Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

GitHub - lmnr-ai/lmnr: Laminar - Open-source DataDog + PostHog for AI agents / RAG apps. Fast, reliable and insightful. Written in Rust 🦀. YC S24.

Sep 11, 2024 - news.bensbites.com
The article introduces Laminar, an open-source platform designed for observability, analytics, and prompt chains for complex LLM apps. The platform is likened to DataDog and PostHog for LLM apps, offering OpenTelemetry-based instrumentation, semantic events-based analytics, and a modern stack built for scale. Laminar hosts background job queues of LLM pipelines, turning their outputs into metrics. It is written in Rust, uses RabbitMQ for message queue, Postgres for data, and Clickhouse for analytics.

The article also provides a guide on getting started with Laminar, either through their managed platform or self-hosting with Docker compose. It explains how to instrument Python code using Laminar, including automatic instrumentation of LLM calls and a simple `@observe()` decorator for tracing inputs/outputs of functions. Users can also send events in two ways and create Laminar pipelines in the UI to manage chains of LLM calls. The article concludes by directing readers to client libraries and documentation for further learning.

Key takeaways:

  • Laminar is an open-source platform for observability, analytics, and prompt chains for complex LLM apps, comparable to DataDog and PostHog.
  • It offers OpenTelemetry-based instrumentation, semantic events-based analytics, and is built for scale with a modern stack including Rust, RabbitMQ, Postgres, and Clickhouse.
  • Laminar can be easily started with a free tier on their managed platform or self-hosted with Docker compose.
  • Laminar provides automatic instrumentation for LLM calls and a simple '@observe()' decorator for tracing inputs/outputs of functions. It also allows for the creation and management of Laminar pipelines in the UI.
View Full Article

Comments (0)

Be the first to comment!