integration setup and ensure traces are being properly captured.

Langfuse provides tools to manage separate environments for development, staging, and production.

This is configurable in the administration settings for enhanced security.

You can invite team members to collaborate on your Langfuse projects through the administration panel.

Use the password recovery option on the login page.

API rate limits and usage quotas are documented in the API section of the FAQ.

This error is addressed in the troubleshooting section. Langfuse handles tens of thousands of events per minute while maintaining consistent low-latency responses (50-100ms on average).

Характеристики

Тип Агент
КатегорияИнфраструктура и MLOps / Мониторинг и управление логами
Цена есть бесплатный тариф (от $29/мес)
Платформа Веб + десктоп
Системы windows, docker, api, web
Для когоIndividual
Язык сайтаen
GitHublangfuse/langfuse
Рейтинг4.98 (48 отзывов)
Просмотры956 618
Запуск2024-09-04

Умеет

agents api workflow-automation realtime open-source self-hosted

Интеграции

GitHub openai slack Docker OpenTelemetry mem0

Платформы

api web windows docker

Теги

llm evaluation llm debugging application debugging llm tracing opentelemetry

Соцсети

X GitHub LinkedIn

Исходный код

langfuse/langfuse

Найден в источниках

aiagentsdirectory aiagentslist github-lists moge producthunt

Похожие в разделе «Инфраструктура и MLOps»

AWS
Облако Amazon: Bedrock, SageMaker и инфраструктура под большинство AI-сервисов.
HPE GreenLake
Комплексная edge-to-cloud платформа, предоставляющая гибкую ИТ-инфраструктуру и управление как услуга для гибридных сред.
Needle
The Cursor for Tools and Knowledge.
CrewAI
Automate complex tasks with AI agents working together.
Vessium
Automate complex business tasks by describing them in plain words.
whoop.com
чё-то я внезапно закоммитился на какое-то довольно большое количество выступлений, поэтому проанонсирую их и заодно поре

Предложить сайт в каталог

Пришлите ссылку — остальное мы выясним сами.

Мы рассмотрим, что вы прислали, и добавим в каталог, если подойдёт.