Friday
Room 5
11:40 - 12:40
(UTC+01)
Talk (60 min)
Open-source Observability for Your LLM Applications... Tracing your chains!
Observability is a critical aspect of modern software engineering, especially for complex systems powered by LLMs. Simple monitoring solutions often fall short in providing the depth of insight required to fully understand and optimize these systems for production usage. The good news is that we already have an full industry standard (not just AI related) that simplifies and normalizes the integration of traces, logs and metrics into your monitoring stack - Opentelemetry!
This presentation will introduce participants to the concept of open-source observability and tailor it LLM applications, highlighting the advantages of leveraging the open-source ecosystem for comprehensive system insights.
We will delve into the openllmetry project, a set of extensions built on top of OpenTelemetry that gives you complete observability over your LLM application. Because it uses OpenTelemetry under the hood, it can be connected to your existing observability solutions that you already have at your monitoring stack - like for datadog, Honeycomb, and others.
Best practices for deployment, monitoring, and performance optimization will be shared, along with real-world case studies demonstrating the impact of effective observability on LLM applications, with sample trace chains built with LangChain and Redis a vector store.